- Use Extract naming convention to name extract functions that fill vbo/ibo
- Separate extract functions into separate file (for clarity)
- Make simpler iter loops to avoid as much overhead as possible
- Separate loose elements looping functions to avoid iteration complexity
(unfortunately this makes the code more verbose).
- Some iter functions are threadable and tagged as such.
- Add multithreaded iteration for extract functions that supports them.
Some elements indices can be written by multiple thread at once and
since setting an element is only garantee to be atomic for single verts,
we cannot reliably use GPU_indexbuf_set_line_* in this case.
To avoid this, we create another iteration for the problematic IBOs
- Add Vcol support
- Add UV/Tangent support
- Add Orco support
- Add Paint Line Mask support
- Add Adjacency Lines support
- Add EditUV face/line/point/facedot support
- Add EditUV data flag support
- Add EditUV StretchAngle/Area support
- Add Facedots select index support
- Add Weight support
- Add Mesh Analysis support
This makes it possible to have each attrib use a contiguous portion of the
vertex buffer, making attribute filling much more easy and fast as this is
how they are store in blender Custom Data layers.
This allows to render only a subset of an index buffer.
This is nice as we can render each material surfaces individually and the
whole mesh with the same index buffer.
This was a read after free error. This only fix the undefined behavior.
The result is still not correct in certain cases (see T67226).
We want to include this for 2.80
Text datablocks should always have a 'single user' flag set, and they
usually do not have any user (since neither text editor itself, nor
Freestyle usage are text users - the second is odd btw...), the only one
am aware of is the script node (e.g. for OSL).
Add text case was simply not doing anything, so added.
Load text case was doing things in inversed logic (setting user count to
zero in BKE, then setting 'real user' flag in ED code). Made it the
other way around (BKE ID creation code should not care about usercount
usually, this is up to higher-level code to decide what to do
(operators, RNA...).
Note: tried to check all cases, but there might very well be some more
hidden bugs here...
This reverts commit b0b2546d4a.
Please do not sweept the dirt under the carpet like that! If there is a
bug, either report it or fix it, but do not silence it!
Previously when switching modes, the code didn't check if we were in the
correct view for the masking mode.
Reviewed By: Sergey
Differential Revision: http://developer.blender.org/D5288
This was caused by TAA offset being computed as the 2nd sample even if the
sampling was reset afterwards.
The fix is to update the matrices after any potential reset.
Selecting multiple armature objects and entering weight paint mode
wasn't working.
Entering weight + pose modes at once was only done when
object mode locking was enabled. Now it's done even when
mode-locking is off - because it's still inconvenient to setup
and useful default behavior.
Resolves issues raised by T66949
If the parent collection was out of view we were not taking its
properties into consideration. We need it even when not drawing the
parent to set active/inactive values for its children.
Related Task: T66948
Reviewers: brecht
Subscribers: Zachman
Differential Revision: https://developer.blender.org/D5272
Basically layer_collection_sync was calling BKE_base_eval_flags right away while
iterating over the bases.
However when a parent/sibling collection is to influence the collection flag of
an object that exists in more than one collection, it is too late since we
deselect the object in BKE_base_eval_flags right away.
Related to T64312.
Reviewers: sergey, brecht
Differential Revision: https://developer.blender.org/D5243
Based on feedback from @matiasmendio
There was a problem with the control points because it was very difficult to know what point move. Now the points are moved apart and makes easy to know what point use.
Like Blender renders without a Z channel. The single layer case assume that channel
names are just R/G/B/A without any layer name prefix, and would not read channels
like "Image.R".
Carefully tested for regressions with the openexr project tests images, so this
should be safe.
`layer_used` runtime data, which controls the drawing of dots in the UI was not getting refreshed properly.
This used to happen in the drawing code, but was no longer working for reasons explained in:
{rB2b09062defa093a243b5fe64b099accb07b440a3}
The solution was to update each layer manually in the operators:
* ARMATURE_OT_bone_primitive_add
* ARMATURE_OT_delete
* ARMATURE_OT_dissolve
* ARMATURE_OT_fill
* ARMATURE_OT_merge
* ARMATURE_OT_separate
* ARMATURE_OT_bone_layers
* POSE_OT_bone_layers
Differential Revision: https://developer.blender.org/D5281
When objects are not in the view layer, just return false rather than throwing
an error. As far as the script is concerned the object is not visible or
selected when it's not in the current view layer.
When ng mask layer operations (add, remove, move) the
original mask was not tagged to be updated resulting
in missing data on the copy. The Masking function
assumes that the copy and the original is always
structured the same.
Reviewed By: brecht, sergey
Differential Revision: https://developer.blender.org/D5283
It is possible that undo will change current scene frame and that was
not synchronizing current frame from scene to movie clip user.
Reported as a part of T66519.
Reviewers: brecht
Reviewed By: brecht
Maniphest Tasks: T66519
Differential Revision: https://developer.blender.org/D5280
`DRW_STATE_CLIP_PLANES` has to be enabled independent of the workbench material.
Reviewers: fclem, jbakker
Differential Revision: https://developer.blender.org/D5278
Applying/undoing incremental changes didn't fit well when
mixed with periodic snapshots from mem-file undo.
This moves to a much simpler undo system.
- Uses array storage with de-duplication from `BLI_array_store`.
- Loads the buffer into existing text data,
for better performance on large files.
- Has the advantage that Python operators can be supported
since we don't depend on hard coded undo operations.
Solves T67045, T66695, T65909.
If one of the objects had invalid selected edges, it would lead to a
crash since none of the for loops were checking for whether the edge
slide data is valid.
We could refactor the macros to create a new
FOREACH_TRANS_DATA_CONTAINER_WITH_DATA
However we are too close to 2.80 final release so we manually skip them
for now.
Note: TRANS_DATA_CONTAINER_FIRST_OK cannot be used either for the same
reason.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D5274
Based on feedback from T66675, the current interface is confusing,
leading to users accidentally overwriting their preferences.
Now, once factory settings are loaded preferences are not saved
and the save button is shown (as when auto-save is disabled).
Tool tips note that factory settings require manual saving
afterwards.
Also rename preference menu items for better consistency
(settings/state -> preferences).
Update the default init step values to be the same for all caches.
This is actually a small hack as these values are not used on the
creation of the first cache. But the default init value is 1, so this
will not be noticeable anymore.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D5271
We only had a very limited, specific handling of that in collection
duplication code, but this has to be handled at a much more general
level in Object copy code itself, since it makes no sense to duplicate
rigidbody object data without adding new copy to relevant rigidbody
collections...
WARNING: This is a fairly risky rework of rigidbody handling logic
when copying an Object data-block. It is *NOT* considered safe enough
for 2.80 release.
I tried to take into account copy flags to not mess with other IDs
(collections) when we are copying outside of Main, and also not do deg
tags when this is forbidden, but risk of something going wrong here is
too high...
When a mesh is pinned in the properties panel, Blender crashes when you
click the "Add Custom Split Normals Data".
The code calls `ob = ED_object_context(C)` which returns NULL when the
mesh is pinned in the properties panel, causing a segfault when trying
to get the mesh via `ob->data`.
A new function `ED_mesh_context(C)` avoids this by first checking
whether a mesh was pinned in the context. If not, it checks the pinned
object's data. If that's not there, or it's not a mesh, it returns the
active object's mesh. Finally it returns NULL if there is no active
object, or if the active object is not a mesh object.
Reviewed By: brecht, mont29
Differential Revision: https://developer.blender.org/D5223
Fixed: The Collada Exporter only supports export of
Lambert Shaders. But Shininess is not supported with
Lambert Shaders. The exporter must not add Shininess
to the Shader data!
Fixed: The Collada Importer adds an illegal value of -1
for reflectivity when this parameters is not defined in
the imported collada data. Now reflectivity is only
set when the import data contains a valid value.
Discarded: The Collada Importer handles shininess in a
dubious way. I have discarded import for now.
This needs to be reworked carefully in 2.81.
Differential Revision: https://developer.blender.org/D5262
working
Mouse values [for 'paint_last_stroke_update()'] gone missing in
rB4f616c93f7cb.
Thx @campbellbarton for hinting at convention to get mouse from sculpt
caches.
Reviewers: campbellbarton
Maniphest Tasks: T66742
Differential Revision: https://developer.blender.org/D5267
The dvert layer was not assigned to the mesh data if it had to be
created by the dpaint modifier.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D5263
Second completely unrelated issue reported in same task (tssst...),
caused by missing doversion code of some changes in recent-ish 2.80
development (tsst... again).
Proper fix for the issue (reverted rB8a5a8282ce48 was merely sweeping
the dirt under the carpet...).
That switch on ID types had not been updated since quiet some time it’d
seem... Using the ID_Type enum type now, so this should not happen again
(compiler will yell at you if you forget to add here any new ID type ;) ).
Also had to change a bit the code layout then, to deal with fake NLA ID
type...
Note that for now, Screen IDs remain excluded from the ID menu usage,
due to T67004.
Was a mistake in normals calculation: need to consider all grids for correct
average in the center of the face.
Reviewers: brecht
Reviewed By: brecht
Maniphest Tasks: T66712
Differential Revision: https://developer.blender.org/D5254
There are two aspects to the problem:
- Dependency graph update for compositor preview was missing
updates flush.
Apparently, update for new frame style of update will take
care of flushing, but not the update tagged style of update.
This goes to a legacy dependency graph and is to be changed
at some point, but not so close to the release.
- Movie clips were missing from the compositor dependency graph.
This fixes part of T66519.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5256
The code didn't check if we should initialize the transform auto IK parameters.
Cleaned up the code a bit too.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D5260
Images were being freed often because the tagging for when the image was last
used was missing.
Differential Revision: https://developer.blender.org/D5252
Particles can not be used with a destructive modifiers, so we can not
maker such configuration fully reliable.
Not sure this specific setup ever worked in 2.7x, maybe DM index was
somehow reset somewhere in particle system in older Blender version,
or maybe all of Blender version were crashing.
Anyway, seems to be very easy to avoid obvious index past the array
boundary in the mapping,
Reviewers: brecht, zeddb
Reviewed By: brecht
Maniphest Tasks: T66812
Differential Revision: https://developer.blender.org/D5257
Some values (not properly handled by do_versions?) could cause
'rna_backup_restore()' to throw an error resulting in the whole preview
generation not being saved to the .blend file.
I've checked folowwing demo files:
- race_spaceship.blend
- wasp_bot.blend
Here the offending seetings were:
- `UnitSettings` (`length_unit`, ...)
- `FFmpegSettings` (`ffmpeg_preset`, ...)
For now, these are now excluded in `exclude_props`
Reviewers: brecht
Differential Revision: https://developer.blender.org/D5255
'viewselected_exec()' [which uses evaluated view layer] would query
BASE_SELECTED flag, but it was not set on base_eval, but on its
base_orig.
reason for this is 'ed_object_select_pick()' would modify base flags
[through 'ED_object_base_select()'], but not tag scene
ID_RECALC_BASE_FLAGS.
thx @sergey for hinting.
Reviewers: sergey, brecht
Maniphest Tasks: T66668
Differential Revision: https://developer.blender.org/D5235
This was caused by loose pointers.
This diff takes care of clearing the fields of other bones before deleting the pchan.
Reviewers: brecht, sergey
Reviewed By: brecht, sergey
Differential Revision: https://developer.blender.org/D5258
When using opengl attributes such as gl_VertexID in a shader its
location is set to -1. This location was used without checking in an
attribute array. This fails big time when called from python. For
example `print(shader.format_calc())` made python crash immediately.
this fixes https://github.com/JacquesLucke/animation_nodes/issues/1141
Reviewed By: fclem, brecht
Differential Revision: https://developer.blender.org/D5231
There was an workaround implemented for specific the Intel HD Graphics
4000 GPU on Windows platform. Other GPU from the same series also need
this workaround.
We will enable the workaround for specific drivers.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5239
The issue was the the copy data function didn't copy the active canvas number.
So it would always be 0 and thus use the first canvas when trying to bake.
Also fix not copying unused type data (unused canvas/brush settings).
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D5220
When the slot of the material is empty, the loop to lock the materials tried to use the material but this was NULL.
Check if material is NULL before using it in other operators too.
We need to indicate a preferred size to avoid `RGN_FLAG_TOO_SMALL` the first time the region is created.
Differential revision: https://developer.blender.org/D5238
When a mesh datablock is pinned in the properties panel,
`context.object` is `None`. This in turn causes `obj.mode` to raise an
`AttributeError` exception as `None.mode` doesn't exist.
Since there is no (fast/simple) way to check whether the owning object
is in edit mode or not, the properties will be disabled. Not ideal, but
better than spewing an exception on every panel draw.
Reviewed By: campbellbarton, brecht
Differential Revision: https://developer.blender.org/D5237
Although the Auto-Run Python Scripts flag isn't used,
this is still a security risk since exclude paths list is.
Further this isn't what users would expect when loading a file &
only some preferences were being loaded so it's not useful.
The issue was caused by un-initialized local storage for volume
intersection hits which are supposed to be stored in per-thread
KernelGlobals.
Fix is to make thread_shader() be the same as thread_render() in
respect of KernelGlobals.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5230
It should be based on the mesh bounds before modifier stack evaluation, but
some modifiers were causing it to be recomputed. The flag to disable texture
space recomputation was not preserved through modifier evaluation.
Differential Revision: https://developer.blender.org/D5225
Seems the deform group index and deform vertices went out of sync somehow.
Added extra NULL pointer check, which seems to be safe and matches checks
in other places in the neighbourhood.
When `BKE_mesh_new_from_object()` cannot convert an object to a mesh, it
returns `NULL`. This case was not handled at all in
`BKE_mesh_new_from_object_to_bmain()` or `curvetomesh()`, causing a
segmentation fault.
This commit fixes the segmentation fault, and leaves the curve object as
a curve object.
Reviewed By: mont29, brecht, sergey
Differential Revision: https://developer.blender.org/D5217
Note: Although this fixes the issue (as in, it prevents a crash)
BKE_reports are not working because of CTX_wm_window_set().
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D5210
The menu shortcut context for 'Refresh All' was wrong and the actual operation
exectuted in the menu for the 'Move' command was different from the shortcut.
A mouse click in the window will trigger the `VIEW3D_OT_cursor3d` operator
before the viewport is available. This causes a segfault in
`GPU_viewport_engines_data_validate()`.
Other callers of `WM_draw_region_get_viewport()` already check for `NULL`
being returned and handle it gracefully.
Reviewed By: jbakker, fclem
Differential Revision: https://developer.blender.org/D5216
Init the scene of the draw context when selecting. When using face dot selection on
when the subsurf modifier is active on the cage, the scene needs to be
valid. It is read from the context in the
`DRW_mesh_batch_cache_create_requested` and used in the `isDisabled`
method of the SubSurfModifier.
Reviewers: fclem, sergey
Differential Revision: https://developer.blender.org/D5214
Restore some context overrides which were originally happening in 2.79.
Made it a bit more paranoid, by storing old values and restoring them
afterwards, so the context is not being changed at the end of the function.
When the object we iterate over is not part of the depsgraph, we cannot
get the evaluated copy to export. This workaround is temporary to avoid
a BLI_assert() failure getting the evaluated mesh of this object.
This will be handled more elegantly in the new AbstractHierarchyIterator
that I'm working on, but that requires a bigger change than we should
allow this close to the 2.80 release candidate.
This fixes a problem described in T58686.
This is intended for Python access and allows temporary negative frames
which is a problem when accessed from the UI.
This allows:
seq.frame_final_start, seq.frame_final_end = frame_range
However it allows invalid sequence strips,
using a label is fine since there is a duration property here which
can change the length.
Blender 2.8 features significant improvements in the creation of particles.
Removed hard limit and increased soft limit.
Patch by Gottfried Hofmann.
Differential Revision: https://developer.blender.org/D5120
The issue was caused by modifications to planar track tagging clip for
copy-on-write, which was invalidating its cache and forcing current
frame in 3D viewport to be re-load.
Ideal solution would be to share movie cache across original and
evaluated movie clips which will reduce memory usage. However, doing
such ownership changes so close to the code freeze is not something
comfortable to do.
When accessing evaluated objects, make sure access to an
evaluated dependency graph is done. This solves possible
access to NULL data on redo.
See https://developer.blender.org/D5209
I don't know if it was the intended behavior or not, but having brush
and canvas data at the same time with dymanic paint, would lead to the
object trying to act as a brush and a canvas at the same time.
We can't currently handle this with the new depsgraph, and it could
legitimately lead to bad feedback loops.
So now, to be more consistent with the GUI, I've made it only use the
current set type (brush or canvas) as the final type of the object.
That is, you can only have a object be a brush or a canvas, not both at
the same time.
The operator was using a non-evaluated depsgraph to get the evaluated
scene, which caused the crash.
This fixes the crash reported in T66605, but not the problem where
sometimes object origins aren't set.
We check if the previous iteration (sample) was using a valid double buffer.
If it wasn't, we request another iteration.
This fix the issue for viewport,viewport render and image render.
Related to T65761 Eevee render inconsistency between 3D View, Viewport render, and F12 Render
The AMD PRO driver on linux PROXY check also fails. Now the
configuration ATI/Unix/Official driver will also bypass the
Proxy test.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D5205
The `tselem->id` pointer can also be used for non-ID data (according to
this comment in DNA_outliner_types.h:
```
/* XXX We actually also store non-ID data in this pointer for identifying
* the TreeStoreElem for a TreeElement when rebuilding the tree. Ugly! */
```
As such, I don't mind adding a `NULL`-check in the
`is_object_data_in_editmode()` function. After all, when there is no
object, its data certainly is not in edit mode.
Allows to disable keyframes from movie clips in dopesheet.
Reviewers: brecht
Reviewed By: brecht
Subscribers: sebastian_k
Differential Revision: https://developer.blender.org/D5203
Don't update animdata after rendering scene
Rendering host scene from sequencer is not supported, removed code is unnecessary.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5199
Previously in 2.79 we were using a specialized drawing using derivedMesh.
Now the subsurf modifier tag each center vertex as facedot and let the
DRWManager pick it up.
Some modifiers (deforming ones) do not clear the tag so we can use this
technique even if there is deforming modifiers after subsurf modifiers.
Looks like that code was not updated when we switched to unicode, it was
still returning axtended ascii codes (iso-8859-15 ones I think)...
That was breaking some chars, which have a very different value in
unicode. Found while working on Text section of the Manual! ;)
Moved the caching code from direct calls in DNA to dependency graph.
In fact, not much was needed to be done apart form removing the direct
cache updates. The rest seemed to work fine.
Possible to avoid full sound file re-load, but doesn't seem this is
causing any issues.
The fill material has an alpha > 0, but the show fill is disabled, but as the fill tool is considered the fill as visible and the stroke is not drawn, the fill area had a gap.
Cell fracture redraw option was crashing.
Change operator popup behavior to close the UI before executing
since the operator may perform operations which change/remove the UI.
Before the primitives were closed only with cyclic flag, but this doesn't create geometry for the gap. Now, a new geometry is created using close stroke function to have geometry for sculpt or edit.
After testing the file, the problem was related to the fill material. As the fill material had the alpha channel set to 1.0, the fill triangulation was calculated, but the fill was disabled.
Now, the fill flag is checked in order to avoid fill triangulation update.
Following the advices of @Germano Cavalcante (mano-wii) , I have exposed as a workaround the free function to be called from draw manager for selection.
Now, the free function is not called for selection inside gpencil draw_scene, but it's called from draw_manager.c.
The real fix would be create a new Scene_finish callback in draw manager, but as the release of 2.80 is almost here, we fix this with a workaround that must be removed when new callback is in place.
Differential Revision: http://developer.blender.org/D5193
Preserving/rebuilding relashionships in recursive instancing was simply
not supported at all, code handling that was assuming a single level of
instancing.
This commit makes the following changes:
* Mixing DupliCollection and DupliVerts/Faces in the recursive chain is
now supported (by using a same GHash in all cases, differences of
persistent_id handling in hashing and comparison is now down inside the
relevant functions).
* When both "keep hierarchy" and "parent" options are enabled, code will
attempt to parent new objects to their version of instancer (instead of
parenting them systematically to the root object). This will preserve
the hierarchy much better.
* Collection is removed from dupli empties that have been made 'real'
(the duplication flag itself was already cleared, but the link to the
instantiated collection was kept).
Use more granular dependency graph traversal, which allows to ignore
dependencies which are not related on transform.
Reviewers: mano-wii, brecht
Differential Revision: https://developer.blender.org/D5184
For bone owners we want to do this in evaluated domain since
BKE_pose_where_is() / BKE_pose_where_is_bone() rely on (re)evaluating
parts of the scene and copying new evaluated stuff back to original.
Fixes T66080, T66397
Reviewers: sergey
Maniphest Tasks: T66080
Differential Revision: https://developer.blender.org/D5189
The issue was in the optimization code path for opaque shadow rays
which was wrongly considering all primitives in the node to have
same visibility flags.
When in texture paint mode and in solid mode the object that is being
texture painted will be rendered by the workbench engine with textures.
All other objects would render the same. For other cases the texture paint
draw engine will still draw the texture.
The texture mode draw engine now only drawn the masks. The opacity
sliders influences the texture mask.
This change has been implemented conserably. In the future we need to
look into making this better, like adding support that every object
can be colored differently. Currently when rendering in the workbench
we can have up to 3 different color types active (what the user selected,
the fallback in case no materials have been configured and this one,
forcing textures)
Reviewed By: fclem, brecht
Differential Revision: https://developer.blender.org/D5190
Changing the brush size for example,
was adding redundant undo steps in paint-modes.
For now, don't store undo steps for property changes
in paint & edit modes.
While not ideal, this is similar to 2.7x behavior.
We want users to go to the current version for their current version
when possible if not point to latest.
/dev should really only be for development related work. End users
should not be browsing /dev unless they are reading about upcoming
features ahead of time.
slow poll was altered in rBc252fe7a32ab after recent fix, now be a bit
more specific about meshes only.
Reviewers: brecht
Maniphest Tasks: T66433
Differential Revision: https://developer.blender.org/D5185
Now the simplify code works correctly in 3D space. Before it was trying
to project the strokes down into a local 2D space, but that didn't work
nicely for strokes with overhangs or big changes in the stroke
direction.
The code in question was simplified as well which lead to some nice code
reduction.
Reviewed By: Antonio Vazquez
Differential Revision: http://developer.blender.org/D5183
Make sure particle system edit never points to a modifier or particle system
which becomes inactive.
This is needed because copy-on-write will change pointers of them and those
pointers are supposed to be restored from particle system evaluation. But
since the particle system is disabled it never updates pointers.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5180
Now it is possible to enable MatCap lighting on top of textured models
in the workbench engine.
Reviewed By: fclem, brecht
Differential Revision: https://developer.blender.org/D5182
When doing image rendering with grease pencil, it reused the view of
workbench or EEVEE. These views might be offsetted due to TAA.
This shifted the view a tiny bit. We will not reset the view in between
render engines.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D5171
There were some keymap conflicts and inconsistencies here. After this change, it works as follows:
Snap toggle on/off: X
Snap pie: Shift-X
Viewpoint pie: V
This is now consistent in all editors and the 3d View conflict should be gone
'is_data'
since the move from 'cycles_shader_compat' (2.79) to
'PrincipledBSDFWrapper' (2.8) we lost the ability to (automatically) set
this colorspace setting for imported textures. This was useful for e.g.
normalmaps to always assume 'Non-Color' data.
This adds the possibility to overide and uses this for normalmaps...
Fixes T66368
Reviewers: mont29, brecht
Maniphest Tasks: T66368
Differential Revision: https://developer.blender.org/D5176
From fix for T66311, c6a199e254
Poll functions should never loop over entire scene,
in this case behave as other transform operators,
only checking the active screen.
This was meant to help users see which users are likely useful to add, but
it's not so obvious and there are cases where it's useful to re-append the
same workspace.
This fixes crashes, wrong names and inability to append workspaces in
edit mode. We now bypass the append operator so we can easily return
a datablock pointer and work in any mode.
could happen when used on multiple objects with multi edit, and skin
modifier was not present on all participating objects
Reviewers: brecht
Maniphest Tasks: T66311
Differential Revision: https://developer.blender.org/D5165
The issue was that the end point would be extrapolated and it would lead to
very high values if the curve had a near inf slope. Now we use the actual end
point value and only extrapolate values that are outside of the start and
endpoint range.
Differential Revision: https://developer.blender.org/D5151
The root of the issue comes to the fact that part of dependency graph
is being removed, without doing any further remapping.
This was happening because only materials used by objects were pulled
in, so when material mode is changed some material became unused and
removed from the dependency graph and freed, causing object or its
data to point to a freed memory in its materials array.
Simplest and safest way to solve this is to pull materials referenced
by both object and object data. This causes somewhat higher memory
usage but keeps evaluated state of scene in an always consistent state,
without any need to tag/update object's data on material mode change.
Don't think it is a problem in practice.
Reviewers: brecht, fclem
Reviewed By: brecht, fclem
Differential Revision: https://developer.blender.org/D5172
CDData checking on file load was not taking into account deprecated
CD_MTEXPOLY datatype, which unfortunately shows same weird glitch as
CD_PAINT_MASK and CD_FACEMAP ones...
Note that it was annoying (due to amount of warnings in console), but
totally harmless, since that data type is just deleted anyway.
This commit also generally cleans up the CD_MTEXPOLY deprecation code, we
have a system to handle that, let's use it, instead of defining local
static values to replace it...
Two issues here:
- Evaluated object data is to only be updated for selection only after modifier
stack is done its job. Otherwise it's possible to have selection batch update
called on an input data, at the same time as original object data is being
evaluated.
- If object's modifier stack did not create its own evaluated mesh (in case
when there is no effective modifiers, for example) can not update selection
on object's data, as it might cause threading issues between objects sharing
same data.
When we create the cage procedurally, we need to remove any edge split
modifiers. Since the new depsgraph in 2.80 we were removing the
modifiers straight from the evaluated object (it is a copy anyways).
On top of that we need to reset its eval data state (BKE_object_eval_reset)
to make sure the call to BKE_object_to_mesh to generate the cage would take the
new modifier stack state into account.
However doing so was freeing the low poly mesh we use later to convert
the normal space.
The solution (and this patch in fact ;) ) as suggested by Sergey Sharybin is to
use BKE_mesh_new_from_object() directly as well as force the modifiers to be
recalculated when any edge split modifier is removed.
The root cause seems to be an assumption in
[generate_video_frame()](https://developer.blender.org/diffusion/B/browse/master/source/blender/blenkernel/intern/writeffmpeg.c)
that we're always using 4 bytes per pixel. This is not true when using
QTRLE in RGB mode, because that uses the RGB24 pixel format (so 3 bytes
per pixel). Just updating the `linesize` property doesn't fix it though,
but just creates a crash somewhere else.
This at least fixes the crash by always forcing RGBA to be written, even
when the user selects RGB.
There are several aspects to the fix:
- Always calculate bounding box for meshes and curves from dependency
graph evaluation function.
There is a reason why mesh was tagged for geometry update, so can not
be spare here in attempts to avoid calculation.
- Copy bounding box and texture space to original object for active
dependency graph.
This matches object-level bounding box and allows to remove bounding
box evaluation from RNA.
It causes some false indication of a failed build on soma platforms
where it times out.
Windows and macOS needs extra work to properly support python paths
and path to the bundle.
This can now happen without poking git every time by doing local
modifications on a builder prior to commit.
As Usual...
Note that we have to edit each C-defined of those callbacks that set
Node->id pointer, instead of just moving the generic call to
`id_us_plus()` in `node_init()` below the call to
`ntype->initfunc_api()`, since that one manipulates RNA pointers, which
are supposed to already handle usercounts themselves on assignment.
The C callbacks are moving back to manipulating directly DNA, so it's
their responsability to handle user count then (Py code should do that
automatically through RNA assignement).
... instead of overiding it (previous behavior).
In practice it's not really noticeable.
This means an object with alpha will never be more opaque when enabling
xray.
Use full saturation (since some intensity is lost when blending),
Set blend back to 0.5, made color look dull which isn't good when
some of the colors are similar (faded orange/yellow).
The wireframe drawing for face masks is intrusive as selected wires
were solid white and always drawn. This made it hard for users to see
the exact color near edges.
This patch draws only the border of the selected faces,
edges between two selected faces are not drawn at all.
Reviewed By: brecht, fclem
Differential Revision: https://developer.blender.org/D5147
There are several aspects to the fix:
- Always calculate bounding box for meshes and curves from dependency
graph evaluation function.
There is a reason why mesh was tagged for geometry update, so can not
be spare here in attempts to avoid calculation.
- Remove texture space evaluation from RNA accessor.
Such data is to be evaluated by a dependency graph.
Don't see a reason to be different here: we never force evaluation of
any kind from RNA.
- Copy bounding box and texture space to original object for active
dependency graph.
This matches object-level bounding box and allows to remove bounding
box evaluation from RNA.
Before this change using Sequencer input for the scene strip would
permanently enable sound playing back for that strip. Going back to
Camera would have been still playing sound from the strip, which is
rather misleading since rest of the nested sequencer is not used.
The default keymap conflicted selecting by curves and channel.
Caused by 325b0ad2ed
Revert to original keys, use click instead of press
to avoid conflicting with box-select.
Now, when close a geometry with cyclic is possible generate new geometry for the gap.
The cyclic operator now supports multiframe edition too. Before only worked with active frame.
Also added the corresponding missing menu options and the new F keymap. All these features were missing, pending of the fix of the alpha glitches in stroke already done.
gflags emits a few unused variable warnings since the main
CMakeLists.txt raised the warning from w4 down to w3. This
restores it back to w4 in the remove_strict_flags macro.
ffmpeg defines some of the math constants if they are not
found before including any of its headers, this lead to
a build warnings about M_E, M_LN2 and M_SQRT1_2 being
redefined once BLI_math_base.h gets included.
With Eevee the user interface, 3D viewport and small icon materials previews
are rendered on the same GPU. This can lead to some choppy interaction when
dragging sliders. Delaying the icon preview render until the user is done with
that helps a bit, though it's no guarantee.
Before this change doing something like `--verbose 10 --debug-cycles`
did not properly set verbosity, only using those arguments in an other
way around was leading to a correct verbosity level.
This option from the very beginning of its existence needed more work
to make it work correct and this was never done.
This option was working fine during continuous playback, when there
are no skipped frames, but it was failing when AV-sync of framedrop
was enabled.
It was never working correct when jumping between frames, including
rendering on a farm which usually does frame-range based rendering.
With copy-on-write things became even more tricky, since the "stuck"
flag was never preserved between re-evaluations.
Fixes T65683: Sticky Option in Floor Constraint for Bones Not Working
This works for most situations, however if you have:
```
A
|-> B
|--> C
```
And only A and C are selected, C will be shown nested under A, instead
being by its side.
I still have to think on how to address these cases since they are
slightly misleading.
Related: T65263.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D5134
This makes things work more similar to 2.79 until there is a toolbar for
mask editing. Select & slide is now always on the left mouse button.
For the LMB keymap the context menu is on the right mouse button. For the
RMB keymap the select & move is on the right mouse button, which is almost
the same as select & slide but for historical reasons there is still a
distinction.
Differential Revision: https://developer.blender.org/D5070
Changing the workspace or mode from one window may need to change the
active tool in another window since two different workspaces
may share an object.
This wasn't working with multiple windows,
WM_toolsystem_do_msg_notify_tag_refresh could use a workspace
from a different window to the screen that owned the area.
Instead of fixing, remove these since they aren't needed anymore
since changing modes now refreshes the tool system.
* Mask Display menu is now part of the header
* Sidebar now has a separate Mask tab
* Add context menu when in mask mode
Differential Revision: https://developer.blender.org/D5102
Newer OpenSubdiv brings fixes and improvements for non-manifold meshes,
which fixes some crashes we've experienced in the recent past when using
Gregory patches.
Additionally, thing new version of OpenSubdiv brings sparse patches,
which allows to multi-thread topology refinement step.
The edit mesh is culled by object. When the object is not visible on
screen, the edit mesh may still be visible. This change will not cull the
edit mesh anymore.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D5144
Blender startup time and shader compilation is a big factor when running
hundreds of tests, so now all renders in the same ctest run in the same
process.
This was previously reverted due to skipping other tests when one test
crashed. Now if a test crashes, Blender is re-run with the remaining
tests so we get results from them still.
This time both full `make deps` and final compilation is tested on
a freshly installed CentOS 7.
The thing is: OpenImageIO is not configured to use an external PugiXML
library, so it was compiling its own.
At the same time the OpenShadingLanguage library was commanded to use
an externally compiled PugiXML. This caused some sort of discrepancy
which lead to Blender-link-time errors. Could be linking error, could
be namespace related, could be ABI related. In any case since we do
have PugiXML in the OpenImageIO already lets just stick to it.
It was originally needed for various migration needs, now CUDA
binaries are always to be built for 64 bit platforms and never
to be built on 32bit platforms.
Memory leaks happened when using final multi view rendering together with workbench.
Workbench assumed that the textures were always NULL
Reviewers: fclem
Differential Revision: https://developer.blender.org/D5136
Grease Pencil already reset the default view so the draw_manager wouldn't assert. During multi view rendering the same assert still happened. This patch will reset the default view when starting to render a new view.
Reviewers: fclem
Differential Revision: https://developer.blender.org/D5137
Following @fclem comments, I have implemented the use of Stencil for Solid line strokes. For Dots, Boxes and Lines with texture, the stencil is not activated because "kill" some artistic effects.
We have done test in greasepencil-object branch and all it's working as expected and the FPS are equal, so the stencil hasn't any appreciable impact in the drawing time.
The Stencil is used in groups of 255 and the passes are done in the same way. If the stencil is not use for the type of stroke (Dot/Box/Texture), the drawing is grouped as much as possible to reduce GPU overhead and limit the times the stencil bit must be cleared.
For doing this patch I had to add 2 new functions for reading private data to Draw manager. We decided add these function as a temporary solution while the Draw Manager implements the option to clear the stencil by groups. When this option will be implemented, these functions must be removed.
Thanks to Clément for his help and support. It's always a pleasure working with him.
Review by: @fclem
Testers: @mendio @pepeland
See D5126 for more details
This function is a workaround to use the stencil in grease pencil and reduce the number of clears.
Reviewed by @fclem.
Note: This function will be removed when draw manager supports stencil clearing by group.
Previously displacement relinking was trying to be smart and seems to be
broken in some cases. This fixes all cases by brute force.
We copy the whole branch linked to the displacement socket and tag it
accordingly. Then we only relink the nodes that are not tagged.
Moreover, we bypass bump nodes inside the dispacement trees so that the
resulting bump is the same as cycles.
Fix T66000 EEVEE: Unexpected results when displacement interact with nodegroups
We fix by separating the drawing of wire xray objects.
These wire objects gets drawn before normal wires and set the stencil to
0x0 just like the solid counterparts. Also a prepass is done to "dig"
through non-xray solid.
This reverts commit e061cb4437.
This function is problematic on some operating systems.
# Conflicts:
# source/blender/draw/intern/draw_manager.c
# source/blender/gpu/GPU_texture.h
# source/blender/gpu/intern/gpu_texture.c
Precompiled headers were sharing the PCH file between debug and
release builds which is 'bad'. Adding the configuration to the
path fixes the issue.
Reported on chat by @mano-wii
This only handles the issue after file load.
To make it an all-around solution we need to deal with
this in ed_object_select_pick, do_outliner_item_activate_tree_element(),
...
Note, the performance of the tests we run here is still bad since we have plenty
of panels around. But better than the crash.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D5116
Am not even sure that it is possible to use fstrings at all when UI
translation is required (that is, is a sensible, reasonable way that
does not make things even more complicated than they already are), but
one thing is certain, this won't be trivial to get it working, so
definitively not a job for now.
Instead just do not use fstrings for UI translatable strings.
Fix things to make test actually fail as expected (one cannot compare
functions to strings, so no more sorting for now).
Not sure how to actually fix the test though, not even sure test make
any sense anymore actually, with all those weirdo gizmos and tools
keymaps thingy...
- Add 'Navigation Buttons' preference, used for 2D views
(previously this couldn't be disabled).
- Add "Off" option for 3D view axis.
- Support minimal axis with navigation buttons.
Separate Images does not have any effect on Movie strips, so this removes this option from the Movie menu.
Author: Peter Fog
Differential Revision: https://developer.blender.org/D5130
The issue was that we were still working around a Xorg bug that has been solved since a very long time:
https://bugs.freedesktop.org/show_bug.cgi?id=22515
The second issue was that the global "confirm on release for mouse clicks" was used for keyboard shortcuts as well.
Reviewed By: Campbell
Differential Revision: http://developer.blender.org/D5128
Not much to be done here really, besides adding yet another hack to that
giant pile of hacks that are mballs...
So to avoid newly created copy of basis of mball to influence to mball
computation, we simply switch it to Empty object for the depsgraph
update run.
Not nice, but working (besides a weird change of basis obflag which
should not happen), and presumably reasonably safe change.
Previously settings were removed, now add to the rigid body world automatically
even if it's a bit ill defined, since this is confusing for users.
Fundamentally the concept of a rigid body world collection could be revised, and
left only as an optional thing.
clnor editing code was simply not checking at all whether it has
something to work on... Guess nobody had idea to edit custom normals on
a mesh that has no normals before! :P
This should probably be handled in a poll function too, to completely
disable those tools when there are no faces/loops, but let's keep it to
minimal changes at that point.
The previous commit disable the fast drawing if the background texture was not ready, but it did not detect the Painting mode, so the fast was always disabled.
Now the check is done inside paint mode.
There were some problems in the engine because the data was saved inside e_data struct, but this struct is reset sometimes and the background texture is not valid.
Now, the data has been moved to stl->g_data and all creation and free has been moved to use stl->g_data. This fix also some small memory leak for the Buffer GPUBatch data.
The background texture has been moved to texture list because must be available all the time. When is not drawing, the texture is removed to safe memory. Also, if the mode is painting and the texture is not ready because it was removed by Draw Manager, the texture is reloaded with the background image again. This ensure the background image is always visible when painting.
Also I have used this patch to reduce the size of texture used for background to 16F instead of 32F and the blank texture to 1x1 pixels instead of 16x16.
Reviewed by: @fclem
See D5115 for more details
* Follow Blender code style a bit more closely
* Fix mixed tabs and spaces
* Remove old README now that it's part of the script
* Make less tied to specific Blender version numbers
Pre-calculates values needed for unprojecting to avoid
a matrix invert and extracting projection matrix dimensions for
every call to GPU_matrix_unproject.
Use for gizmo selection drawing.
Using a capitalized app name fits the platform guidelines. Since macOS file
systems are case insensitive by default this should not break scripts that
assume lowercase.
Separate temporary windows, like Preferences, should be restored (if minimized) before bringing to top
Differential Revision: https://developer.blender.org/D5105
Reviewed by Brecht Van Lommel
This solution checks if `persp_mat` has changed. If positive, the selection buffer is redrawn.
Differential Revision: https://developer.blender.org/D5081
Switch to Gregory basis patches which are tangent continuous across their
boundaries.
Originally we've used BSpline basis patches to be more compatible with the
old subdivision code, but a lot of things changed anyway.
This reverts commit 36faf739a7.
Somewhat annoying but this change had some unforeseen consequences,
which lead to an actual bug.
Since this change was not sufficient to get original report fixed
is easier to simply revert for now.
Fixes T65842: Hair disappears when clicking on particle system name
- Move strip name to the top (easier to see which strip you are adjusting, and more consistent with other areas)
- Move Mute next to this, since it affects everything, both audio and video, and completely disables the clip
- Mute now greys out all the Strip panels, since none of them apply if the strip is disabled
- Rename Info to Time, since now this panel only includes timecodes and related controls
- Move Lock to the Time header, since you are locking the time controls
- Move clip resolution to Source, since it's not related to time
Migrate old legacy code to the draw mamager/object mode. The old legacy
version did not work with wireframe. By migrating the code
to modern draw manager code we have mode control on the drawing process.
Still background images do not work with OIT, the cause seems to be that the transparent pixels are treated as background pixels.
Also There are some artifacts when working with Holdouts and DoF, this
is because the draw engines do not pass the correct alpha values.
Reviewers: fclem, brecht
Differential Revision: https://developer.blender.org/D4638
The Pose Bone Selection used normal matric multiplication, but that
mismatched the Depth buffer from all draw engines. They used the
optimized matrices from common_view_lib.
This change will use the optimized version, so the depth buffer matches
and the render artifacts would be correct.
Please note that bone selection is not using shcfg and therefore render clipping is still off.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D5100
Was only toggling transform gizmos, which is too specific
in cases where users might want to remove gizmos from the view.
This also follows the top-level button which is most prominent in the UI.
Even though we are in UI freeze, we agreed that this should be better, and so we are changing a few things:
- Clearer separation of controls that affect the image transform vs the video
- New Transform panel houses Flip X/Y, Offset and Crop
- Flip X/Y now uses toggle buttons like we do for mirroring elsewhere (clearer + takes up less space)
- Video panel only includes things that relate to playback, ie Playback Direction, Strobe etc.
- Backwards/Forwards playback is now an enum rather than a toggle (we should always use enums when it's not an on/off switch)
- Rename Input panel to Source
- Just more immediately understandable and correct
- Move Deinterlace here since it's source file dependent
- Move Source panel to be a top level panel
- Merge Info and Timecodes panels
- Move Lock toggle to Info panel (was previously attached to name field which made no sense whatsoever)
- Name field now uses full width and doesn't add redundant text in front of it
- Re-arrange tabs to be Strip, Modifiers, Proxy & Cache, View
- Strip and Modifiers should be together
Reviewers: brecht, iss
Differential Revision: https://developer.blender.org/D5098
The other built in modifiers, except the generator modifier, seems to
update the depsgraph thought some RNA magic.
However the generator seem to be a bit special and doesn't get included
into this. Now we manually update the depsgraph on value changes to the
generator modifier.
The rotation matrix included the global object offset too. Now we only
take into account the actual offset that what sent to the function.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D5094
Also fixes bug where mesh with multires modifier is invisible in render
while in sculpt mode.
Differential Revision: https://developer.blender.org/D5099
Was noticeable when sequencer uses scene with a compositor.
The way it was using render API was forcing a single render layer.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5095
When performing F12 render on a transparent film all solid objects where
a bit transparent. Single Pass AA and no AA passes were not visible at
all.
Issue was that the first frame was not handled correctly making these
artifacts. This commit changes the order of `GPU_state_init` and
`DRW_state_reset` so the state is correct during the first pass.
The mask tools in the Tool Area of Clip Editor where never brought back,
after decision was made to postpone proper masking tools to 2.81.
This brings back the Mask Tool Panel in the Mask tab of Masking Tool Area.
This Diff also adds a better Context Menu, which respects whether you're
in Masking or Tracking mode. I have added the most important operators
there, with a focus on those that are otherwise harder to access.
Reviewers: sergey, billreynish
Subscribers: sebastian, brecht, hype
Tags: #motion_tracking
Differential Revision: https://developer.blender.org/D5075
When button text doesn't already show the label,
include the label in the tool-tip.
Without this the descriptions for icon-only buttons
don't always make sense.
This also gives a tool-tip for icon-only popovers.
This changes behavior from 2.7x, where selection & cursor could be
occluded by other objects.
Doing this without z-fighting in 2.8x isn't so simple because drawing
the text geometry is separated from edit-selection.
Change behavior since this doesn't seem like an important difference.
Fixes assert drawing text edit mode.
Initial trigger to this work was T65782, requesting faster autoscroll
when current folder contains thousands of items. That was a fairly
simple change, just needed to make scrolling steps variable based on
'distance' between current position and desired one.
But several other issues showed up while working on this, among the most
annoying ones, the scrolltimer could keep running forever in some cases,
failing to detect properly an 'end condition', we could even get some
'bouncing' in extreme corner cases, edited item was not always properly
visible in the end, etc.
So as usual with UI, this ended up in a frustrating equilibrium game of
finding the optimal solution among several tradeof, taking unexpected
large amount of time... At least new code seems to work OK in
all possible (reasonable) cases, that will do for now.
When compute preemption is available we schedule more work which is more
efficient. However the CUDA driver appears to be incorrectly reporting this as
unavailable, even though it should be supported starting with Windows 10 1803
and Pascal and Turing (10x0 and 20x0) graphics cards.
This reduces render time by about a 25% difference on our benchmark scenes. On
Linux compute preemption appears to be reported correctly.
An Alembic file saved by 3DS Max caused Blender to crash when importing.
Either the UV indices in the file are out of bounds or they are written
in a way we don't expect. In either case, this now no longer causes Blender
to crash.
When using RNA to alter the type of socket only the type was changed.
the typeinfo was not updated. Internally the File Output Node used RNA
to update the sockettype. making the socket invalid. When users save the
file and reopened the typeinfo was used. Also the color of the node was
determined via the typeinfo.
Another thing that happened was that the socket conversion was ignored
when empty node groups were present. The empty node groups were
optimized away before the needed data conversion was determined.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4938
Use depth buffer to order the uv edges correctly to always draw selected
edges on top.
We still use the double drawing workaround for points to keep the smooth
antialiased display.
Properly initialize custom data layers to default values when copying from
a mesh that does not have all the same layers.
Differential Revision: https://developer.blender.org/D5003
Percentage formatting of x% (used elsewhere) when using percentage display factor
Differential Revision: https://developer.blender.org/D5050
Reviewed by Brecht Van Lommel
Fixing small error in UI_GetThemeColorShadeAlpha4fv() clamping blue channel twice instead of alpha
Differential Revision: https://developer.blender.org/D5048
Reviewed by Brecht Van Lommel
Replace relative threshold for pseudo inverse in sharp remeshing modifier with
0.1 as proposed in the original paper.
Also change pseudo-inverse implementation that works with dynamic heap-allocated
matrix to static 3x3 version, for performance
Differential Revision: https://developer.blender.org/D5078
It was possible to have animation player step happening after manual
jump prior to the dependency graph evaluation.
Now we forbid changing scene frame if it was manually requested to
be changed.
Sometimes when the end caps pointed strait up (z axis) their faces would
become inverted.
It seems like the code tried to rely on a certain vertex order to always
happen. However this edge case did manage to produce an order that would
produce inverted cap faces. So now I introduce a normal check instead
so we can be certain that the face will have the normal direction we
want.
Reviewed By: Jacques Lucke
Differential Revision: http://developer.blender.org/D4138
Now, when you add an slot and then draw, the automatic created material will use the empty slot, instead to add a new slot and leave a empty one.
This fix only works if you add only an empty slot. We could remove any empty slot in the middle, but as this has no impact in the drawing or file, it does not worth the CPU time to review every time the material list. Anyway, the user can press delete button to remove any empty slot.
If we see this is a big problem in the future, we can reopen the bug and add this cleaning function.
argument into account
- if a rotation is given, always take it into account [which means
setting alignment to ALIGN_WORLD prior]
- caused by recent rB06fe2a5e0c5d
Reviewers: brecht
Maniphest Tasks: T65878
Differential Revision: https://developer.blender.org/D5085
The PBVHs raycast function calls `isect_ray_tri_epsilon_v3` with epsilon `0.1` which is inaccurate and may result in the problem presented in T65620.
The solution is to use `isect_ray_tri_watertight_v3` instead `isect_ray_tri_epsilon_v3`.
This can positively affect other areas as well.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D5083
Part of the issue was caused by missing speaker objects in the depsgraph
used for post-processing.
Remaining part was caused by missing scene sound update for this depsgraph.
This reverts commit a412f49e75 and 873c756e5d. The change was only
supposed to affect grease pencil, but also changed the cursor for sculpt mode,
where it's not clearly visible. Since this has not been quickly resolved I'm
reverting the commit.
Ref D5036.
Merge code will generate temp normal editing data for affected loops,
but since it will later (by setting some edges/faces to smooth) alter
and extend affected clnor spaces, it will also need temp normal editing
data for some other loops around those vertices...
Using those clnor editing data in that code is a bit of an abuse, but on
the other hand that struct stores exactly what we need.
So simply added an option to generate that editing data for all clnors
of affected vertices.
Without these buttons the functionality of the popover is
incomplete compared to the Graph Editor panel. To support
this the operators have to read the active F-Curve from
the context, instead of directly scanning animation data.
Expanding the context would also help Python operators.
- was using wrong offset [index instead of index * 4]
- also minor correction to variable naming
Reviewers: fclem
Differential Revision: https://developer.blender.org/D5082
The PYTHONPATH environment variable sometimes causes issues,
clearing it in the helper batch files is easier than talking
every user individually though the process of removing it
to see if it resolves the issue they are having.
This only replaces ICON_GREASEPENCIL with ICON_OUTLINER_DATA_GP_LAYER icon
Differential Revision: https://developer.blender.org/D5073
Reviewed by Dalai Felinto
This adds the icon intended for Grease Pencil layer data, in SVG but currently marked as blank
Differential Revision: https://developer.blender.org/D5073
Better to make internal code naming match official/UI naming to some
extent, this will reduce confusion in the future.
This is 'breaking' scripts and files that would use that feature, but
since it is not yet officially supported nor exposed in 2.80, as far
as that release is concerned, it is effectively
a 'no functional changes' commit.
Changes all GP layer icons to Pencil and highlights selected layer with background color.
Differential Revision: https://developer.blender.org/D5073
Reviewed by Dalai Felinto
This adds NA_SELECTED to notifier when selecting Grease Pencil layers so Properties Editor will update
Differential Revision: https://developer.blender.org/D5073
Reviewed by Dalai Felinto
- MPEG4/DivX has a maximum value of 65535 for the timebase denominator.
- MPEG1 and 2 have a list of supported frame rate ratios. These use
ratios like 24000/1001 and need those exact numbers.
This fixes an issue introduced in c5b1e7cd4e
where the correct ratio was passed to FFmpeg, but not with the identical
numbers FFmpeg has in a lookup table.
When exporting annotations using `bpy.ops.render.opengl` the annotations
where only correct in the first frame. In the second frame the
annotations was shifted by half the resolution.
This change will reset the GPU matrices when the annotations are
rendered.
Partially reverts commit 3b58bf3c7e
Green highlights aren't as bright as yellow/orange & were
reverted for the default theme, do the same for the light theme.
This patch does not bring functional changes, but it is a good change if we want to use these utilities in areas other than those using BMesh (eg painting editors).
This is also a step to replace `ED_view3d_select_id_validate`. That function erroneously checks `V3D_INVALID_BACKBUF` which causes it to update unnecessarily.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D5072
The issue was caused by a bad usage of GPUOffscreen.
The Framebuffer was created using a window framebuffer and used
in a viewport callback when another GPUContext was bound.
This change allows up to 3 framebuffers per GPUOffscreen.
Most common case will be using 2 framebuffers (one for init and
one for drawing) but in the case of more (bad usage) it will just
degrade performance a bit.
Always compute CD_ORCO undeformed coordinates now for rendering, same as before.
There is still a refresh issue to be fixed, when switching from solid to textured
mode in the viewport.
Computing such undeformed coordinates can be expensive and is not actually needed
if the mesh is only using e.g. UV maps. This was the same in 2.79, at least now we
are skipping the computation when there are no deforming mdifiers on the mesh.
This saves memory and evaluation time for simple static meshes with e.g. a
subdivision surface modifier. If no CD_ORCO layer exists then we assume the
actual vertex coordinates are equal to the original undeformed coordinates.
This was because the VAOs were not updated if an instance batch was
reusing a VBO containing instances attributes which was reinitialized.
Now we ensure the Batch will reconfigure the VAOs if the VBO is 0.
While user should never do that, it appears many end up using a 'view
layer' instancing collection as RBW collection, and even worse, have
objects in that unique collection.
Therefore, when removing RB simulation from an object, which among other
things has to remove it from the RBW collection, it would fully delete
the object from the blend file.
This fix merely checks the usercount of RB-removed object, and if it is
at 1 (which means object was in a single collection), it adds it to the
scene's master collection first.
There were various changes only applied to specific builtin templates even
though they should apply to all, since it's possible to do e.g. use grease
pencil objects in the general template or use mesh sculpting in the grease
pencil template.
Also, nearly all changes now apply to builtin templates only. This was
already the case for most, and the distinction seems to have been more by
accident than intent.
- The Start value and the Playhead value both used 'frame_start', which only tells you where the strip was originally inserted. Instead 'frame_final_start' is now used, which results in the correct timeline values.
- When scaling the sidebar some of the the labels weren't scaling correctly, this has been fixed
- Use Hold Cut/Cut instead of Hard/Soft Cut, for consistency
Patch by Peter Fog
Differential revision: https://developer.blender.org/D5065
This makes Shift and Ctrl work properly to extend and subtract selections.
This also moves Cut Links to Ctrl-RMB, which doesn't conflict with the selection tool.
Allows it to be preserved during copy-on-write update when on-geometry
related update is needed.
This is a required part for T63537, where we need to preserve the entire
evaluation data when object is tagged for only RECALC_COPY_ON_WRITE.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5023
Use lower level dependency graph evaluation which doesn't run any handlers,
doesn't touch sound system, doesn't inform image editors about changes.
Should probably move such evaluation steps to a helper function in DEG module,
but that is more like a cleanup.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5067
This was a design decision, but now we have decided to change it using the active material for the strokes using deleted material.
If the material slot is empty a new material is created to keep the strokes visible.
The issue was that the valid color modes was checked on the old image
format, not the new one. So if you switched formats it would not
correctly check if the settings were valid.
When they are occluded or when the snap is done for the generated meshes vertices, it was inconvenient.
An ideal solution needs to be discussed, but for now, for vertices, keep the behavior similar to the pre 2.8 versions.
This reverts commits
dcec863b2f,
f11929c145.
It wasn't possible to set the cursor without accidentally selecting a
column of keys & (vice-versa).
This conflicts with Alt->LMB to select column of keys.
Shift-RMB is already bound to scrubbing.
The behavior for loading factory settings wasn't clear for users.
This commit changes the behavior:
- Loading factory settings always disables auto-save
for the current session.
- The internal setting to skip saving on exit is now exposed
in the preferences (when enabled).
- The menu item "Load Factory Settings (Temporary)" has been removed
since it's always temporary.
This way users can always reset factory settings without
having to consider the combination of options that might cause their
preferences to be overwritten at exit.
If they want to enable auto-save for the current session
this can be done from the preferences.
Cheap tip: anything that is not "Camel Case" and/or that is more than
a few words long should use `TIP_` translation, not `IFACE_` one.
Also added several missing strings (including the one reported in D5056
by Jean First (@robbott), thanks).
Collections are a tad annoying with all their caching of objects... When
we modify content of a children collection, we need to tag DeG for CoW
update of all of the ancestors.
For now keeping that recursive tagging helper private, but would not be
surprised if we found more similar cases and needed to expose it to more
code...
`IFACE_` is for short strings always shown in UI (like labels of buttons,
menu entries...). Every thing else, especially when more than a couple
of words, must use `TIP_`.
Note that there are probably many other similar cases... This code is
really legacy, should use library_query helpers and other modern
BKE_library code instead of doing its own dirty cooking...
This was due to a double offset of the wireframe. We also reduce
the wireframe offset. The look of the wireframe overlay changes
a little with on distant wires.
FFmpeg uses a fraction of integers to indicate the frame rate, whereas
Blender uses `int / float`. When a custom frame rate is used with
non-integer base, the FPS and Base settings were multiplied with 100000
before passing to FFmpeg as `int`. This could overflow when a high
enough FPS setting was used, which is the case when importing a video of
almost-but-not-quite-integer frame rate into the VSE. The overflow
caused FFmpeg to return an error "The encoder timebase is not set",
which is rather cryptic for users.
The new solution is to take the max int and divide that by the frame
rate, and use that ratio to pass to FFmpeg. This won't overflow, and
thus allows exporting arbitrary frame rates.
Was very easy to reproduce by rendering sequencer with sound strip.
Need to use evaluated scene to open movie handle, since that is the only
scene which has proper sound handle with everything else attached to it.
A lot of areas were querying sound information directly using audio handle
which does not exist on an original sound IDs.
This change basically makes it so it's possible to query information about
given sound ID, without worrying about whether it's loaded or not: if it is
needed to load it first it happens automatically (no automatically-opened
handles are left behind though).
While this seems a bit extreme to open files on such queries it is still
better than the old situation when all sound handles were opened on file
load, no matter if it's needed or not. Besides, none of the changed code
paths are performance critical, just handful of tools.
Fixes T65696: Sequencer fails to create a new sound sequence strip via Python
Fixes T65656: Audio strip - SHIFT K crashes Blender
Reviewers: brecht
Reviewed By: brecht
Subscribers: ISS
Maniphest Tasks: T65696, T65656
Differential Revision: https://developer.blender.org/D5061
Apparently the `rna_Armature_editbone_transform_update` function was incomplete because it didn't copy all mirrored transform values.
I also noticed that the same logic seen in `rna_Armature_editbone_transform_update` is also seen in `ED_armature_edit_transform_mirror_update`.
So the solution is expose and use that logic that updates a mirrored bone. Thus deduplicating and fixing T65671.
Reviewers: brecht, zeddb
Differential Revision: https://developer.blender.org/D5058
It could be argued this was correct behavior, since auto-save
defaults to 'on' nevertheless, auto-saving settings once
the user has disabled auto-save can lead to accidents.
Don't reset the preferences flag when resetting preferences.
Improvements to behavior for gizmo tool-tips.
- 2D gizmos no longer cancel tool-tips on cursor motion
(matching the behavior of UI widgets).
- 3D gizmos still close on motion since 3D gizmos may have a large
on-screen area which would cause them to stay visible even after the
cursor has been moved a large distance. The motion threshold is used
so they don't close on unintended cursor motion.
- Changing highlighted gizmo now cancels the tool-tip & resets the timer.
This code now expects to wrok from fully evaluated data, however when we
keep original, we are actually working from data just copied from orig
one.
Ideally, we'd do a single depsgraph update/eval *after* we have created
all new required data, but that is tricky to do properly in that code
without risking breaking one thing or another.
So for now, just going for the simple, if not optimal solution, and just
repeatedly re-evaluating whole deg every time we duplicate an object to
be converted. Yep, dummy, but simple and... safe. ;)
Redo panel would be hidden (when 'keep original' was not set), due to
same kind of (un)selected issue as in T65301 (see previous commit).
Further more, not all MBall objects of the family were properly removed.
We need to properly select new objects (and deselect 'source' ones) when
converting to another type while keeping original ones. Otherwise poll
check of the operator fails, and redo panel cannot be shown.
Note that this is actually a design flaw in redo system currently, since
*new* state has to still allow last operator to be ran, when it should
actually be previous step in history that matters here...
How to reproduce: use 1-10 to change the visible collection. If the
collection was globally invisible, it would be set to globally visible.
This was a left over from the previous collection visibility design.
Now that we have a more clear separation between temporary visibility
(i.e., layer collection visibiilty) and a global visibility setting
(i.e., collection visibility) we should keep them separated.
Now texture storage of images is defined by the alpha mode of the image. The
downside of this is that there can be artifacts near alpha edges where pixels
with zero alpha bleed in. It also adds more code complexity since image textures
are no longer all stored the same way.
This changes allows us to keep using sRGB texture formats, which have edge
darkening when stored with premultiplied alpha. Game engines seems to generally
do the same thing, and we want to be compatible with them.
This new cursor is used instead of the ARROW because it was too disruptive while you are drawing.
The change affects all paint modes that are used Brushes.
See D5036 for details.
Reviewers: @brecht @billreynish @mendio
Cursor designed by: @billreynish
The OpenGL state was not set to blender defaults when using
`DRW_render_to_image` path. This is only used when doing F12-rendering.
Cause of changes with the `RESTART_INDEX` hair rendering was rendering
the restart_index as an actual vertex index.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D5039
This option can not be supported by a new granular dependency graph,
and, especially, copy-on-write.
It was always doing full update ever since initial commit of new dependency
graph which we are using here in the studio for the past years and lack of
this option was never brought up.
Fixes T65557: Delay refresh option in armatures is broken
Consists of few simple steps.
Don't tag object data's bounding box as dirty before re-evaluaiton
Most of the time this actually tags evaluated mesh which will be tossed
away few lines below anyway. And this causes issues if the evaluated
mesh is actually owned by the mesh datablock.
The bounding box now has clear separation between original object data
and object data after modifiers, so this should not be causing any
issues.
Free evaluated mesh when owner mesh changes
Ensures that evaluated mesh shares the same settings as its owner when
updates related on animation system happens.
Depsgraph: Update mesh when its geometry settings changes
Allows to have options like autosmooth animated.
Reviewers: brecht, mont29
Differential Revision: https://developer.blender.org/D5030
Most of the time this actually tags evaluated mesh which will be tossed
away few lines below anyway. And this causes issues if the evaluated
mesh is actually owned by the mesh datablock.
The bounding box now has clear separation between original object data
and object data after modifiers, so this should not be causing any
issues.
This commit fixes crash but the possible 3D sound is still missing.
Also, sound from sequencer coming from a scene strip is also missing.
This is partially a regression, at least for F12 case. The editing
was already behaving wrong: it was needed to have scene strip to be
actually rendered in preview.
- Right click select always activates on press
since this was added as a workaround to left click select conflict.
- Left click has the option to set this to drag or press.
Make this change based on feedback on 8778dd0c8b.
Add invalidation to strip add functions
Add invalidation for scene and movieclip strips
Skip invalidation for sound strips
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4989
3D engine settings were used to select image scaling method.
Use higher quality scaling only for rendering.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4995
We need to find a better solution for mixing normal refreshes in some corner
cases. We actually had similar issues in 2.79 anyway, not crashing is more
important.
Modulo makes no sense here at all, is from an old bugfix to prevent a crash
that is no longer an issue.
Differential Revision: https://developer.blender.org/D4727
When rendering the Sequencer using viewport rendering the annotations
were not scaled to respect the scene render size. This was because the
matrices were calculated based on the output size and not the actual
rendersize.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5032
Ninja was unable to see the dependency between the cpp
that generated the pch and the compile units that used
it. Explicitly managing this now makes precompiled headers
work with both msvc and clang, with both msbuild and ninja
based generators.
Didn't touch rna to make sure that the Python API does not break.
Maybe I missed some, but these don't seem to be used anywere internally.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D5034
Use an additional pose bone flag so we can keep track of mirrored bones that should be autokeyframed.
Reviewed By: Brecht
Differential Revision: https://developer.blender.org/D5033
This is a mix of solutions rBe60b18d51d58 and rB52af5fa31fbc.
What happened was that when a node of the BVH gets 0 vertices, the batch is untouched and therefore still drawn.
Ninja has issues detecting the implicit dependency on the
precompiled header output for freestyle. Disabled ninja
support for now until a proper solution can be found.
This is enabled with bl_use_eevee_viewport = True. It allows external engines
to generate an Cycles/Eevee shader node setup to emulate their materials in
the realtime viewport, or to leave it to users to do manually.
Removed bl_use_exclude_layers and bl_use_shading_nodes that did nothing
anymore. This should not break API compatibility, any scripts setting those
should continue to work the same as before.
Also adds descriptions for some RenderEngine settings.
Currently only obvious setting which can be animated is Smooth.
The rest requires more proper support from animation update on
the Curve datablock.
But at least with this change it's not a "dependency graph fault"
While fix in dependency graph for those is trivial, adopting underlying
code to become aware of animated settings is quite a rabbit hole: there
are non-obvious hidden inter-dependencies between settings, which are
especially tricky for NURBS.
For until we have more dedicated time for this disabling animation.
There is no obvious threading-unsafe code in the localization.
The main source of issues were the new_node/new_socket pointers
which are no longer used during node tree duplication.
Respect do_time flag in on_visible_update, matching behavior of old
dependency graph and avoids unwanted animation updates.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5026
- Proper support for dragging in empty areas using tweak events to either move or box select
- Avoids transforming immediately on press, same as the built-in keymap
When using python offscreen rendering the background was not drawn in
EEVEE. Users would expect that when calling the `draw_view3d` method
that it shows the same as in the 3d viewport.
Seems to be a difference between the meaning of draw_background in b279
and b280. In B279 the world background would be drawn. In B280 the
transparency is changed.
Reviewed By: campbellbarton, fclem
Maniphest Tasks: T61768
Differential Revision: https://developer.blender.org/D5022
Every time you build cmake it unpacks numpy. It is seemingly
very unhappy with OUTPUTS being directories rather than files.
The creating of the directory is not needed, so that step was easy
to resolve, next problem: was there was not a target that outputs
the tarball so it too was deemed out of date. Losing that dependency
as well, fixes the issue, while this is not great if we ever update
the libs in svn, we are planning to drastically change the way we
package python for windows so this is ok for now, but needs to be
addressed with the next python update, T65547 is the tracking ticket
for this.
Tested with VS2015/2017/2019 with both msbuild and ninja generators
This brings down the build time for freestyle with MSVC from a
minute to 10-20 seconds.
vs2019 bf_freestyle debug before: 60464 ms after: 11028 ms
vs2019 bf_freestyle release before: 56984 ms after: 20526 ms
Differential Revision: https://developer.blender.org/D2606
Reviewed By: brecht , sergey
In the Intel HD 4000 driver a shader has to be deleted in the same context in which it is created.
However, because you can't use a rendering context on different threads, to maintain the multithreaded compilation, the solution was to use the `GL_ARB_get_program_binary` and copy the binary generated for the shader and generate a shader on the main context using that binary.
This solution is limited only to Intel HD 4000 and windows.
Reviewers: fclem
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D5019
This includes refactoring of the `mouse_action_keys` to
make it easier to just detect if there is a key under
the mouse. The refactoring mostly consists of extracting
methods and reducing vertical scope of variables.
Reviewers: billreynish, brecht
Differential Revision: https://developer.blender.org/D5024
make.bat now supports optional parameters to restrict
the formatting to a specific folder. Multiple paths
may be given
example:
make.bat format source/blender/blenkernel source/blender/gpu
Depth testing was off as it used the precomputed ModelView matrix. As
draw engines currently use a different approach the depth was sometimes
a bit off making the color disappear.
This change will use a different vertex shader that will write the
correct depth. I expected the same change to be needed in the bone
selection overlay but was not able to reproduce it.
Reviewed By: fclem
Maniphest Tasks: T64615
Differential Revision: https://developer.blender.org/D5006
When using the world space cavity shader together with viewport or image
rendering only a single pattern was used. This was that the iteration of
the cavity shader was updated when the cache is initialized.
Now the cavity iteration is updated together when the TAA samples are
updated.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D5021
This change makes it so sound handles are created for evaluated scene,
sequencer and speakers. This allows to have properly evaluated animation
on them.
For the viewport playback sound uses regular dependency graph.
For the final render sound uses dependency graph created for render pipeline,
which now also contains sequencer and sound datablocks.
All the direct sound update calls are replaced with corresponding dependency
graph recalc tag.
For some cases like the edit mode context menu we don't currently have a fixed
title since it's dynamic. This should be improved, but we should at least still
show Call Menu if there is no menu title.
- added import of transparency and emission into principled BSDF Shader
- added support for importing all default collada material parameters
* diffuse
* emission
* index_of_refraction
* shininess (mapped to BSDF Roughness)
* reflectivity (mapped to BSDF Metallic)
* transparency + transparent mapped to BSDF Alpha)
* ambient (creates unconnected texture node)
* specular (creates unconnected texture node)
* reflective(creates unconnected texture node)
- added support for exporting collada material parameters:
* diffuse
* emission
* index_of_refraction
* shininess (mapped to BSDF Roughness)
* reflectivity (mapped to BSDF Metallic)
* transparency + transparent mapped to BSDF Alpha)
- prepared support for exporting the following parameters
but currently commented out:
* ambient (creates unconnected texture node)
* specular (creates unconnected texture node)
* reflective(creates unconnected texture node)
Problem: For now we only allow export of principled BSDF based
materials. I am not sure from where to get ambient, specular
and reflective as those values are not included in the
principled BSDF Shader (wip).
Some old Win32 files could have the brushes damaged and this produces a segment fault. Now, if the brush is damaged, a new set of brushes is created.
Related to T61413
If the layer collection had a parent that was disabled we wouldn't draw
it altogether.
Note: In this case I'm moving any enabled collection with a disabled
parent up in the hierarchy.
All in all I'm happy with this but I would like us to tackle the 1-20
shortcut, Ctrl+H menu and Collections viewport panel to do something
similar.
Nothing special to mention about regression test itself, it basically
mimics the one for `BLI_task_parallel_mempool()`...
Basic performances benchmarks do not tell us much, besides the fact that
for very light processing of listbase, even with 100k items,
single-thread remains an order of magnitude faster than threaded code.
Synchronization is just way too expensive in that case with current
code. This should be partially solvable with much bigger (and
configurable) chunk sizes though (current ones are just ridiculous
for such cases ;) )...
Function would not clear the static scheduler pointer, which lead to
crash (mem use after free) when trying to re-init and use the task API
again. Should not happen in Blender itself, but could in other cases
(like some future gtests ;) ).
The constraint icons are now no longer shown in the Outliner again.
This needs a more robust implementation that also works for bones & pose mode.
The new constraint icons are still shown in the Constraint Properties.
Each constraint now has a unique icon, just like modifiers.
Icons are designed by Andrzej Ambroż & Alessio Monti, with user input.
Patch by Alessio Monti.
Regression introduced by rB7fe3d1e7d718 (fixing T57934).
This effectively reverts rBrB7fe3d1e7d718, since changing the type of an
object is a very bad idea in general, and would need a careful and
complex check of all of its usages (many object usages assume a specific
type of object...).
Instead, we simply remove duplicollection on those objects, and give a
warning in case dupligroup was effectively used (in many reported cases,
dupligroup was a 'mistake setting', without actually instancing
aniything).
Note that the otehr idea to fix that versioning issue (to create a new
empty object for the instancing) is much less easy than it might look
(one would need to take into account potential animations, relations
between objects, etc.). Doable probably, but way overkill for a
corner-case 'bad' usage of the feature in the first place.
For offscreen rendering a high definition color buffer is needed.
Without it there are banding issues when doing multi-sampling viewport
rendering.
Reviewed By: fclem
Maniphest Tasks: T65287
Differential Revision: https://developer.blender.org/D5009
The image information for multilayer openexr files are only shown on the
first draw. In the second draw the render result property is filled of
the image, making blender think the image is a render result and it want
to display information about the renderpasses. Which are not present.
It could be that in the first draw the image data is not requested as
the buttons are drawn, before the main area
This change will display both information. We should investigate how to
actually detect if this is a render result or not as similar code is
present in the `node_shader_buts_tex_environment_ex`.
Reviewed By: brecht
Maniphest Tasks: T65345
Differential Revision: https://developer.blender.org/D4987
Now, the file does not have to be reloaded in all cases.
Instead, just scripts are enabled and all depsgraphs freed.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D5000
This is how it worked in 2.79 and it is how it is expected to be working.
Avoids unintended icons update during animation playback.
Fixes T64318: Update of material icons during animation 2x performance penalty
There is a keymap conflict with ctrl+tweak.
Therefore, I did not include this yet.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4999
This is probably just one of the related issues.
Root of the problem was that compositor job was using original scene and node
tree for compositing. It is not guaranteed to have all the evaluated data.
Switched compositor job to use it's own render-pipeline-like dependency graph
which has everything evaluated in it.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4998
The root of the problem goes to the fact that node tree copying
uses source tree and nodes for a temporary storage.
This makes it so multiple dependency graphs can not be reliably
evaluated from different threads if they are using same original
node tree.
Solved by doing the following:
- Commonly used tree copying function (which is used by library
manager) keeps source tree, nodes and sockets untouched.
- All the related areas (like node tree's callback) now have
const qualifier on the input.
- Areas which needs to have those temporary pointers assigned are
now using explicit function.
Would be really cool to get rid of those temporary pointers
completely, but this is a bit tricky due to hairy nature of the
code. Can happen any time now though: is easy enough to generalize
the new pointers mapping.
Note that this change is only intended to solve the crash.
The fact that icons shouldn't be updated on playback will be fixed
as a separate change.
Reviewers: brecht, fclem
Reviewed By: brecht, fclem
Subscribers: brecht, fclem
Differential Revision: https://developer.blender.org/D5002
We cannot do refcount operations in a non-Main ID, this is forbidden.
While that whole func could probably use some love and refactor, for now
sticking to minimal changes and just moving refcounting op after mesh
has been transferred to Main database.
Restart index can have been changed in another context and the static
var can get out of sync. A better solution is to set the restart index
when binding the VAO. It also have less perf impact.
Fix T65364 Corrupted mesh display on macOS
To make this work, other changes were necessary:
* To select a specific marker you have to click more exactly on the icon.
* Moving markers with click-drag only works when starting on a marker.
Additionally this patch implements that all markers are deselected,
when the user clicks in an empty area.
Reviewers: billreynish, brecht
Differential Revision: https://developer.blender.org/D4975
Material setup for quick explode was assuming a shader with a "BSDF"
output socket connected to the material output node whereas other socket
names are valid ("Shader", "Emission", "BSSRDF", "Holdout"...)
1.) The Blender order of applying transforms is:
Scale
Rotation
Transformation
Reasoning: This order ensures there is no shearing, which happens
when you do scaling after rotation, see also:
https://blender.stackexchange.com/questions/1806
The Collada exporter now exports in the order how the transforms
need to be applied upon import.
2.) Also removed obsolete #if 0 lines
Currently we can not export Decompsed Transforms in combination with
Armature asnimations. As a temporary workaround enforce export
of transformations as Matrix for armature objects.
This commit improves Blender's AppStream metadata with minor punctuation and grammar improvements to the description, and a new `<release>` tag for the current release.
The `<release>` tag benefits Blender in the following ways:
- Flathub users get to see the actual release number instead of "stable"
- Make it easy for all packagers to include changelogs, so Blender's users gets to see what's new in 2.79
- Blender gets featured in GNOME software to reward it for having Release information in its AppStream metadata (KDE Discover [[ https://bugs.kde.org/show_bug.cgi?id=389509 | plans do do the same thing ]])
If the prior objection to adding release data to the AppStream metadata was the ongoing maintenance burden, I volunteer to take it on for subsequent Blender releases.
Author: @ngraham
Differential Revision: https://developer.blender.org/D3065
On GPUs that support it, we now present OpenGL contents via CAMetalLayer. This
fixes frame skipping issues found in T60043. If the system does not have a Metal
capable GPU, NSOpenGLView will continue to be used.
Patch by Tomoaki Kawada, with some changes by Brecht Van Lommel.
Differential Revision: https://developer.blender.org/D4619
This is in preparation of an the upcoming fix where we need to use a Metal
layer to avoid performance issue when drawing with OpenGL. Note that we already
only officially support 10.12+, the difference with this change is that Blender
will not start at all on 10.9 and 10.10.
This caused the file browser to open in c:\windows with the fonts
folder selected instead of opening c:\windows\fonts\ and listing
the fonts.
Reported on chat by @blendify
Previously, bright edges (e.g. caused by rim lighting) would sometimes get
halos around them after denoising.
This change introduces a log(1+x) highlight compression step that is performed
before denoising and reversed afterwards. That way, the denoising algorithm
itself operates in the compressed space and therefore bright edges cause less
numerical issues.
The fix itself is by Germano Cavalcante (mano-wii).
But since I was investigating this with him, I'm including here an
assert in EDBM_select_id_bm_elem_get to help catching this sooner
in the future
Bringing the same logic we do in the outliner restrict column callback
and the menu call.
Also removing the "change depsgraph" logic there. Isolate collections
should not affect depsgraph relations (if it does it is to be tackled
separately anyways).
Users could change the master collection flags, but they should not.
That would not effectively affect the master collection objects
(depsgraph flag evaluation ignores master collection flags).
However we use the layer collection flags of the parent collection when creating
a new child collection.
We *could* solve this differently by creating a new RNA type for the
master collection (and layer collection) and hook this with rna refine.
But this patch seems to work well enough and it is simpler.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4981
It may be good to move the normals update out of the drawing code. But it was
already there for the non-multires sculpt cases, and does not have an obvious
place since we bypass the depsgraph and want to avoid the cost of updating the
normals multiple times when multiple events are handled before a redraw.
Centralize logic for when to use the PBVH for drawing, fix missing tests in
mask drawing, fix missing tests for multiple windows, only do more expensive
update for all viewports at end of the stroke.
Was missing re-set evaluated mesh to NULL.
Fix T65302: Crash after selecting one of multiple materials
Fix T65346: Blender crashes when selecting material in editing mode and LookDev view
Fix T65344: Blender crashes in edit mode when changing object
Object could and polygon count should ignore multiple instances of
the same data, yet count it at least once even if linked.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4986
Make sure object is properly evaluated to the new state when
it is expected to.
Reviewers: brecht, mont29
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4985
The draw manager used to determine if the view transform should be
applied by checking if the scene was not rendered to an offscreen image.
As the sequencer and texture painting needs to render to an offscreen
image with the view transform applied we need to separate the
`do_color_management` from the `is_image_render`.
Reviewed By: fclem
Maniphest Tasks: T64849
Differential Revision: https://developer.blender.org/D4909
More accurate determination when to draw the PBVH and when to draw the
regular mesh. PBVH drawing is done for Multires, Dyntopo and normal
sculpting with no active modifiers.
Maniphest Tasks: T62070
Differential Revision: https://developer.blender.org/D4731
It makes no sense to control this with a slider, since it became difficult to scale in normal ranges, and scale factors aren't represented well this way.
Still needs a proper solution for older compilers with custom static assert.
macro that can conflict when there is an assert on the same line in multiple
files.
Fix T63790: Crash when applying Scale of Object with vertex parent
Made apply transform to fully live in the evaluated domain.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4976
Was happening when modifier stack detected that mesh is not deformed
and is not modified and attempted to share result across multiple
objects.
This was introduced in 2f77119.
Now functions which are supposed to return mesh owned by caller will
do so again. Shouldn't be a huge impact on memory print since the
data layers are referenced.
- Return Toggle Meta to Strip Menu
- Move position of Lock/Mute in Strip Menu to match position in Context Menu.
- Remove Lock icon.
Author: Peter Fog
Reviewed By: billreynish, ISS, brecht
Differential Revision: https://developer.blender.org/D4957
This also fixes commit rB1fd7b380f4cf8a0489b405de2819f228a4da5ea2 which
didn't do allocation for effect strips properly.
Reviewed By: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4970
A lot of drawcalls don't use the object's properties and don't
need a dedicated DRWCallState. We allocate a unique one at
the begining and use it for all calls that uses the default
unit matrix.
This cleans up a bit of duplicated code and some confusion about
what was culled and what wasn't.
Now everything is culled based on the given object pointer.
If the object pointer is NULL there is no culling performed.
This type of indices is not natively supported on modern GPU and
gives warning on some implementation. The memory savings it
provides is also quite minimal and unlikely to be visible on
nowadays hardware.
This remove some uneeded struct members and makes primitive
restart always enabled by default. This can be broken by addons
if they are not careful enough but many other states have this
problem.
Also leverage GL_PRIMITIVE_RESTART_FIXED_INDEX if
ARB_ES3_compatibility is supported. This removes all API calls
to change restart index depending on indices length.
Modifier keys were disabled, removing support for shift-clicking
on the transform gizmo to constrain to the plane.
The gizmo keymap matches the default, no need to re-define it.
Use drag-only keyamp to adjust gizmos which use drag motion
to change a value. By default gizmos still activate on press.
This allows for left click select to be used with transform
without the gizmos getting in the way.
Even though this isn't necessary for right click select
it allows click events to pass through and be used by tools
which can be useful.
Resolves T63996
Now a small threshold is used for mouse input,
avoiding delay when gizmos are activated on drag.
Tablet input threshold remains unchanged since
it's easier to make small movements when using a tablet.
A larger threshold for non-cursor input is now used (typically keyboard)
which improves usability when the "Pie Menu on Drag" key-map preference.
The issue is that generic ANIM_animchannel_keyframes_loop can't
handle non-FCurve keyframes, so every selection operator does
its own looping over special key types, and the box and region
select operators didn't have code to deal with the summary.
This adds code to handle that in a similar way to other ops.
This is more of a temp urgent hack than a proper fix, chenages required
for the later are too involved for 2.80 at that point of time and will
be done later.
That commit merely keeps the whole existing logic for 3D painting
textures, but instead of re-creating a pool of threads for *every* (!)
stroke evaluation (i.e. mouse move event during painting), we instead
use Blender's task scheduler.
That remains sub-optimal, there are likely more improvements possible in
that code, but it should address the main issue reported (which is
presumably caused by windows thread launching being rather heavy process).
Before it was not possible to see everything in the outliner
when there are e.g. long icon rows. This is because Blender
did not allow panning the view to the right, since it did not
know the actual width.
Most of the code to compute the width correctly was there already,
but there were a couple of issues that made it not work.
* The tree width was computed before the tree was drawn.
This does not work, since the width is only known, after
it is drawn.
* Every `TreeElement` stores its right-most position in `xend`.
However, in the current code, the `xend` of e.g. an object is
the position where the text ends. The `xend` of the icons
is stored in the sub-tree-elements. Therefore, to compute
the maximum width, you may not skip the closed tree elements.
* The current drawing code had an early exit when the icon row
would not be visible anymore. This also skipped the calculation
of `xend`. So it would work correctly, when the icon was visible
a little bit, but not when it was not visible at all.
This patch fixes these issues. So even in more complex files, the
width is computed correcly. At least I haven't found a case, where
it does not.
Unfortunately, some optimizations had to be turned off, to make
it correct.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4887
Fix for GPencil and Cycles Render draw type. GPencil objects were only
shown when overlays were turned on. The cause of this is an
optimization we did to not populate any draw engine when an
external renderer was used with overlays turned off.
This will check if there is any visible GPencil object in the scene. if
so it will still perform the loop.
`DEG_id_type_any_exists` can check if any object of a certain type_id is in the
result. This check is also being used to check if there are any
visible grease pencil objects as a precheck in `DRW_render_check_grease_pencil`.
Reviewed By: brecht, fclem, antoniov
Maniphest Tasks: T65191
Differential Revision: https://developer.blender.org/D4962
The idea is to share a mesh data-block as a result across all objects
which are sharing same original mesh and have no effective modifiers.
This mesh is owned by an original copy-on-written version of object data.
Tricky part is to make sure it is only initialized once, and currently a
silly mutex lock is used. In practice it only locks if the mesh is not
already there.
As an extra bonus, even viewport memory is also lower after this change.
Reviewers: brecht, mont29
Reviewed By: brecht, mont29
Differential Revision: https://developer.blender.org/D4954
Depsgraph: Only invoke callbacks when there are changes
Only affects when an evaluated dependency graph is requested via
context.
Makes it cheap to call when there are no changes made to the graph
Transform: Ensure depsgraph is evaluated when needed
Fix based on D4455 from Campbell, utilizes some recently
introduced functions and allows to not have any extra checks
in the transform code.
Fixes T61904
Fixes T62135
Reviewers: brecht
Maniphest Tasks: T62135, T61904
Differential Revision: https://developer.blender.org/D4967
three issues here:
- when curves had multiple nurbs, the active vert is per nurb
[curve_create_edit_data_and_handles() wasnt taking that into account]
- code could go wrong when points where hidden
- upon selection, tag curve ID_RECALC_COPY_ON_WRITE for batch cache
update
Reviewers: brecht, fclem, sergey
Maniphest Tasks: T64829
Differential Revision: https://developer.blender.org/D4943
The calculation of the mix color was not using the bottom color when the alpha was lower than 1.0.
Also added clamp code to avoid values outside valid ranges.
When XRay is turned on the mesh analysis did not draw correct. This
change won't draw the mesh analysis overlay when xray is turned on.
In terms of giving the user a visual feedback of this limitation we
render the overlay options inactive.
Reviewed By: brecht, fclem
Maniphest Tasks: T65225
Differential Revision: https://developer.blender.org/D4965
The previous fix to the spike bug T64582 was not really right.
This fixes that one properly and restores the desired curving
profile in the bug's example.
sidebar UIList
nodesockets were actually drawn, but immediately drawn over by widgets
from the widget draw batch cache.
solution here is to "widgify" nodesocket drawing as well.
Reviewers: brecht
Maniphest Tasks: T57090
Differential Revision: https://developer.blender.org/D4963
Minor changes to recent gizmo click/drag logic 08dff7b40b
Changing the gizmos highlighted part in the invoke_prepare
callback is too error prone since it needs to run
before it's known which operator will execute.
Add back 'drag_part', since it simplifies click-drag use.
While this isn't essential with custom keymaps per gizmo
it avoids having to define a keymap in the case a drag
event needs a different action.
Remove click-drag support for tweak gizmo,
rely on keymap events instead.
This is needed for some gizmos to use modifiers keys
without having all gizmos use all modifier keys (see: T63996).
The collada exporter allows to export transforms either as Matrix
or as a decomposition of Translation, Rotation aand Scale.
The decomposition option was falsely named "TransRotLoc".
I renamed it to the much more descriptive word "Decomposed".
The tooltip already contains sufficient information, so there
is no need to change that.
It was possible to use a drag event for a gizmo
that dragged away from the gizmo, changing the active gizmo.
Now use gizmo located at the location that was clicked on.
Some versions of the Windows SDK headers have a dbghelp.h that will emit
C4091 warnings, repress them just this once, since the warn can be helpful
in other places.
When exporting an object we can choose the transformation type 'Matrix'
or 'trans/rot/scale' When exporting an animation we have the same choice
regarding the used transformation type.
However we must make sure that animations and objects use the same
transformation type within one colleda export. The user interface is
now reworked such that the correct settings are always guaranteed.
I also reworked the tool tips
Use evaluated object as an input for mesh construction. This ensures
all dependencies are ready.
Reviewers: brecht, mont29
Reviewed By: brecht, mont29
Differential Revision: https://developer.blender.org/D4955
This removes the dither patterns visible in wireframe mode.
This does decrease de depth perception but many users complained
about the visual noise it produces.
This reverts commit ebf924e0b7. This is causing
the transform tool to start immediately which leads to lots of accidental
changes. We can bring this back when that issue is fixed.
This call modifies geometry but does not inform anyone about
changes.
After this change it's possible to load edit mesh, then request
evaluated dependency graph and do some interesting things with
the updated object.
This is part of T63244: object.to_mesh ignores object.update_from_editmode
This commit extends dependency graph API with an argument which
denotes that all custom data layers are to be preserved. This
forces modifier stack re-evaluation with more inclusive mask.
Far from ideal, since this might fail in certain configurations
with indirectly used objects which might be missing layers needed
for the current object evaluation. But this is how it worked for
a long time, so should be good enough for until more sophisticated
solution is found.
In order to use this new behavior two things are to be passed:
- Pass keep_all_data_layers=True
- Pass a valid dependency graph.
The dependency graph is only needed if keep_all_data_layers=True
and is NOT to be passed if keep_all_data_layers=False.
If keep_all_data_layers=True the dependency graph MUST be passed.
Reviewers: mont29, brecht
Reviewed By: mont29
Maniphest Tasks: T64994, T64794
Differential Revision: https://developer.blender.org/D4940
This reverts part of commit b7eba20236. Polling
is causing issues in scripts, and the minor usability improvements are not worth
the extra work this may cause at this point in the release cycle.
Fixes T65149
Code for extending sharp edges assumes ADJ pattern and this
example uses TRI_FAN pattern. This change doesn't fix TRI_FAN
mark sharp bug at least won't infinite loop any more.
This patch makes BLI_task_scheduler_create wait for all worker threads to have started before
returning to caller. For very short workloads (BLI_taks_test) there is the chance that the
worker threads have not fully started yet, and the main thread is calling pthread_join at
the same time as pthread_setspecific is being called on the worker threads which causes a
deadlock on pthreads4w.
Differential Revision: https://developer.blender.org/D4936
Reviewed By: mont29, sergey, brecht
The declaration and implementation of BLI_path_name_at_index were
out of sync leading to build warning
C4028: formal parameter 1/3/4 different from declaration
This marks the headers in the LIBDIR as system headers
and changes the warn to /W0 on msvc versions that support it.
This resolves some warnings we would had to completely
repress otherwise.
Caused by 5adfc51a0f, sharing keymaps caused changing tools to
unregister gizmos and remove their keymaps.
Workaround for now by not removing the keymap.
Issue introduced on 86eefefdc1.
There was a `if scene:` check in the original code, as it turned out
there was a reason for that. We gotta be careful with these assumptions.
This removes the soft and warm studiolight.
This tweak the outdoor and studio presets for more contrast.
It also introduces a rim studiolight that add more variation.
This mode was supposed to be the new default since 2 years already.
But apparently, it was tackled only for doversion, but not for new
outliners (see 7f596d39df).
Loading factory preferences from the preferences window and triggering a
redraw then would cause the failing assert.
We shouldn't mess with window-manager data when loading preferences
only.
- Make edges darker in vert & face select mode (making more
contrast to not loose the topology). Downside is less select
edges visibility in vertex mode. But I'm confident that it's not
as painfull as it seems.
- Make select faces less saturated to have more color contrast
between select faces and edges.
- Make unselected faces white to increase contrast with faces and
edges. The brightening is negligeable for bright surfaces and
help readability on darker surfaces. Reminder that if the faces
overlays are too distracting (i.e: uv mapping, or texturing) they
can be toggled off in the overlay panel.
Reviewers: billreynish, campbellbarton, brecht
Reviewed By: billreynish, campbellbarton, brecht
Subscribers: brecht
Differential Revision: https://developer.blender.org/D4941
When viewport samples are set to 1 simple scenes with volumetrics crash.
EEVEE volumetrics needs to init the post processing buffers. With recent
changes the need for post processing buffers are known after the cache
init. But they are constructed before the cache init. This lead to null
pointers.
Reviewed By: fclem
Maniphest Tasks: T64922
Differential Revision: https://developer.blender.org/D4942
Explicitly disable particles in edit for now.
Those were not rendered already, but were attempted to be converted
to Cycles structures (if UVs were not needed for hair nothing was
rendered, but if UVs are needed then crash happened).
Would be nice to bring hair in edit mode back, but this is a bit
more involved change, which will be done later.
- Move Auto Render into Options panel in sidebar
- Move Pin to the right and remove emboss
- Move Background to the right and use greying out to avoid jumping UI elements
Also remove emboss from pin toggle in Image Editor for consistency
It's closer to the default matcap now, but slightly less metallic and dark. The
reason to use studio lights as default is because the roughness and metallic
parameters of the material then have an effect, and because Texture color mode
does not work for matcaps.
- Need to assign current scene in the builder: it is used to
route relations for object's customdata.
- Tweak relation from scene to object for the customdaat: this
didn't work before because the render pipeline scene has no
view layer component.
Fixes T65044: Crash when Rendering (F12)
Evaluation must never go to original objects and query them:
this is a huge violation of the entire idea of separating
state across viewports and render engines.
Allowed this to only happen for active dependency graph, where
we at least know order of dependency graph update and user
input.
While support for gizmo specific keymaps remains, this should only
be used if a gizmo-group is doing something that requires one.
There was also a hidden limitation that meant only the last registered
tweak keymap would ever be used.
For now leave this using the generic keymap since all
tweak modal keymaps were using the same template anyway.
While internally these are separate gizmos,
there is no reason to have a keymaps for each.
Also prefix the gizmo with "3D View"
since there are other kinds of transform gizmos.
This is too generic flag, and it might be used by anything, starting from
changes in transform ending with changes in ID properties.
The check here is to be as specific as possible. If that is not possible
the decision must be documented.
Related on T63111.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4923
Tagging original ID introduces a conflict of interest when a separate
graph is created and is tagging objects to be re-evaluated with its
context.
This is part of the problem in T63111: tags within a temporary dependency
graph affects viewport and vice versa, which makes logic to wrongly
consider that something did change in the scene and that baking is to
be redone.
This effectively reverts db3bfd0, but this time everything seems to
be updating fine in the viewport.
Root of the issue (beside lack of control over operators evaluation in
keymaps, since tools always get tried first), is that wurrent WM code
only allows one modal op to start from a single event (returning
`PASSTHROUGH` in that case is useless, WM code still considers event as
consumed and does not try any further handling of it).
Solution is then to change trigger events of resize/link operators from
`LEFTMOUSE` to `EVT_TWEAK_L`, which makes more sense anyway, imho, for
drag operations...
Fix T64693: Can't resize selected node.
Fix T64933: "Dots" in node editor can no longer be selected by just clicking on them.
Fix T64660: no access to node connectors when a node is inside a layout frame.
When using texture drawing the material alpha was not set correctly, It
used the `shading.xray_alpha` as this was the default set in the forward
renderer.
This change makes it so a minimal dependency graph which only includes
compositor and sequencer is built for the render pipeline purposes.
Tricky part here is that it's only compositor itself and sequencer who
to use this dependency graph and IDs from it. Render engines are still
to be provided original IDs because:
- They will create dependency graph for the given scene, and currently
it is not possible to create dependency graph from CoW scene.
- IDs from the compositor/sequencer dependency graph are "stripped",
as in, they wouldn't have all view layers, collections or objects
required for proper final render.
This creates annoying mess of mixing evaluated and original scene
access in various parts of the pipeline.
Fixes T63927: Compositing nodes - drivers don't really work
Reviewers: brecht
Maniphest Tasks: T63927
Differential Revision: https://developer.blender.org/D4911
This is used by driers and this is a first step towards support of
scenes used for only compositor or sequencer.
Fixes T61014: Assert adding a driver that uses a single property of a scene ID
Noticed when was looking into T64764, F7043663.
This is a weird case when proxy group is not a group, but is the
same as linked object.
Remove useless relation which was causing cycle, but had no functional
meaning.
Pair programming session with Brecht.
Reviewers: angavrilov, brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4928
At some point when the transform evaluation was split into EVAL and FINAL
the b350edc was re-introduced by EVAL being run prior to proxy_group's
matrix is evaluated.
The File output node stores it settings locally, but the stereo settings
were not displayed, making users only able to use the default settings
of the node.
The cause of not displaying the buttons are was a NULL-pointer check in
`uiTemplateImageFormatViews`. The NULL pointer was used to check if
multiview was enabled. in case of the file output node this check was
performed by the node, so the nullpointer check could be ignored.
Reviewed By: brecht
Maniphest Tasks: T62767
Differential Revision: https://developer.blender.org/D4929
Added new feature: Collada: global axis rotation upon export (UI)
The new feature allows to specify the target rest coordinate system upon export.
This allows for example to export a character that is in Blender orientation (Y forward)
to match the Secondlife orientation where (-X forward)
- Refactor:Added new utility methods to collada_utils
Made BCMatrix class more powerfull
moved Blender related structures into new BlenderContext class
added class wrapper to encapsulate ExportSettings structure
Added blender context getters to ExportSettings
added access methods to BlenderContext into ExportSettings class
Moved class BCMatrix into BlenderContext
moved utility functions from collada_util into BlenderContext
replace own function for parenting by a call to ED_object_parent_set()
- Cleanup: removed obsolete parameters from methods
renamed parameters for better understanding
cleanup whitespace and indentation
removed obsolete comments
`object` is the superclass of all objects. Old-style code could still be
using `class SomeClass(object)` and assigning something else to `object`
could have unexpected results.
This is now also documented in the
[Python style guide](https://wiki.blender.org/wiki/Style_Guide/Python)
on the wiki.
Differential Revision: https://developer.blender.org/D4922
Reviewed by: sergey
That would break proxy behavior after a library reload.
The usual super-annoying loop-back pointers... At least that one is
easily detectable and can be fixed in-place.
Found while investigating T64764.
Missing dependency graph update.
Ideally need to introduce more clear ID_RECALC flag, and maybe go over all
of the operators (some of them might not use dependency graph still).
Masks were not really covered by Copy-on-Write due to mistake
in the dependency graph. After correcting that mistake a lot
of tools became broken, so majority of the patch is related
on making it so access to evaluated/tessellated masks is done.
When accessing evaluated mask state make sure access to an
evaluated dependency graph is done. This solves possible
access to NULL data on redo.
Fixes T64899: Re-doing new point addition causes crash
Reviewers: brecht
Reviewed By: brecht
Maniphest Tasks: T64899
Differential Revision: https://developer.blender.org/D4918
Move keymap handler logic for keymap and gizmo handlers
into their own functions.
This makes it possible to refactor keymap handling
without large changes or duplicating code.
Those are two cases where keeping infamous backward `parents` pointers
of collections in sync is kind of impossible to do... So rebuilding
those relationships from scratch instead.
Fixes e.g. a crash when undoing, then reloading a library, and likely
many more weird ones like that.
Uncovered while investigating T64764.
It's not always possible to keep 'by hand' parent relationships valid in
collections hierarchy. Add functions to remake those
(re-using/factorizing code from `readfile.c` `lib_link_collection_data()`
function).
Can't stress again how painful it is to have those kind of backward
relationships in our data structures, those *always* end up being
serious issues to keep in sync... Should only be generated on the fly
when needed, period. :(
Note that this only fixes the core issue reported (caused by own dummy
mistake in rBd0df7fb3b94ea), investigating that report uncovered at leat
two more issues, including a crasher (when reloading after an undo)...
Not re-reading linked data-blocks in undo/redo case also means that we
do not touch to their usercounts. Even worse, lib_link process in
readfile will increase those (for cases where local data uses linked
one).
Whole data management code is now heavily relying on valid consistent
refcount of all IDs, so we cannot allow that anymore.
Simple solution here could have been to then not increase that one for
linked IDs in `newlibadr_us()`, but unfortunately that would not be
totally bullet-proof, as some local users of linked data may be added or
removed by an undo step...
So I cannot think of any other solution than the ugly brute force one,
i.e. going over the whole Main database and recompute linked IDs users
count... Should not be a big issue performance wise though, this is
fairly cheap process.
This will be needed in undo/redo case, since we do not re-read linked
IDs, their usercounts become total garbage (especially in 'used by local
ID' cases)...
No functional nor visual change.
This is a partial revert of 0910932e71.
The toggle option was introduced on 6640bcca74.
This allow us to simplify the outliner draw code so it uses the icon as
defined in the RNA (as oppose to get the value there once again).
Currently if a panel becomes empty (draw simply returns), it stays
at the last non-empty height. This seems to be caused by some legacy
checks that may be completely obsolete, but the safest fix is to at
least allow resetting height when the panel is open.
Use a default update function for user preferences that tags
dirty and redraws (if changed).
This avoids relying on button changes which fail in some cases.
The unit system is designed for displaying and editing and not for
rendering. Eevee, Workbench and GPencil used these settings to convert
the focal length and sensor size to world units. Making depth of field
render differently with Cycles.
For now we will remove the scale in the draw engines to match cycles,
until we implemented a camera parameters specific scale.
Reviewed By: brecht, fclem
Maniphest Tasks: T64988
Differential Revision: https://developer.blender.org/D4925
Final small tweaks and fixes by Peter Fog:
- Remove commented lines
- Set 'Both' handle selection before Left and Right
- Small tweaks to Strip and Context menus
Any further tweaks will be postponed for 2.81
When using FXAA when rendering to an image the alpha channel was not
correct what lead to visual artifacts.
These artifacts come from the FXAA function that overwrites the alpha
channel with the original Luma of the texel. In the shader this can be
turned on or off. But at the end it always overwrites the alpha with the
luminance.
We didn't use this feature, but the alpha of the resulting pixel still
contained the luma value what lead to render artifacts.
By overwriting the alpha channel with the original alpha we remove these
artifacts.
Reviewed By: fclem
Maniphest Tasks: T64947
Differential Revision: https://developer.blender.org/D4924
Remove the clear allocation flag as it has little impact since there should
be very few allocation per redraw.
Make BLI_memblock_alloc and BLI_memblock_iterstep much more cache efficient
removing them almost entirely from performance profiles.
This will have multiple benefit.
TODO detail benefits (culling, more explicit, handling of clipping planes)
For now the view usage is wrapped to make changes needed more progressive.
The issue was that Audaspace options ended up in the cmake cache though
they should not be there.
Also reverting indentation change by @ideasman42.
Thanks to @mont29 for reporting and helping with the fix.
Don't allow changing it for painted images until they have been saved, similar
to sidebar panels. This could be solved better, for now the important thing is
not to lose changes.
There may well be more vertex shaders that need this, but I couldn't find them
in my testing.
Differential Revision: https://developer.blender.org/D4921
Order 'Matcaps' first instead of 'Flat'.
Order 'Material' first instead of 'Single'.
While we don't have to order defaults first, it's strange
to have obscure options first (in the case of 'Flat').
Track-pad & NDOF events were using KM_NOTHING which wasn't included in
the RNA enum, causing the value to be an empty string in exported key-map
(which then failed to load back).
Add back 'Nothing' value, keep it last since it's not used often.
We are not exposing RNA_ObjectBase in the 2.80 API.
Thus we can't have operators relying on it (e.g, CTX_data_visible_bases,
CTX_data_active_base, ...). Otherwise users won't be able to override
context for these operators.
This commit keep the CTX_data_.*bases() functions around so we don't
need to change the operators and potentially break things that late into
2.80. However as far as the Python scripters are concerned there is no
base to be overriden, ever.
That also simplify the guessing game addon developers have to play when
trying to override an operatori context. They still need to find whether an
operator requires editables, visibles, selected, ... objects. But at
least they don't need to find out whether the operators need base or
object.
- Move Live Unwrap to UV menu
- Move Auto Merge Threshold into a sub-section of Auto Merge to make the relationships clearer
Differential Revision: https://developer.blender.org/D4919
Reviewers: Brecht Van Lommel, Pablo Vazquez
This small fix in the GLSL shader seems do to the trick: now smoke won't jitter when using the adaptive domain.
The previous workaround rB3891ad8e0317 is still needed too, i.e. the bug that caused jitter this time was not related to the previous one.
T64679 mention a desire for a solution that is not in a per-case basis.
However until then we are still better off with this working then not.
Specially since changing individual theme elements works, while reset
theme was not working.
Adding a constant yields quadratic time complexity which can
have quite a big impact on some scenes.
I used the file from T64901 for testing.
In the test file, the time it took to execute `wm_draw_update`
changed from `0.60s` to `0.51s`.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4916
The maximum particles per task of 256 was outdated and lead to too much thread
contention. Instead define a low fixed number of tasks per thread.
On a i7-7700HQ, creating 4 million particles went down from 31s to 4s.
Thanks to Oscar Abad, Sav Martin, Zebus3d, Sebastián Barschkis and Martin Felke
for testing and advice.
Differential Revision: https://developer.blender.org/D4910
When rendering viewport to an offscreen buffer the buffer was
constructed for non anti aliasing (0 samples). This made the objects
that are drawn by the `object_mode` including `wireframe` draw type
non-anti-aliased.
The offscreen buffers will be constructed based on the user setting for
viewport multisampling (`U.ogl_multisamples`). The same setting will
also be used when previewing scene strips in the sequencer. For now
this only improves wireframe drawing in the scene strips. To improve the
Anti aliasing in the scene strips we need to get finer control in the
draw manager. This will be part of a different patch I am preparing.
Please note that this patch also cleansup some unused code in the offscreen rendering (FSAA code was still existing, but never called)
Reviewed By: brecht
Maniphest Tasks: T64849
Differential Revision: https://developer.blender.org/D4907
When doing offscreen rendering (Viewport Render or Sequencer Scene
strip) EEVEE and workbench used the wrong window coordinates. These
coordinates included the border that was not drawn.
Reviewed By: brecht
Maniphest Tasks: T64505
Differential Revision: https://developer.blender.org/D4864
This is old logic that no longer makes sense in the new depsgraph, and causes
issues when multiple threads try to modify the same bevel object.
Differential Revision: https://developer.blender.org/D4913
The task_scheduler was not being explicitly freed, leading to
unpredictable behavior when the process was exiting. The test
would pass, but would sometimes segfault at process shutdown.
This adds an additional way to scrub the playhead, by holding Alt and dragging with the left mouse button, in addition to Shift-RMB.
It's easier to do this, especially with pen input. The other method is still kept, in case you have Emulate 3 Button Mouse enabled, in which case Alt-LMB pans the view
And of course, you can still scrub simply by dragging the playhead, without holding any modifier keys.
Right click select is unaffected.
After new dependency graph API this is no longer needed: all the access
to dependency graph is done explicitly.
Still leaving this flag for C, but that might also be gone in the future.
This was still a proof of concept tool needing further development. Hiding this for now.
It can return in future releases with further development.
See https://developer.blender.org/T57210 for details.
`BKE_collection_object_add_from()` would not check wether collections
were local or not... Trivial to fix.
Note that here I assume we do not use that function in some special
cases where we would like to edit linked datablocks. Think that is
reasonable stance, though.
these are about greasepencil vertices and in these spaces we are not
editing greasepencil objects.
Fixes T64861
Reviewers: antoniov
Maniphest Tasks: T64861
Differential Revision: https://developer.blender.org/D4901
thx @Gvgeo for adding the python/RNA case as well.
Reviewers: brecht
Maniphest Tasks: T64867
Differential Revision: https://developer.blender.org/D4902
Previously, if a boolean happened to use an icon there was no way
to make it display as a check-box from Python scripts.
The previous logic meant we ended up having to edit the RNA.
Since booleans with icons don't work well with the split-property layout
(now used for most of the interface).
Icons were being removed from RNA then added back using awkward Python
ternary expressions in the interface scripts.
The toggle argument now has an unset state (-1).
- toggle=True: no checkbox (emboss).
- toggle=False: always use a checkbox (no icon).
- toggle=(unset/-1): depends on the icon status, default as before.
Since toggle=False was default, this isn't used in existing UI logic.
This removes the extra highlighting of lights that are active as this is not supported
Differential Revision: https://developer.blender.org/D4903
Reviewed by Brecht Van Lommel
See T64324 for discussion of normals ui.
Consensus on blender.chat was to use Alt-N to pull up Normals Menu,
overwriting the old use of that key (Flip normals).
There are still shortcuts for Recalculate Outside and Recalculate Inside
which are likely the more common needs.
Now we always draw a roundbox behind the active object icon, and only change
the text color if the active object is also selected. This matches the 3D
viewport better.
More updates to Sequencer menus from Peter Fog.
- Rename Cut operators in the UI to be more clear
- Re-organize the context menu to be nicer and more useful
- Select menu is clearer
- other assorted fixes and changes for clarity
* Replace Log view transform with Filmic Log.
* Remove Rec.709, DCI-P3 displays that were incomplete and outdated.
* Remove outdated RRT and Film transforms, replaced by Filmic.
* Remove camera responsive curves that don't work with HDR colors.
* Rename Default view transform to Standard.
We're breaking compatibility now for 2.80, so that we can add future
improvements on a clean config.
Part of the code was contributed by George Vogiatzis in D4782.
Differential Revision: https://developer.blender.org/D4900
While this consistency is useful, green selection in edit mode does not stand
out as much as orange. This can cause problems for some users, it seems to
depend on the person. Overall the risk of making the change at this point in
the release cycle is too high.
This reverts commit 5827a47280.
See T64324 for discussion re improving normal editing ui.
As next step, remove the face_strength tool settings because
menu operator now includes that. Move face_strenth enum to
better place.
Remove normals toolbar panel because only thing left
(normal_vector) can stay hidden for copy/paste.
Remove add vector and multiply vector menu entries as
they are useless without ui method for specifying operand,
and they are very low utility operations anyway.
The outliner should not be the only way for users to change these settings.
The Python API was extended to keep these properties positive and keyframable.
Differential Revision: https://developer.blender.org/D4889
The new selection code was redrawing everything for each update.
Use the gestures wmGenericUserData to store the cache between
executions and ensure it's freed.
The properties inside are still hidden when the toggles are off, but we really should use greying out here instead.
That needs changes to DNA/RNA to work though.
Previously, render parts were stored in a linked list and every tile update
searched the entire list for the correct part. As a result, the overhead
of searching tiles increased quadratically w.r.t. the number of tiles.
By hashing the parts based on their location, this operation is much faster,
significantly reducing the tile update overhead for small tiles and/or large
renders.
For example, rendering an empty scene in 1080p at 1spp and 8x8 tiles goes
down from 9.22sec to 1.45sec on my laptop.
Reviewers: brecht, sergey
Differential Revision: https://developer.blender.org/D4896
This patch updates the polling that enable/disables Edit Menu items. Slight Undo History menu changes
Differential Revision: https://developer.blender.org/D4846
Reviewed by Brecht Van Lommel
- add movies to channel above free slot (no check for 2 free slots),
add sound to the free slot
- don't override channel settings if `channel` property is set
This is just a hack. Propper implementation should be done along with T59540
Reviewed By: brecht
Differential Revision: https://developer.blender.org/T59540
A few icons were missed in the last update.
- Added dedicated icon for showing hidden objects in the Graph Editor
- Added new, more descriptive icon for F-Curve snapshots
- Tweaks for Toggle Full Screen and Collections icons
This should be the final set for 2.80.
Thanks to Andrzej Ambroż for contributing this icon set.
Cosmetic change for consistency according to the naming guidelines in 2.80.
Also place Render first, Viewport later to match other areas in Blender
such as the sampling panel in EEVEE/Cycles.
This also replaces the Use Alpha setting. We now have these alpha modes:
* Straight: store RGB and alpha channels separately with alpha acting as a
mask, also known as unassociated alpha.
* Premultiplied: transparent RGB pixels are multiplied by the alpha channel.
The natural format for renders.
* Channel Packed: different images are packed in the RGB and alpha channels,
and they should not influence each other. Channel packing is commonly used
by game engines to save memory.
* None: ignore alpha channel from the file and make image fully opaque.
Cycles OSL does not correctly support Channel Packed and None yet, we are
missing fine control over the OpenImageIO texture cache to do that.
Fixes T53672
For some reason there seems to be no way to do the very simple and
obvious task of converting a scale vector to a matrix via mathutils.
The Matrix.Scale constructor does something complicated instead.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4893
This fixes poor Cycles panel ordering, with Freestyle and Custom Properties
appearing at the top.
For most cases order of registration is still the easiest way to control
order and it's recommended to keep using that. This is mainly to solve a few
cases where we want a few built-in panels to appear below add-on panels.
This patch removes the hover highlight that can sometimes remain after moving out of the Outliner space
Differential Revision: https://developer.blender.org/D4822
Reviewed by Brecht Van Lommel
This patch gives new Windows users a better default preference for fonts folder
Differential Revision: https://developer.blender.org/D4725
Reviewed by Campbell Barton and Brecht Van Lommel
Due to popular demand, this switches the number key mode order, so that it goes:
1: vert, 2: edge, 3: face, 4: object, 5: sculpt and so on
This has a number of downsides in practice, because it works less well for objects other than meshes, which now have a gap in the keymap between 1 and 4. I will try this change anyway, due to popular demand.
Also use V for viewport pie, which makes it easier to switch viewpoints on laptops especially.
SVN takes a long time to sync even if there are no updates,
the `code_update` parameter gives the option opt out of the
SVN updates.
This is a developer option, people just wanting to build
blender and not do any development are highly recommended
to keep using the `update` method.
The previous fix 8a6414ed46, resolved selection picking but didn't
work for box/circle/lasso select.
- Add ED_select_buffer_utils.h for general select-buffer operations
unrelated to edit-mesh.
- Circle select still needs to cache select-id's for each update.
The same was done for Windows, but some extra changes were needed to make it
work on macOS. This is required because the Blender quit dialog now contains
additional settings for image saving.
- Rename Frame menu to Navigation, and move to View menu
- Add missing Zoom entry to View menu
- Move the Snap/Offset items to the Transform menu
- Remove Crossfade Sounds from the Strip menu, now that it is in the Transitions menu
- Added more separators where it makes sense
Suggestions by Peter Fog (tintwotin)
Add missing menu entries:
- Select Box
- Sound Crossfade
- Toggle Meta
- Gap Remove
Also:
- Rename menu entries to use Playhead rather than the ambiguous Frame
- Use icons for the rest of the Add menu categories
Thanks to user tintwotin for pointing out the missing items
Unify all desktop icons on Mac, Win and Linux to fit with the correct, updated brand guidelines here: https://www.blender.org/about/logo/
This is using :
PANTONE 716 C = #EA7600 (RGB 234 118 0)
PANTONE 647 C = #236192 (RGB 35 97 146)
Thanks to Yevgeny Makarov (jenkm) for compiling this icon update.
See T63623
New icons from Andrzej Ambroż / Jendrzych:
- Bespoke icon for creating new Collections
- Special icons for Rigid Body and Rigid Body Constraints (Physics Properties)
- New icons for Holdout and Indirect Only toggles in the Outliner
- New generic Cursor icon for cases that are not related to either Orientation or Pivot
Many other tweaks to existing icons, including:
- Add & Remove Keyframe
- Weight Paint & Collision Modifier
- Tablet Pressure Sensitivity
- Playback icons
The basic idea is that such new images will be packed into the .blend file
when saving all modified images from the quit dialog. To make this workflow
nicer a number of changes were made to how this type of packed image is
handled.
* "Save Modified Images" now packs generated images into the .blend file.
* "Save As" for packed images now automatically unpacks the image, so that
it's easy to save automatically packed images. "Save Copy" keeps it packed.
* "Save" for packed images now re-saves the image into the .blend file, so
that it's effectively the equivalent of "Save" for non-packed images.
* Empty image filepaths are no longer remapped when saving the .blend file.
Previously it would become e.g. "//../../" which makes no sense for generated
images with no filepath yet.
* Hide unpack button and filepath for such packed images with no filepath.
Unpacking does not work in a predictable way without a filepath, better
to just "Save As".
- Add back info icon at the top.
- Remove icons from the buttons. None of the OS's we support use icons here
- Remove button align. It doesn't work well with the active default highlighting
- Center-align the text inside the buttons
To make this look even nicer, we should add more padding all around the edges, but that could be done separately.
If the parent of a collection has a setting disabled, the children should have
that setting inactive.
In some cases a column may affect another one. For example, disabling a
collection to render should make holdout and and indirect only inactive.
In View Layer it works for both objects and collections.
For Scenes mode, it works only for collections.
Differential Revision: https://developer.blender.org/D4888
Editing properties like generated X/Y size clears any changes to the image,
and it's not obvious that this is destructive. Now if the image has been
painted on or baked to, buttons to Save or Discard changes will appear and
editing the properties will be disabled until doing one of these.
There is now a checkbox to enable/disable depth of field per camera. For Eevee
this replace the scene level setting. For Cycles there is now only an F-Stop
value, no longer a Radius.
Existing files are converted based on Cycles or Eevee being set in the scene.
Differential Revision: https://developer.blender.org/D4882
This adds a new dialog that is shown whenever a file is closed.
So, either when a new file is opened, or when Blender quits.
The dialog allows to save unsaved changes. Furthermore it also
allows saving images that have been modified in Blender, but are
not saved yet.
Known limitations:
* Images that have no file path and have not been packed before,
are not saved.
* On MacOS the old dialog is shown when Blender quits.
Reviewers: brecht, billreynish
Differential Revision: https://developer.blender.org/D4860
This was discussed in the past, the idea is to get the outliner even more compact.
Also to let users to use the viewport selection restriction further.
It would be nice to haven indication that an invisible column is set, but this is
no particular to the selection column.
very straightforward: initialize default to the same hard limit as the
RNA properties.
Annoying part is that it's not trivial to make RNA to use same BLI functions,
so leaving that behind for now.
The goal is to prevent assignment of temporary or evaluated meshes
to objects from the main database.
Majority of the change is actually related on passing reports around.
On a positive side there are more error prints which can become more
visible to scripters.
There are still possible further improvements in the related areas.
For example, disable user counting for evaluated ID datablocks when
assignment happens. But can also happen later on as a separate
improvement.
Reviewers: brecht, campbellbarton, mont29
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4884
This is to simplify the usage of Volumetrics.
Now it automatically detect if there is any Volumetric material in the
view and allocate the needed buffer if any.
- Move the Channels setting to the View panel
- Remove double separators in the View panel
- Close the Scene panel by default - it only applies in rare cases
- Move the Frame Overlay from the header to a sidebar panel
Previously it used a strange ghost icon, and it make the other controls jump around
Now uses clearer naming
scene.update() needs to be replaced with view_layer.update() after
rBe693918d4074
note: will go over other occurances of scene.update() in a different
commit
After recent dependency graph API changes it became easy to violate
ownership design (since its changed) and request removal of out-of-main
temporary mesh. This confuses Blender and can cause crashes/locks.
Fixes T64731: Blender freezes after assigning result of object.to_mesh()
This patch does not bring any functional change, but it does expose
some utilities that can be very useful to correct occlusion and
performance problems of Circle Select and similar.
Creating a selection context still makes it easier to track issues.
Makes the result of object.to_mesh() and bpy.meshes.new_from_object()
to be the same as what is visible in the viewport.
This makes Cycles to respect modifiers enabled in edit mode, and should
also easy some scripter's work. The final render still needs some work,
which, maybe, will be about forcing objects out of editing modes.
For existing files, it will use the setting from Cycles or Eevee depending on
the render engine in the scene.
Differential Revision: https://developer.blender.org/D4874
We can not access ensured-to-be-evaluated dependency graph from the
render API: some of it is running from within evaluation which makes
it possible for engines to access list of evaluated IDs.
Solved by passing dependency graph to viewport functions, similar to
the final render functions.
Blender menu:
- Move Preferences to the top as it is used often.
- Introduce two new sub-menus
- About - links from Help menu, plus link to license section in blender.org
- Support Blender - dev fund, store, they used to be in Help menu.
Help menu:
- Add links to Tutorials and Support section in blender.org
- Add link to Developer Documentation (wiki)
- Always show Python API Reference link
One of the usecases is to create mesh from an object is a manner similar to
how Apply Modifiers does it, and have it in the bmain so it can be referenced
by other objects.
This usecase is something what went unnoticed in the previous API changes, so
here is a followup.
Summary of changes:
* bpy.meshes.new_from_object() behaves almost the same as before this change.
The difference now is that it now ensures all referenced data-blocks are
original (for example, materials referenced by the mesh).
* object.to_mesh() now creates free-standing Mesh data-block which is outside
of any bmain. The object owns it, which guarantees the memory never leaks.
It is possible to force free memory by calling object.to_mesh_clear().
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4875
EEVEE assumed that the OGL renderer did FSAA, as the FSAA was removed we
needed to revalidate this assumption. The temporal sampling only inited
the matrices during init phase. As now we need to update the matrices
for every sample rendered the code for updating the matrices was
isolated in a new function `EEVEE_temporal_sampling_update_matrices`.
Reviewed By: fclem
Maniphest Tasks: T64646
Differential Revision: https://developer.blender.org/D4871
* Change circle to roundbox around active icons, so they don't overflow.
* Change text color to indicate selected and active state.
Differential Revision: https://developer.blender.org/D4650
This also makes `IDP_CopyProperty` the "opposite"
of `IDP_FreeProperty`, which is what I'd expect.
Two refactoring steps:
* rename IDP_FreeProperty to IDP_FreePropertyContent
* new IDP_FreeProperty function that actually frees the property
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4872
This new filter allows to select only some type of keyframe (keyframe, breakdown, etc).
This was a request of artists that are used to work with other 2D softwares.
Also some cleanup to remove aninmatable option to some properties.
- Use single column layout
- Move the rather obscure Backdrop toggle into the View menu
- Use correct units for Crop and Offset (pixels)
- Re-organize the sidebar in 2 main groups: Adjust & Info
Adjust includes Compositing, Offset, Crop, Video, Color, Sound
Info includes strip name, path, start/end and other data
Original changes by tintwotin, with changes and adjustments on top by me
When using Intel GPU EEVEE did not display anything. This was due to an
internal shader compilation bug inside the intel drivers. We had fixed
this for other vertex shaders. The same change we have to apply to other
vert shaders that want to limit the need of Matrix multiplications.
The kernel does not use AVX2 vectorization, and trying to use BVH8 was
leading to an empty scenes.
Fixes T64624: Ctest : Win32 + AVX2 fails virtually all cycles tests
Main goal here is to make it obvious and predictable about
what is going on.
Summary of changes.
- Access to dependency graph is now only possible to a fully evaluated
graph. This is now done via context.evaluated_depsgraph_get().
The call will ensure both relations and datablocks are updated.
This way we don't allow access to some known bad state of the graph,
and also making explicit that getting update dependency graph is not
cheap.
- Access to evaluated ID is now possible via id.evaluated_get().
It was already possible to get evaluated ID via dependency graph,
but that was a bit confusing why access to original is done via ID
and to evaluated via depsgraph.
If datablock is not covered by dependency graph it will be returned
as-is.
- Similarly, request for original from an ID which is not evaluated
will return ID as-is.
- Removed scene.update().
This is very expensive to update all the view layers.
- Added depsgraph.update().
Now when temporary changes to objects are to be done, this is to
happen on original object and then dependency graph is to be
updated.
- Changed object.to_mesh() to behave the following way:
* When is used for original object modifiers are ignored.
For meshes this acts similar to mesh-copy, not very useful but
allows to keep code paths similar (i.e. for exporter which has
Apply Modifiers option it's only matter choosing between original
and evaluated object, the to_mesh() part can stay the same).
For curves this gives a mesh which is constructed from displist
without taking own modifiers and modifiers of bevel/taper objects
into account.
For metaballs this gives empty mesh.
Polygonization of metaball is not possible from a single object.
* When is used for evaluated object modifiers are always applied.
In fact, no evaluation is happening, the mesh is either copied
as-is, or constructed from current state of curve cache.
Arguments to apply modifiers and calculate original coordinates (ORCO,
aka undeformed coordinates) are removed. The ORCO is to be calculated
as part of dependency graph evaluation.
File used to regression-test (a packed Python script into .blend):
{F7033464}
Patch to make addons tests to pass:
{F7033466}
NOTE: I've included changes to FBX exporter, and those are addressing
report T63689.
NOTE: All the enabled-by-default addons are to be ported still, but
first want to have agreement on this part of changes.
NOTE: Also need to work on documentation for Python API, but, again,
better be done after having agreement on this work.
Reviewers: brecht, campbellbarton, mont29
Differential Revision: https://developer.blender.org/D4834
Note that the same issue actually showed without a frame node, when
trying to click-drag on a socket on its 'inside node' part, you would get
same behavior.
Only solution I can see here is to prevent Node selection to go on when
user clicks on one of its sockets, there is no way afaik to make
drag-from-socket to start if we keep select-node operator running
modal, since both operators (NODE_OT_select and NODE_OT_link) use the
same shortcut, if select (which is checked first, being a Tool) returns
modal, then event is considered handled it seems, even though
Passthrough is also returned...
* Align category icons and color use with Default theme.
* Set colors for new properties and regions.
* Fix several issues.
Needs further testing to tackle all editors but this puts it at much more
usable level compared to how it was.
When normally editing curves, deleting the last keyframe also
deletes the curve. Thus if for some reason it didn't happen,
e.g. maybe due to removing keys directly via Python, skip the
bad curve instead of resetting the channel to zero.
This constraint requires full bone data to evaluate, so it can't
use the generic single target matrix system. Calculating it is
thus useless waste of CPU time.
Helps to:
* Make a connection between what we see in Outliner, Viewport and Properties editor
(clicking on the obdata icon takes us to Edit mode, which shares colors).
* Quickly tell which mode we are in.
* Armatures have now distinct color when in Object, Edit, and Pose modes.
Missing is Pose mode which is currently cyan in the viewport and obdata category in properties,
it would probably need its own special category though as it doesn't match the others.
The choices are now World, View and 3D Cursor.
This breaks Python API compatibility, add-ons that add objects with this
parameter will need to be updated.
Differential Revision: https://developer.blender.org/D4706
Both windows.h and wm_types.h define HKEY, moving wm_types after windows.h
sidesteps the issue, since HKEY isn't used in fsmenu.c it doesn't cause
problems.
We need to make those paths validation async, since some infamous OS is
unable to provide a quick way to check whether a path is valid or not...
Not much to say here, this is basic wmJob stuff really... We don’t even
need complex threaded coordination locks or atomics, given how simple
the changes are here.
Fake-tested with some `sleep()` calls to add atificial delay, seems to
work as expected.
This is the angular diameter as seen from earth, which is between 0.526° and
0.545° in reality. Sharing the size with other light types did not make much
sense and meant the unit was unclear.
Differential Revision: https://developer.blender.org/D4819
This removes the ability to hit X or Del to delete data from inside the File Browser.
This was extremely dangerous because it didn't use the system trash/recycle bin. This made it a very dangerous operation with no way to get data back.
Later on, we should add support for the OS trash through native API's, but until then we will remove this feature from the keymaps.
This addresses T61412 and T64104
This uses the Blender icon that previously activated the splash.
Menu items which apply to the application are located here:
startup file, app templates & preferences.
Due to recent changes the default aa samples in the viewport was set to
16, but should have been 8. This is due to how the old viewport quality
setting was interpreted by the workbench engine. This patch will respect
the same way to version the viewport quality to viewport_aa as the
workbench used to do this.
For anyone interesting on polishing pixels, this is still one of the few
places in Blender where when renaming the name "jumps" a tiny bit when
editing. Most of the other places (id rename, UI list rename, ...) have
the text drawing in the exact same place while editing it and drawing.
A better solution would be to not use the callback mechanism anymore for
cases like this where the dependency graph will free volume data, but
that would be a bigger refactor.
No major API change here, only in the outliner restriction column
variables (e.g., show_restrict_column_selectable > show_restrict_column_select).
* Get rid of _INSTANCE (introduced on b1af682001).
* Differentiate (everywhere but the API) between HIDE (temporary) and VIEWPORT (global).
* Use the expected icon for restrict viewport (same as objects and modifiers).
* selectable > select
In the outliner we don't have yet a clear distinction for the users of when
to use each of the visibility settings.
This changes that by properly naming the property and their tooltips.
I'm also unifying the naming between the rna properties and the outliner
for collections and objects (e.g., so collection and object hide_select
have the same tooltips everywhere).
The API did not change.
Since drivers on Bone properties are really supposed to be stored
in Armature data and access bones via its bones[] collection, this
lookup path should work efficiently.
Mass lookup of bones by name was already done through hashes,
but they were built temporarily every time that was needed. This
simply replaces it with a common hash table computed immediately
after file load, copy, or Edit to Object mode switch.
This was fixed for active tool gizmos but not viewport gizmos.
Note that this applies only to the Industry Compatible keymap, which is where this issue manifested itself.
Left-click select broke that behavior, since it puts both action and
select buttons on the same physical mouse button...
To support this behavior again, we have to split selection process in
two steps, hence make it modal... While I remain rather skeptical about
that global design decision, and complexity it adds to many UI/UX areas,
this solution ends up being OK-ish I think.
Thanks to @brecht for some final tweaks on the patch.
- Fix 'ddisable' typo
- Use clearer names for the operators in the Collections context menu
Instead of Set Exclude / Clear Exclude, we now use Enable in View Layer / Disable from View Layer
On Windows 10 with a post Februari 2019 Intel driver, the box selection
is not working. It only detects the object centers, but not the drawn
triangles of the basic engine.
Reviewed By: fclem
Maniphest Tasks: T62947
Differential Revision: https://developer.blender.org/D4857
The complexity in this patch comes from the fact
that the current operator system does not support
multi-step user interactions well.
More specifically, for this to work, we need to show
a confirm dialog and a file browser afterwards.
We decided that it is easier to keep everything in
a single operator, instead of creating separate
operators that invoke each other.
So, now the `WM_OT_open_mainfile` operator invokes
itself in different states. It implements a simple
finite state machine to manage the states.
The dialog itself is expected to be improved in
a future commit. See D4829 for more details.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4829
While shapekeys RNA path handling does support keyblocks (ShapeKeys)
with an `id_data` wrongly set to the geometry instead of the shapkey,
this is really not something to promote. ideally it should be removed at
some point. At least, let's not generate such RNA pointer if we can
avoid it.
Here it was breaking proper showing of animated shapekeys in the UI,
since this expects animated datablocks to have their own animation data
(and 'rooting' keyblocks to their mesh instead of their shapekey ended
up defining their animation in the mesh, not in the shapekey).
The refactoring of texture handles did not take into account that render
services are shared between multiple render session. Now the texture
to handle map is also shared between render sessions.
New option to disable the follow drawing path. Before it had only a switch property, now there is a list of options.
Modes:
Path: Follows drawing stroke and rotate with object.
Object: Only follows object rotation.
None: Don't rotate.
The owning ID reference of RNA pointers is supposed to point at the
ID container that owns the object, but for PoseBone.bone it wasn't
updated to point at the Armature data ID instead of Armature Object.
This caused issues, like pose_bone.bone.driver_add() adding the
driver to the Armature Object animation data, which violates the
intended design of the animation data structures.
Since RNA code generally assumes that all pointers that don't
refer directly to an ID remain within the current ID, a custom
getter is required to fix this.
This is a big change that cleanup a lot of confusing code.
- The instancing/batching data buffer distribution in draw_instance_data.c.
- The selection & drawing code in draw_manager_exec.c
- Prety much every non-meshes object drawing (object_mode.c).
Most of the changes are just renaming but there still a chance a typo might
have sneek through.
The Batching/Instancing Shading groups are replace by DRWCallBuffers. This
is cleaner and conceptually more in line with what a DRWShadingGroup should
be.
There is still some little confusion in draw_common.c where some function
takes shgroup as input and some don't.
This is in order to have VAO handled by thoses batches instead of using a
common VAO. Even if the VAO has no importance in these case using a batch
will help when transitioning to Vulkan.
Tagging dirty was failing for UI scale, line width, hinting & others.
Checking for an update function means all update functions
need to tag preferences as dirty.
Since this check was added to prevent the active section marking the
preferences as dirty, only exclude this property.
This is necessary to correctly do low-level keyframe manipulation
in tweak mode, and the logic is complex enough that re-implementing
it in Python is impractical.
since own rB5d9d32fd1fa3 (mirror bbone scaling in editmode)
- bbone scaling in posemode was missing immediate updates
- bbone scaling in posemode could crash
This now properly supports mirroring in posemode as well.
note: for bbone scaling, I made both X-Axis-Mirror options (editmode
option as well as posemode option) valid.
Fixes T64091
Reviewers: brecht
Maniphest Tasks: T64091
Differential Revision: https://developer.blender.org/D4851
The most difficult part is handling parent-child relations correctly:
when a parent is applied, the children should be moved accordingly,
and when applying a child, it should not include transformation from
unapplied parents. All this requires walking bones as a tree, instead
of a flat list.
Limitation: Applying bones with non-uniform scaling without also applying
children will distort non-rest posing on said children for reasons related
to T54159 (basically, non-uniform scale plus rotation creates shear, and
Blender matrix decomposition utilities don't have tools to deal with it).
Reviewers: campbellbarton, brecht, mont29
Differential Revision: https://developer.blender.org/D3775
Was missing copy-on-write tag since lamp itself has no geometry or
transform.
Now tagging for animation, and taking care of special case in the
dependency graph.
This makes it easier to set up materials with emission and transparency.
Importers/exporters and add-ons are recommended to now use these rather than
creating separate transparent BSDF and emission nodes.
Cycles lights now use strength and color properties of the light outside
of the shading nodes, just like Eevee. The shading nodes then act as a
multiplier on this, and become optional unless textures, fallof or other
effects are desired.
Backwards compatibility is not exact, as we can't be sure which renderer
the .blend was designed for or even if it was designed for a single one.
If the render engine in the active scene is set to Cycles, lights are
converted to ensure overall light strength remains the same, and removing
unnecessary shader node setups that only included a single emission node.
If the engine is set to Eevee, we increase strength to remove the automatic
100x multiplier that was there to match Cycles.
Differential Revision: https://developer.blender.org/D4588
Cycles now uses the color space on the image datablock, and uses OpenColorIO
to convert to scene linear as needed. Byte images do not take extra memory,
they are compressed in scene linear + sRGB transfer function which in common
cases is a no-op.
Eevee and workbench were changed to work similar. Float images are stored as
scene linear. Byte images are compressed as scene linear + sRGB and stored in
a GL_SRGB8_ALPHA8 texture. From the GLSL shader side this means they are read
as scene linear, simplifying the code and taking advantage of hardware support.
Further, OpenGL image textures are now all stored with premultiplied alpha.
Eevee texture sampling looks a little different now because interpolation
happens premultiplied and in scene linear space.
Overlays and grease pencil work in sRGB space so those now have an extra
conversion to sRGB after reading from image textures. This is not particularly
elegant but as long as engines use different conventions, one or the other
needs to do conversion.
This change breaks compatibility for cases where multiple image texture nodes
were using the same image with different color space node settings. However it
gives more predictable behavior for baking and texture painting if save, load
and image editing operations have a single color space to handle.
Differential Revision: https://developer.blender.org/D4807
Curve function had two arguments:
- for_render, which was originally supposed to be used to control
whether viewport or render visibility for modifiers is to be
used.
- use_render_resolution, which sounds like it is supposed to control
whether viewport or render resolution for curves is to be used.
What is totally confusing is that those arguments were used
interchangeably: sometimes use_render_resolution would control
modifiers visibility.
This commit makes it so there is one single argument for this.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4850
For wome reason, tag to mark normals as needing recompute was commented
out instead of updated to new system, during conversion of DynaPaint
code for 2.8 new evaluation stack... Trivial fix.
Since these are temporary properties, changing them shouldn't cause
undo pushes. There already is a flag to disable that, but since each
operator inherits its own properties from the base, RNA define code
also had to be changed to allow inheriting the flag.
This patch fixes an issue with the ColorRamp overflowing its drawing bounds
Differential Revision: https://developer.blender.org/D4843
Reviewed by Brecht Van Lommel
- Fix Alt-key navigation (was conflicting with Alt-click to set straight lines)
- Use consistent shortcuts for brush strength and size
- Use direct shortcuts for switching tools
Expose REPLACE and CYCLE_AWARE, and add AVAILABLE for completeness.
These flags are generic and safe to use, and necessary to match
the behavior of certain UI options.
The change was introduced on b1af682001.
We may still change the name functionality of this anyways - we are
not really changing only the instancing visibility here at the moment.
So there is no reason to break the API in the meantime.
And in the future we can simply add a new API entry, leaving the old one
to be slowly deprecated.
All buttons will get their own block's `func` by default, this cannot be
considered an error, do not assert in that case either.
Fixes bunch of very annoying asserts e.g. when using FBX exporter...
See T61578 for discussions and mockups.
Visibility Options
==================
We are adding more granular control over restriction columns in the outliner,
exposing "indirect only" and "holdout" as options, and change the way
users enable/disable collections in a viewlayer.
We also rename the object viewport restriction to hide instance.
So the options we have are:
Collection
----------
* Render Visibility
* Instance Visibility
* Selectable
(View) Layer Collection
-----------------------
* Enable
* Holdout
* Indirect Only
* Viewport
Shortcuts
=========
Isolate Collection
------------------
* Ctr + click isolates the collection.
It turns all its parents and children "visible", and all the other
collections "invisible".
If ALL the collections were already properly set, we re-set the
collections to their default value.
Set Collection Inside Collections and Objects
---------------------------------------------
* Shift + click: Set/unset inside collections and objects.
We only set objects values as well when we are in View Layer mode and
(obviously) when the objects have a matching property.
Icons
=====
Little reminder that we will need better icons for holdout, indirect only, and
probably instanced (nothing wrong with the current, but it differs from
the proposal when it is turned off).
Also, we need to decide where do we want the modifier/bones/... icons to
be (in which column) and ideally make sure their icons match the ones we
use for collections/objects.
At the moment those are using the screen icon, which is not being used
by collections.
Reviewers: brecht, billrey
Subscribers: pablovazquez
Differential Revision: https://developer.blender.org/D4823
Blender startup time and shader compilation is a big factor when running
hundreds of tests, so now all renders in the same ctest run in the same
process. If a test crashes, the remaining tests in the same category will
be marked as skipped.
Benchmarked on a quad core with ctest -j8.
cycles: 118.1s -> 94.3s
eevee: 66.2s -> 29.2s
workbench: 31.7s -> 8.6s
- Silence now has an optional sample rate parameter.
- Fix: wrong length reported by modulator and superpose.
- Minor formatting, include and documentation fixes.
It was a missing invert matrix update and also setting the OB_NEG_SCALE.
This defeats the purpose of not computing it in the draw manager but this
is more local and is have not a dramatic impact on performance.
This adds the ability to assign shortcuts to boolean settings and enums
via the right click menu.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4808
Versionning code generating collections from old 2.79 scene layers was
setting new collections' library from scene, but it also needs to
properly set the `LIB_TAG_INDIRECT` tag, otherwise localizing code does
not know what to do, and most of append process fails.
Internally tool settings have been moved to the 3D view.
Added the ability for to draw panels from another space/region
so they can be mirrored in the properties editor.
This makes the properties editor navigation bar less saturated. Internally
the icon theme coloring was refactored a bit to move more towards the button
drawing code.
Monochrome colored icons don't work well on a dark background, so now we can
add a border around them. Note that most icons in the interface will remain
without a border, just the outliner and properties editor navigation have
colored icons and those will get a border. Other icons continue to be drawn
in the text colored without a border.
Differential Revision: https://developer.blender.org/D4787
By default the texture is locked and only the box is aligned when Follow Drawing Path is enabled.
Before, when the Follow was disabled, the texture was always aligned to top and it was not affected by object rotation. Now, the texture always is rotated with object rotation.
The DrawManager assumed that the first render layer was the render layer
where GP needed to render on. In viewport this is always the case as we
only have a single view layer. When rendering this can be multiple when
multiple layers are rendered and composited in the compositor.
In stead of the assumption that the first render layer is the render
layer we need to draw on, we search for the render layer with the same
name as the viewlayer.
Reviewed By: fclem, brecht, antoniov
Maniphest Tasks: T63099
Differential Revision: https://developer.blender.org/D4818
In recent changes the viewport_quality setting was not working what
users expected. This change will separate the anti-aliasing method that
is being used.
We now have three settings:
* scene.display.render_aa: Will be used during `Render Image`.
* scene.display.viewport_aa: Will be used during `Viewport Render Image`.
* userpref.viewport_aa: Will be used in the 3d view.
The viewport_quality setting has been replaced by the viewport_aa
setting as it was the only thing in currently controlled.
Reviewed By: brecht
Maniphest Tasks: T64132
Differential Revision: https://developer.blender.org/D4828
When using multiple viewlayers and switching between them the selection
gets buggy. The reason for this is that the select_id is updated based
on the index in the viewlayer. This makes the select_id not unique as
objects might be shared or not shared at all.
This fix will update the select_id on all objects in main. This will be triggered in all the selection operators.
Reviewed By: sergey, brecht
Maniphest Tasks: T55617
Differential Revision: https://developer.blender.org/D4824
This change the logic a bit, when 2 bsdfs using SSS are mixed, we use the
one with the biggest radius (on a per pixel basis).
This change from previous behavior which was to select input 1 if radius
was greater than 0.
While not exactly optimal, that's the only disambiguation solution I
found doable for translations (i18n). ID names tend to sneak
everywhere, including in places where we have no access to i18n context
currently (like menu labels e.g.).
Other alternative would have been to use 'Lightweight', but that is a
tad too long, we love short UI messages as much as possible.
The genius who decided that it was critical to rename ID lamp to light
in 2.8 is welcome to find a better solution...
Part of T43295.
* Timeline has summary collapsed to show only one row of keyframes.
* Cavity default is screen space curvature (faster).
* Scripting workspace text editor uses syntax highlight and line numbers.
* Marker lines enabled by default in all animation editors.
* Movie clip editor pivot point default to median.
Ref T63986.
This is in order to create less shading group when using duplis.
Data for dupli objects keep in draw manager state until the source object
changes so retrieval is fast.
Note that this system could be extended to all meshes.
This is really close to BLI_mempool but uses an array to keep track of the
chunks of memory. There is no tagging necessary to clear the whole
structure so reuse is fast.
Naturally supports iteration but does not support freeing.
Not sure why newid was set in evaluated object instead of orig one,
makes no sense to me, since we want to operate remapping of ID pointers
on orig data? Looks like that was something overseen when that code was
ported to new COW system.
This is a patch for the default keymap in Blender.
It relates to the Dopesheet, Timeline, Graph Editor, NLA and Sequencer
Currently, in these editors, dragging outside of your selections does nothing.
This patch makes it so dragging outside the selection does a box select operation.
It is consistent with how the Node Editor works, as well as the 3D View, if you use the gizmo overlays.
Differential Revision: https://developer.blender.org/D4799
Reviewers: campbellbarton
Many modern computers support a lot of threads (parrallel building
jobs), but are somewhat restricted in memory, when some building jobs
can require several GB each.
Ninja builder has pools, which extend the usual `-j X` make
parallelizing option, by allowing to specify different numbers of
parallel jobs for different targets.
This commit defines three pools, one for linking, one for usual compile,
and one for compiling some 'heavy' cpp libs, when a single file can
require GB of RAM in full debug builds.
Simply enabling WITH_NINJA_POOL_JOBS will try to set default sensible
values for those three pools based on your machine specifications, you
can then tweak further the values of NINJA_MAX_NUM_PARALLEL_ settings,
if you like.
On my system (8 cores, 16GB RAM), it allows to build a full debug with
all ASAN options build with roughly 7GB of RAM used at most, pretty much
as quickly as without that option (which would require up to 11GB of
available RAM at some points).
Review task: D4780.
Since scale is multiplicative, the appropriate way to partially copy
it is to use power. However, the influence slider of constraints uses
linear interpolation. Thus, there is no way to correctly split scale
via constraints without adding this feature.
In addition, this allows inverting scale by using negative powers,
fulfilling the function of Copy Rotation's Invert checkboxes.
A 'Disable and Keep Transform' button for constraints was added. This
allows animators to disable a constraint without moving the constrained
object/bone, making it easier to toggle constriants on and off without
any visual consequence. Typical usage would be a character picking up an
object (enable 'Copy Transform' constraint) and placing it somewhere
else (disable the constraint).
Note that there could still be movement when there are muliple
constraints active. For example, when using this constraint stack
- #1: Copy Transform from Empty.001
- #2: Copy Rotation from Empty.002
and disabling constraint #2, constraint #1 is still active and will
still modify the visual transform of the object. According to our
in-house animators, this is expected behaviour.
Reviewers: campbellbarton, dfelinto, sergey
Reviewed By: campbellbarton
Subscribers: brecht
Tags: #animation
Differential Revision: https://developer.blender.org/D4677
This was trying to emulate the 2.79 Graph Editor.
The toggle took up more room that the filter toggle it revealed, and it made the header buttons jump around
Better to just have the filter toggles showing immediately.
Add a new option that makes the Spline IK solver apply volume
preservation on top of the original scaling, considering the
pre-IK scale of the bone as the goal volume to be preserved.
This basically works similar to the Stretch To constraint, and
allows easily rigging a stretchy chain that uniformly follows
its parent's scaling.
Since the Stretch To behavior is more familiar, the new option
is on by default for newly created Spline IK constraints.
Perviously it was only possible to interpolate from the breakdown poses.
Now you can Push/Relax in regard to the rest pose as well.
For this only one keyframe is needed while the old modes needs two.
- This only applies to left click select. Right click select and the legacy keymap are unaffected
- You can still set the playhead from anywhere, using Shift-RMB, just like how you set the cursor in the 3D View
The hard-coded transparency of just 16 made it hard to see the markers
when the background was busy with keyframes (or strips in VSE).
The rename of the setting is in the following commit.
Reviewers: billreynish
The main reason for this change is to allow setting the
active frame with the left mouse button, while still being
able to select e.g. keyframes with the same mouse button.
The solution is to introduce a new scrubbing region with
a specialized keymap. There are a couple of related todos,
that will be handled in separate commits.
Those are listed in D4654.
This solves T63193.
Differential Revision: https://developer.blender.org/D4654
Reviewers: brecht, billreynish
Most properties aren't altered by the evaluation of the bone,
and can be read immediately from its input operation. B-Bone
properties can be evaluated at the last possible moment.
This provides more freedom in using drivers to connect bone
properties: for example, it is now possible to read raw local
transform values via drivers from a bone that depends on the
reader bone via constraints.
Previously you could only orbit.
Apparently the order of the keys in the modal keymap makes a big difference.
Thanks to users Znio.G and Oskar on Devtalk who provided this solution.
This happened to be a bigger rabbit hole to hell than it originally seemed,
and there are higher priority design tasks to be handled (at this point high
priority design task is more important than high priority bug fix).
After talking to Brecht the decision was made to revert to the known isolated
issue, which will allow everyone in the studio work same as prior to last
Friday.
The remaining bits will be worked on after all the design tasks are out of
the way.
This commit reverts:
4cdb4b9532 Fix T64161: Crashing using undo and multiple windows
064273a4ae Sound: Port more cases to be a part of dependency graph
2e582f8ab5 Sound: Fix access wrong dependency graph
5fc49d9c91 Sound: add stubs to build without audaspace
c68c81a870 Sound: Make sure spin lock is initialized for new sound datablocks
c02534469a Sound: Delay creating sound scene handle for until is needed
9f681bea68 Fix T64144: Crash when displaying audio waveforms in VSE
2f79286453 Cleanup: unused vars
bed8ad6f95 Fix crash in background rendering after recent sound changes
773691310f Fix T64143: Crash when scrubbing in the graph editor
888852055c Sound: Fix for being unable to jump to a frame during playback with A/V sync
6ab7b38464 Sound: More fixes for access of original scene
35db119545 Sound: Fix access original scene during playback
211c4fd2e9 Depsgraph: Make comment about evaluation more obvious
c5fe16e121 Sound: Make sound handles only be in evaluated datablocks
b4e1e0946b Depsgraph: Preserve sound and audio pointers through copy-on-write
4eedf784b0 Depsgraph: Store original sequencer strip pointer
6990ef151c Sound: Move evaluation to dependency graph
d02da8de23 Sound: Delay opening handlers for until really needed
3369b82891 Depsgraph: Add scene audio component
e8f10d6475 Depsgraph: Tag sequencer for update on changes
6e4b7a6e4d Depsgraph: Initial work to cover sequencer
17447ac5a6 Depsgraph: Make sound ID part of the graph
matrix inversion was changed in rB01c75c3765eb from own code to EIGEN
for performance reasons. EIGEN would return a zero matrix on failure
(resulting in the pivot always being at the object origin).
This brings back the "old" matrix inversion code (which has the benifit
of providing a partial solution which makes the local transform center
appear correct)
Reviewers: campbellbarton
Maniphest Tasks: T57767
Differential Revision: https://developer.blender.org/D4804
This allows creating hooks completely without the use of operators.
Also fix subtarget to actually recompute the inverse matrix.
In order to follow the vertex_indices_set function naming convention,
it is necessary to fix makesrna to avoid a naming conflict between
the function wrapper and the property accessor by adding a tag.
Differential Revision: https://developer.blender.org/D4798
Do not iterate over degsgraph when overlays are turned off and
rendering via an external engine. External engines sync data
from Blender differently. The external engine
draws the depth buffer, but that is only needed for overlays.
Reviewers: fclem, brecht
Differential Revision: https://developer.blender.org/D4791
After a lot of thinking about this, I decided that all operation modes
that I've tried over the past couple of years, including the original
2.79 one, have their uses after all. Thus the only reasonable solution
is to add yet another option.
The modes are:
- Strict: The current 2.80 mode, which overrides the original scaling
of the non-free axes to strictly preserve the volume. This is the most
obvious way one would expect a 'Maintain Volume' constraint to work.
- Uniform: The original 2.79 mode, which assumes that all axes have been
scaled the same as the free one when computing the volume. This seems
strange, but the net effect is that when simply scaling the object
uniformly with S, the volume is preserved; however, scaling the non-
free axes individually allows deviating from the locked volume.
This was obviously intended as a more or less convenient UI tool.
- Single Axis: My own variant of the intent of the Uniform scale, which
does volume-preserving if the object is scaled just on the Free axis,
while passing the non-free axis scaling through. I.e. instead of
uniform S scaling, the user has to scale the object just on its
primary axis to achieve constant volume. This can allow reducing the
number of animation curves when only constant volume scaling is needed,
or be an easier to control tool inside a complex rig.
Refactor a bit node selection code, mostly in the socket selection in
one place 'extend' logic usage was inverted, deselection of other
sockets/nodes was not working properly.
Also now make node active when selecting a socket, this makes more sense
and makes the NODE_OT_link_viewer macro much more useful (this is the
only user of the 'socket select' feature currently, afaics).
Being able to compare Eevee reference images is useful for refactoring I'm
working on so might as well add them now, even if we can still improve them.
Workbench tests are just rendering the same files as Cycles and Eevee. This
doesn't really tests many workbench settings until we add tests specifically
for them, but does cover how it it handles the different object types.
Unlike location and rotation, there is a meaningful definition of
overall/average scaling via the total change in the volume. This
adds an option to retrieve that via a single driver variable,
instead of having to use three and an expression.
Using the determinant to compute the volume scaling also allows
detecting flipping due to negative scale - this is impossible
to do via the three variable approach.
The volume_scale functions are added purely for code readability:
'volume scale factor' is easier to understand than determinant.
Differential Revision: https://developer.blender.org/D4803
- was introduced in rB844a17a3d9d2 but apparently never used
- spotted while looking into T57767
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4805
A template_curve_mapping with brush=True was always resetting the curve
to a positive slope (descending from left to right).
Behaviour is now changed so that specifying "use_negative_slope" is
taken into account as well.
Also use this for:
- paint "cavity_curve"
- gpencil "curve_sensitivity" / "curve_strength" / "curve_jitter" /
"interpolation_curve"
Reviewers: brecht
Maniphest Tasks: T64164
Differential Revision: https://developer.blender.org/D4809
Mainly covers RNA callbacks which were still doing direct scene update,
which was causing crashes. Now corresponding ID_RECALC flags are used,
so all scenes can update accordingly.
Also tested animated volume/pitch on strips, which now works as well.
Fixes T64133: Assert after changing FPS
Fixes T64154: Immediate crash when changing the current frame on the timeline
Fixes T64185: Client Crashes when the frame position value is changed
Fixes T64190: Blender Crash using Timeline Editor
Fixes T64128: Click to close bug type on timeline
Fixes T64147: Crash when setting current frame from Python
Fixes T64152: Blender Auto-Close on timeline change
Only a few icons are colored, which caused some random icons to be colored and
others not. We can add it back for specific menus later (like add object or
modifiers).
New icons from Andrzej Ambroż / jendrzych:
- New icons for select tool modes
- Brought back screen icon for viewport visibility toggles
- Added new Instanced visibility toggle (unused currently - meant for forthcoming updates to the Outliner)
- New Gizmo icon for the header
- Many tweaks and alterations to existing icons. Full list on Devtalk
- Use correct icon when Proportional Editing is disabled
Due to some fields in the context being NULL access was happening
to the default view layer, not the active one.
Simply re-arranged context initialization, so it happens before
accessing dependnecy graph.
Fixes T64183: Crash in BKE_sound_scene_playing due to Scene NULL
Fixes T64192: Crash opening a file with a non-default View Layer active
Added working X-mirroring in pose mode with an optional relative mirror
mode.
Reviewed By: Campbell Barton
Differential Revision: http://developer.blender.org/D4765
- Add missing tooltips to Quick Effects and Subdivide Edge Ring
- Add proper tooltip to the Bevel operator
- Clearer tooltips for the extrude operators
Plus a few other tweaks
Suggested by users on Devtalk.
Was too easy to do by mistake while attempting to zoom.
Unfortunately it seems like it's not currently possible to add zooming to the Knife modal keymap, so this is still missing here.
This can easily happen if adding drivers through Python via
pose.bones[...].bone.driver_add(), e.g. in Rigify code: the
bone field doesn't change id_data, so the driver is associated
with the object ID.
To handle this it's necessary to skip from Object to data in
RNA_Bone-specific code both for generic RNA and in the custom
code for drivers. The latter also had to be changed to use the
proper parsed RNA pointer instead of string matching on paths.
Usually the brush type is correct, but for some old files the brush could be a not valid type. In this case, returns the object active material.
This is related to T61413
Steps to reproduce were:
* Duplicate some area into new window (shift-click corner triangle)
* Make it fullscreen
* Close the window again
* Activate the added screen from the menu (the one without the
-nonnormal prefix)
-> Crash (you may have to press "Back to Previous" first though)
When activating a screen, code should check if there's a fullscreen
variant of it and activate that instead. From what I can tell that's
what the code tried to do, but incorrectly.
Same issue as T64045, but things are a bit different for 2.7.
Note: Some adjustments were made compared to the diff mainly for code
readability and made the default ball size 150px.
Reviewed By: fclem
Differential Revision: http://developer.blender.org/D4793
Workbench/Eevee now displays multiple multi-materials correctly.
Iterate over pbvh nodes when doing object iteration. This makes the
rendering process more streamlined and allow for using different materials.
This change will make possible to:
- Add culling pass of each pbvh leaf node. (speedup if zoomed on a small
area)
- Reduce number of lead node iteration.
- Reduce code complexity
Makes the file browser operator property region (lower left) dynamically
sized to content. Previously, the default size would work really badly
for file browsers opened in small windows, e.g. to install Add-ons from
the preferences window. It would be mostly empty but use lots of space
then.
It was crashing due to array out of bounds access. This is not a great fix,
but brings back behavior the same as before now. Perhaps images that failed
to load should be stored separately somewhere.
Was required to have AV-sync enabled. and then simply play playback
would have triggered an assert.
In release builds the sync would have happened to a wrong frame.
Quite straightforward change, which makes it so audio handles are
only created inside of evaluated datablocks.
Exception is adding sound strip to the sequencer, which needs an
audio handle to query length and number of channels. This is done
by temporarily loading sound file into an original datablock, and
then tossing it away.
There is an assert in sound.c which verifies that audio system is
used from an evaluated domain, which should help porting all the
cases which are likely missed by this commit.
Some annoying parts:
- `BKE_sound_update_scene()` is iterating over all bases, and does
special ID tags to see whether sound has been handled or not
already. This can not be done the old fashion now.
Ideally, this will be done as a speaker datablock evaluation,
but seems that would require a lock since audio API is not safe
for threading. So this is not a desired way i'd say.
Possible solution here would be to iterate over ID datablocks
using dependency graph query API.
- Frame jump needs to call `BKE_sound_seek_scene()` directly
because there might be some flags assigned to the scene which
could be clear after operator execution is over.
Need to verify if that's the case though. This is a bit hairy
code, so sticking to a safest and known to work approach for
now.
- Removed check for format when opening new sound file.
Maybe we can have some utility function which queries channel
and duration information, leaving the caller's code clean and
tidy.
Tested following cases:
- Adding/removing/moving sequencer's sound strips.
- Adding/moving speakers in viewport.
- Rendering audio.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4779
Needs to be done in order to localize sound handlers to the evaluated
IDs only. This commit might not be fully optimal, since it does more
things on every scene update request, but that will be solved by the
upcoming change which will put those updates to a dependency graph.
Just adds sequencer operation and links possible sound ID to it.
No functional changes, just moving towards sound system integration
into the dependency graph.
Currently those IDs are not covered by copy-on-write mechanism since
that ruins the current design of BKE_sound, But this change allows to
move towards system where sound handlers are only valid for an evaluated
ID datablocks.
As mentioned in the comment, the icon width computation relies on
big enough margins; however in compact mode they aren't big enough
and the label gets truncated.
These are the internal changes to Cycles, for Blender integration there are no
functional changes in this commit.
Images are converted to scene linear color space on file load, and on reading
from the OpenImageIO texture cache. 8-bit images are compressed with the sRGB
transfer function to avoid precision loss while keeping memory usages low. This
also means that for common cases of 8-bit sRGB images no conversion happens at
all on image loading.
Initial patch by Lukas, completed by Brecht.
Differential Revision: https://developer.blender.org/D3491
This is the groundwork for supporting loading image textures with arbitrary
color spaces through OpenColorIO.
Initial patch by Lukas, completed by Brecht.
Differential Revision: https://developer.blender.org/D3491
This adds our own OSL texture handle, that has info for OIIO textures or our
own custom texture types. A filename to handle hash map is used for lookups.
This is efficient because it happens at OSL compile time, because the optimizer
can figure out constant strings and replace them with texture handles.
At first you could think that this refactor would not be
necessary, because `ACHANNEL_FIRST` exists already.
It contained the small y offset that all channels had.
Unfortunately, a lot of code assumed that
`ACHANNEL_FIRST = -ACHANNEL_HEIGHT`, making the
define pretty much useless. This refactor fixes that
for the action and nla editor.
As a nice side effect, this patch fixes channel box select.
Before there was always have a half-channel offset.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4783
This was an attempt to speed up motion path calculation, which didn't
really work in real world animation files, where animators already
hide and disable all the heavy collections. Filtering approach also
doesn't allow to have multiple frames evaluated in multiple threads
easily.
Filtering also adds extra complexity on keeping the graph in a correct
and consistent state.
Fixes T64057: Blender crash when use motion paths
- Use Display As rather than Draw Type, following naming conventions
- Rename Advanced subpanel to Stretching, since that more accurately describes what it's for
This reverts commit a01bcfa636.
This causes MSVC2019 build to fail with error C2059
Unfortunately this means we have to put up with bad formatting
in Python structs.
Users have requested to have the Annotate tool added to Sculpt Mode. It can be useful for annotating while sculpting.
- Remove Measure tool from the tools_annotate group
- Add Measure tool to object and edit mode toolbar
- Add Annotate tools to Sculpt and Paint modes
Differential Revision: https://developer.blender.org/D4785
This was only used once, other checks were masking out
RGN_SPLIT_PREV which isn't future proof (if other flags are added).
Add RGN_ALIGN_ENUM_FROM_MASK macro, use it everywhere we need to
check the alignment enum.
This is required for some add-ons
Example use:
ma = bpy.data.materials[0]
# create settings
bpy.data.materials.create_gpencil_data(ma)
# remove settings
bpy.data.materials.remove_gpencil_data(ma)
Related to T63707
- Add `render_aa` and `viewport_aa` sampling setting for workbench. 0
samples means no AA, 1 sample uses FXAA and more samples will use
TAA.
The viewport `gpu_viewport_quality` can still limit viewport anti-aliasing
method.
- Use TAA when rendering images. (this used to be CPU based FSAA)
- Removed `R_OSA` related settings.
Reviewers: fclem, brecht
Maniphest Tasks: T60847
Differential Revision: https://developer.blender.org/D4773
Same changes as for Dopesheet and Graph editor:
- Dragging in empty area starts a box selection
- Hold Shift to expand or Ctrl to remove
- LMB Dragging on keyframe moves key
- MMB Dragging anywhere moves selection
* Fix operator adjust settings not working.
* Fix modal operator not working when invoked from search menu.
* Fix tagging to be correct for new depsgraph.
* Fix pose relax doing nothing when start and end frames match
even if the current frame value is different.
* Remove odd 0.3..0.7 limit in adjust operator settings panel.
Can not use evaluated datablock to localize since that could point to
another evaluated datablock, which can not become part of another
dependency graph.
The original code needed to have unkeyed changes preserved, but now
we do have a flush of animation to an active dependency graph, so this
code is not needed anymore.
- Dragging in empty area starts a box selection
- Hold Shift to expand or Ctrl to remove
- LMB Dragging on keyframe moves key
- MMB Dragging anywhere moves selection
This flag becomes stuck in certain situations, causing the marker
camera switch feature to not work.
To fix old files with the problem, clear the flag on file load.
Also, the line that is supposed to clear it has wrong bit math,
causing seq_render_scene_strip to effectively toggle the flag on or
off every time it's run, instead of restoring to original state.
The flag and code were added in 001789d733 by @campbellbarton.
Only draw the depth buffer when overlays are enabled and scene
or view has changed.
When using Cycles in the viewport for every viewport draw call
the depth buffer was renewed. Draw calls happened when a sample
was finished or the status report was updated.
This could waist some CPU/GPU cycles. This change will check when
the depth buffer needs to be updated or when the last known depth
buffer could be reused.
Reviewers: brecht, fclem
Maniphest Tasks: T63525
Differential Revision: https://developer.blender.org/D4775
This panel is arguably the most important in the ViewLayer properties, so
the concept of "1 panel open per context" doesn't work. Especially since
the first panel (View Layer) contains only two settings.
This also registers the Passes panels before filter/override so it's sorted in the same way as EEVEE.
This panel is arguably the most important in the ViewLayer properties, so
the concept of "1 panel open per context" doesn't work. Especially since
the first panel (View Layer) contains only two settings.
This affects the timeline, dopesheet, graph editor, sequencer,
clip editor and nla editor.
Removed structs and enums: `V2D_ARG_DUMMY`, `eView2D_Units`,
`eView2D_Clamp`, `eView2D_Gridlines`, `View2DGrid`.
A main goal of this refactor is to get rid of the very generic
`View2DGrid` struct. The drawing code became very complex
because there were many different combinations of settings.
This refactor implements a different approach.
Instead of one very generic API, there are many slighly
different functions that do exactly, what we need in the
different editors. Only very little code is duplicated,
because the API functions compose some shared low level code.
This structure makes the code much easier to debug and change,
because every function has much fewer responsibilities.
Additionally, this refactor fixes some long standing bugs.
E.g. when `Show Seconds` is enabled, you zoom in and pan the view.
Or that the step size between displayed frame numbers was
always `>= 2`, no matter how close you zoom in.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4776
This new property allows to check if the current datablock is an annotation or is used in a grease pencil object.
This property is required for some Add-ons.
A spelling issues introduced in 5a43406e1b. From quick look the
sequencer seems to be the only special boy. But the change is a bit
too big to carefully re-read for exact spelling mistakes.
when the scene is referenced as a strip from another VSE scene
Fix T49658: Evaluation / animation of f-curves does not correct
for a scene's position within another scene
Solution: Evaluate animdata before rendering scene seqbase.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4755
Otherwise things can become confusing due to copying unwanted curves,
especially if "Only Selected Curve Keyframes" is enabled, and thus
selected keyframes from other curves may not be visible.
Now Copy Keyframes behaves exactly like Paste Keyframes.
This removes the vertical bar separators between items and subitems (for closed
rows), and the vertical lines between the columns of restriction icons.
The vertical bars and lines don't really serve any useful purpose and add some
visual fuzziness and confusion. Best to concentrate the eye on the items of
content, not the separators between them.
Differential Revision: https://developer.blender.org/D4743
When rows were highlighted - for selection, hover, or search - the highlighted
bar would not take up the entire vertical space but instead leave a gap.
That gap generally looks like a separator between items, adding complexity and
fuzziness for no real benefit.
Differential Revision: https://developer.blender.org/D4742
* The cache file datablock is now evaluated as part of the dependency graph,
creating/freeing the Alembic file handle matching the current frame.
Modifiers and constraints depend on this evaluation.
* Cache file handles and readers now only exist on COW datablocks, never the
original ones.
* Object data paths are flushed back to the original for the user interface.
* The cache file keeps a list of all readers associated with its handle, and
automatically frees them when the handle is freed. This kind of sharing of
data across datablocks is weak but we have no better mechanism for it.
Fix T62720: Alembic sequences not working and crashing
Differential Revision: https://developer.blender.org/D4774
This partially mitigates the performance drop in Spring
02_055_A.anim caused by dependency changes in D4715.
Differential Revision: https://developer.blender.org/D4753
Disable cache use, when rendering another scene seqbase.
Now cache limit applies only to one scene, so it can be overshot.
Cache of other scenes can be filled manually still.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4757
In 2.7x UV sculpt was a kind of sub-mode
(a toggle with it's own key-map & drawing code).
Move this to an operator that uses the tool-system,
this simplifies internal logic, especially brush selection
which now matches sculpt and other paint modes.
- Remove toggle used to enable uv sculpt.
- Expose the brush, which was already used but there was no way to
select different brushes.
- Make UV sculpt use paint paint tool slots
(using brushes how all other paint mode currently do).
- Move UV Sculpt keymap to the tools keymap.
- Remove Q to toggle UV sculpt mode,
S/P/G keys to switch tools.
Make Lookdev works with bloom and TAA by rendering it before TAA and fixing
the motion vectors of the lookdev balls.
Rework Lookdev to remove much of its complexity. Use simpler matrices with
more understandable setup code.
Viewport now displays alpha checkerboard pattern like Cycles does when
film alpha is set to "Transparent".
Some small workarounds were necessary for Depth of Field and correct TAA
support.
1. No need to repeat the title inside the panel (this was needed when we
had this as a popup menu).
2. No need to call it "Collections Visibility" since we are already
under the "View" category. It is "Collections" now, short and sweet.
This will make it easier to move the markers area somewhere
else (to the top) in a separate step.
There should be no functional changes, only minor UI changes.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4745
This is a continuation of rB7fdffd735ff24, where I separated the
e.g. frame number drawing from scrollers internally.
This patch changes the API, so that space draw handlers
have to draw these numbers explicitely.
This greatly simplifies the scrollers API for all spaces
that just need scrollers without any frame numbers.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D4747
- Merged SEQ_OFSDRAW with V3D_OFSDRAW and define in the
DNA_view3d_types: Due to this FSAA always kicked in making the
rendering slow.
- Removed `Texture Solid` and `DOF`.
- Now when chosing Solid rendering the settings
of the original scene is used.
- Added a global override to use scene specific shading. In the
Future we will need to enhanced this so user can change the
settings.
- Added support for LookDev. LookDev crashed as it needed the
`evil_C` what was not set
- LookDev mode will always show the scene + world lights.
Reviewed By: brecht, fclem
Maniphest Tasks: T62517
Differential Revision: https://developer.blender.org/D4738
When this button was still RNA based it was FLT_MAX (that changed to
10000 in rB1b8c3774a86e)
Reviewers: brecht
Maniphest Tasks: T64003
Differential Revision: https://developer.blender.org/D4763
Add special case to construct PointerRNA from evaluated NLA strip
to an original one when flushing animation back to the original ID.
Reviewers: brecht
Maniphest Tasks: T63865
Differential Revision: https://developer.blender.org/D4762
Note that unlike some others, this is always enabled for sequencer,
since previous (2.7x) code was already deselecting everything when
clicking in an empty area...
Part of T63995.
Althought this has limited usability currently (only 'nothing' area in
Outliner are empty lines below last entries), better for consistency to
have it here too.
Part of T63995.
Was unconditionnaly behaving that way, now use proper common setting to
control whether we should deselect everything when clicking on an empty
area.
Part of T57918.
Also moved that operator option from 3d-view op to generic
WM_operator_properties_mouse_select() helper, and renamed its label
(since 'Deselect' is already in use).
Part of T57918.
This is a part of T61296: Crash with animated b-bone segments.
Consider animated/driven bendy bones segments as something what requires
special bendy-bones operation and relation in the dependency graph.
This is because it is more beneficial from a performance point of view
to not build operations if they are not needed. But if the property is
animated it is not possible to make any reliable decision based on just
a property value.
Differential Revision: https://developer.blender.org/D4739
Should be no functional changes, just switching code to use more
generic checks now.
One thing which goes a bit deeper than that is check for whether
base is a part of dependency graph. This is now done by explicitly
tagging corresponding ID node (of an object) rather than doing
animation check again.
Allows to speed up lookups like "is property FOO of data BAR animated".
Can be used to optimize object's visibility check, but also allows to
check animation on bones without too much of time penalty.
The cache is shared between both nodes and relations builder.
Currently is not used, just a boilerplate for an upcoming changes in
an actual logic.
- Move connected & projected into individual toggles.
- Top-level proportional editing button now only toggles.
- Use popover for proportional edit-mode falloff and options.
Note that it's no longer possible to toggle connected via key bindings,
although this could be supported again if it's needed.
Resolves T58081
When the modifier uses vertex groups, the set of the bones it actually
needs is precisely defined by the set of the group names. If envelopes
are enabled, this refinement is not available, because any bone can
potentially be used.
This can be used in the dependency graph construction to allow objects
deformed by a part of the armature to be used in constraints on other
bones, e.g. for placing cartoon-style face elements on top of the body
mesh via Shrinkwrap constraints.
Since the list of vertex group names is now used as an input by
the dependency graph, adding/removing/renaming groups should now
be triggering a graph rebuild.
Differential Revision: https://developer.blender.org/D4715
There was a double loop when drawing the offscreen data for the fill tool. This loop did an exponential drawing, slowing a lot the process.
For example, one file was using 7.39 seconds and now use 1.20 or less to do the same process.
The fix is more noticeable with big files, reducing exponentially the time used by the tool. For very simple files the speed change can be unnoticeable.
Aso fixed a hidden bug with stroke opacity that it was unnoticeable due the double loop.
This hides the top-bar by default for everything
besides paint/sculpt workspaces.
- Use the top-bar mainly for active tool settings &
popovers panel options.
(transform / snap settings are an exception for this convention).
- Only show the top-bar (by default)
in paint work-spaces (sculpt / texture-paint / grease-pencil).
- Add an active-tool panel to the sidebar.
- Split 3D view tabs into (Item / Tool / View).
D4721 with minor changes.
Further work is needed for the top-bar and image-editor.
The driver code was almost there, but didn't work because ID nodes
have no outlinks - and using links won't be safe anyway because of
ordering issues. Instead, just loop over all IDNodes.
Animation is fixed simply by referring to ARMATURE_EVAL instead of
BONE in construct_node_identifier - the bArmature ID doesn't have
BONE components in any case, so the old identifier can't work.
During driver var evaluation, the code directly skips to the proxy
object if present, so the dependencies should reflect that, instead
of inventing extra relations to pretend as if the values were
first copied from the proxy - they actually aren't.
The invented dependencies can't work without cycles in complex but
meaningful cases, e.g. drivers on armature data that depend on pose.
This alters the changes from 72cbf966fb to be more correct.
This macro is handy in at least two occasions:
* When you want to check the optimized generated assembly for
some small function that would get inlined otherwise.
* The marked function will be visible in profiling results.
I can't test the implementations for other compilers now,
so I did not include them.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4754
Some deform modifiers (like armature) use passed Mesh parameter to
retrieve some extra data (vgroups in our case), and default to obdata
when it is not available.
This should be fine in theory in the 'apply modifier to obdata' case,
since this is always the first modifier, but here upper-level code
passes **evaluated** object, not orig one, so ob->data is not orig
anymore, and might miss some stuff...
Note that am quiet unsure whether the way evaluated data is passed
around in that apply modifier operator code is actually OK, but for now
it seems to work at least...
This patch implements new cache system.
Aim is to give user more control over cache, so it can be maximally
utilized. This is done through sequencer timeline side panel
in category proxy & cache.
Cached images are also visualized in timeline, controled by
sequencer timeline view->cache menu
Functional changes:
- NOT use IMB_moviecache API
- refactor names of cached image types
- each scene owns 1 sequencer cache
- merge preprocess cache into per-sequencer cache
- cache links images rendered per frame in order as they are created
- add cache content visualization tool
- add RNA properties to control the cache
More info can be found in design notes in blenkernel/intern/seqcache.c
and in https://developer.blender.org/D4443
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4443
This patch fixes T63794 by updating duplicated bones' custom handle references
to point to the new armature's bones.
The problem occurs because B-Bones of a duplicated armature keep pointing to
the old armature's bones. The references are cleared upon entering Edit Mode
because there are no matching `EditBone` for such bones, and `find_ebone_link`
just returns `NULL` in such cases.
Reviewers: mont29, angavrilov
Maniphest Tasks: T63794
Differential Revision: https://developer.blender.org/D4726
drw_batch_cache_generate_requested() is only needed when the geom needs to be
created.
Went from 37fps to 47fps with artificial testcase (lots of bones with one custom shape). Baseline 2.79 is 24fps.
Also fix the drw_shgroup_bone_custom_wire.
If the draw uses the same id as the previous draw, there is no need to read the depth
buffer at this point, avoiding a CPU-GPU sync bubble.
Fixes T62511 Selection is significantly slower in production scenes.
With this patch glReadPixels is not the bottleneck. Regular drawing Is still very
slow so I would suggest fixing the regular drawing first before trying to
improve the selection algorithm.
If the Tabs as Spaces settings is enabled for the text block. This avoids
issues with inconsistent indentation when pasting Python code from another
source.
Differential Revision: https://developer.blender.org/D4512
screen_geom_vertices_scale already checks if areas are too small for header.
And header size will be set in region_rect_recursive from ED_area_initialize.
Differential Revision: https://developer.blender.org/D4711
Viewport drawing has moved to offscreen buffers, and we no longer need to have
depth, stencil, aa samples, sRGB buffers as part of the window. So all that
code is removed now. The depth buffer was the only one still being allocated,
its removal save a bit of memory.
Code by Germano and Brecht.
Differential Revision: https://developer.blender.org/D4708
Similar to many apps:
- Use left click and drag to box select
- Hold modifiers like Ctrl and Shift to remove or expand selections
- Use MMB-drag to use the tool outside of the gizmo area
In the future it would be nice if the transform tools would have this increased flexibility built-in so you could configure it more easily, but this setup seems to at least make it do the most commonly useful thing by default.
Enabling the drawing of the mesh analysis overlay.
Currently the settings are part of the scene toolsettings. What makes sense,
for 3d printing, but does not fit well with the per viewport blender 2.80
overlays.
Reviewers: brecht, fclem
Differential Revision: https://developer.blender.org/D4707
Modifier previews should be implemented by a more generic system.
The current system is already a hack and needed a lot of work
to get it working again in 2.80 and even so that would be replaced by
another system in the near future.
For Vertex Colors we have a work around in place by using Workbench
Vertex Colors. For Vertex Weights we loose the previewing. Not sure
targetting weight is working (even for 279).
Reviewed By: brecht
Maniphest Tasks: T63857
Differential Revision: https://developer.blender.org/D4734
We always keep the children that are not in a collection listed in the
end of the children list due to design. This way we can visually draw
them with dashed vertical lines.
This was already working for alphabetical sorting, however whenever
sorting was disabled, we would end up with a list of children ordered
regardless of their state (whether the child itself is in the collection).
As known as outliner parenting hierarchy take two.
Implemented suggestion by Brecht Van Lommel:
```
The problem is that it's iterating over te_parent->subtree,
while at the same time removing elements from it as tree_to_remove_objects_from.
Further there is a linear lookup to find tree elements corresponding to a child
object, which causes O(n^2) time complexity overall and so poor scaling for many
objects in a collection.
The more efficient solution that also fixes the crash could be:
* Build a map from Object* to a list of TreeElement* matching the object.
* For all objects in the tree lookup the parent in this map, and move or add
tree elements as needed.
```
I removed the grouping of the children not in collection in the end of
the children list when sorting was not enabled. If we think we really
need it back it can be tackled separately.
That said, despite due to performance reasons, I can't see why would
someone not have the a-z sorting enabled. And if they do, it is not the
end of the world to have interleaved children that are in the collection
or not in the parent subtree.
Force Displist to Mesh conversion if there is any modifier.
This is until we find a better way to store the batches per objects.
Also fix draw cache functions that were not returning final mesh edges.
I also had to make the "New" operator a submenu in the
`File Context Menu`, so that you can still select the template.
This partially solves T61599. Currently the confirm dialog
is not shown when an already existing file is opened.
Implementing that requires a bit more work and will be
done in a separate patch.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4732
Previously only a fixed bounding box could be used.
This was not flexible enough.
T63193 will benefit from this refactor.
Reviewers: brecht, campbellbarton
This never really worked as it was supposed to. The main goal of this is to
turn noise from sampling tiny hairs into multiple layers of transparency that
do not need to be sampled stochastically. However the implementation of this
worked by randomly discarding hair intersections in BVH traversal, which
defeats the purpose.
If it ever comes back, it's best implemented outside the kernel as a preprocess
that changes hair radius before BVH building. This would also make it work with
Embree, where it's not supported now. But it's not so clear anymore that with
many AA samples and GPU rendering this feature is as helpful as it once was for
CPU raytracers with few AA samples.
The benefit of removing this feature is improved hair ray tracing performance,
tested on NVIDIA Titan Xp:
bmw27: +0.37%
classroom: +0.26%
fishy_cat: -7.36%
koro: -12.98%
pabellon: -0.12%
Differential Revision: https://developer.blender.org/D4532
If the "Object Children" filter is enabled, we nest the object children inside
the object. If the child itself is not in the collection, it is grayed out,
connected by a dash line, and its restriction flags and contents are not shown.
If "Object Children" filter is disabled, it works as before.
Note: This is not super fast, but at least we traverse the tree only once to get the
children of an object. That said, there is a lot of loops going on here.
Task T63526.
Development notes:
I could use the GPU_SHADER_2D_LINE_DASHED_UNIFORM_COLOR shader, but
that would mean I would need to iterate over the tree twice (once for
each shader) - or do some bigger refactor.
Also I could not get that shader to work. This shader expects float
vertices while the current one is using integers, so converting the code
would make the dash line drawing to diverge from the regular lines even
further.
Differential Revision: https://developer.blender.org/D4696
Shader nodes are now shared with Eevee, so makes more sense to have it in
the core and not be Cycles specific.
Fix T62415: issues with append/link of old Cycles settings.
On file read we need to update group nodes in case the group they refer to
has changed its inputs and outputs. This had O(n^2) time complexity and was
updating all datablocks even if they did not change.
Forward compatibility with that version is already long gone, and removing
it means we can avoid running some complicated code on every file read/write.
Node versioning code was added before there was a mechanism to do versioning
after lib linking. Now integrate with that system and make it less of a strange
exception. Node versioning is now skipped on undo, like other versioning code.
This is a first step towards T61599.
This way the invoke function can be used for the confirm
dialog in a separate patch.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D4723
The offset added allows to generate a vector to determine direction. This direction will be used when rotate the object to rotate texture.
The solution is not 100% perfect, but it's far better that having an unpredictable rotation.
Each spaces top-bar wasn't showing it's own active tool,
Remove RNA access to the workspaces tool since using it is error prone.
Eventually this should be completely removed.
When OSL is enabled, Cycles disables RTTI in some of its modules, which
then breaks vptr sanitizer (part of the 'undefined' sanitizer).
thanks to @brecht for helping tracking down the issue.
If image buffer is not loaded and blender attempts to reload it (during
`BKE_image_acquire_ibuf`) over and over for each frame rendered.
When attempting this reload, image_load_image_file is calling
`BKE_image_free_buffers` and tag the Image to the (GPU) image_free_queue
(because this run on the rendering thread).
If the main thread decide to redraw the UI and go through `GPU_free_unused_buffers` they all get deleted and if that happens before the rendering thread use them ... segfault.
If I replace the environment textures with correct ones (the file does not seems to contain them), there is no crash when rendering.
I used a list of GPUTexture from blender Image to increase and decrease the
reference counter correctly.
This add very little memory and computation overhead.
Since it is a continuous cycle, there's no need to repeat the
name of the previous bone. Also, dot is a common symbol in object
and bone names, so use '/' instead for node nesting.
Add a callback to allow custom node editors to draw their own
backdrop.
Differential Revision: https://developer.blender.org/D4709
Reviewed by: JacquesLucke
Targeting a different object always requires its transform,
and normally dependencies should go to the solver node.
ITASC is quite broken so special case it until fixed.
As far as I can tell, there is no technical reason why the B-Bone
segment thickness scaling can't be separated into two axes. The
only downside is the increase in complexity of the B-Bone settings,
but this is inevitable due to the increase in flexibility.
Updating the file is somewhat complicated though, because F-Curves
and drivers have to be duplicated and updated to the new names.
Reviewers: campbellbarton
Subscribers: icappiello, jpbouza
Differential Revision: https://developer.blender.org/D4716
Currently it is not possible to view the vertex colors of an object. To
optimize the workflow, workbench will need to support Vertex Colors.
The Vertex Colors is a new option in `shading->color_type`. When objects
do not have vertex color, the objects will be rendered with the
`V3D_SHADING_OBJECT_COLOR`.
In order to support vertex colors in workbench the current texture/solid
shading structure is migrated to a primary shaders and fallback shaders.
Fix: T57000
Reviewers: brecht, fclem
Differential Revision: https://developer.blender.org/D4694
I did not rename the other property to `show_expanded_dopesheet`
yet (as suggested in the report), because:
* Would break compatibility (haven't found any addon using it though).
* I'm not sure if this really only affects the dopesheet.
- Resizable areas use 2D view bounds.
- Header uses the button bounds.
- A margin is added to avoid clicking between buttons.
- Region resize edges clamp to the 2D view bounds.
Resovles T61554
Some GLSL compilers seem to not have problems with \ to break preprocessor
directives. I couldn't find other places with similar code, but fixing this
case by case is not ideal and the same issue may come up again.
This adds a new "Automatic" image display method which uses GLSL shaders for
most images. It only does CPU side color management for higher res images
where sending big float buffers to the GPU is likely to be a bottleneck or
cause memory usage problem.
Automatic is the default now, previously it was 2D Texture.
If you had several windows or, after last changes in topbar, several areas, the swith of overlays or Onion Skin was not working by area, but as whole switch, so it was impossible to have different status by window.
This commit tries to fix the rotation problem when the parent is at layer level.
The problem was the object location was not used, so all object not in origin got weird transformations.
The problem occurs because status changes between BackBuffer and Offscreen.
Reviewers: fclem
Differential Revision: https://developer.blender.org/D4703
There is no more sorting happening, so the comment is out of date
by manu reasons. If something extra is needed there it would need
to be done differently anyway.
This improve visibility in edit face select mode by using the face select color
instead of the edge select color (which is in default theme a bit more red). Also
makes the selected edges in this mode a bit more opaque (0.75 instead of 0.4).
Full opacity is still reserved for edge select mode.
This is only working for shading batches for the moment and only if some Custom data layer are not needed anymore.
The collection rate is hardcoded at 60 sec but could be exposed to the user.
This system can be extended and discard most unused batches in the future.
This commit is in prevision of removing BKE_MESH_BATCH_DIRTY_SHADING when changing shader parameters.
Adding this back was a frequent request.
Now the panel can be hidden - some users may prefer only to access this
via a shortcut.
Unfortunately this uses an F-Key we'd intended to keep free for users.
This still keeps the middle block of F-Keys (5..8) free.
Now that B-Bone shape data is kept in bPoseChannel_Runtime, the
armature level cache only holds one quaternion value per bone.
It can also be moved to runtime, and the structure removed.
This has an additional effect that, as far as I can tell, now
the Armature modifier can run as soon as all of the bones it
actually needs are done, thus making T59848 a purely depsgraph
level problem.
It makes no sense to load add-ons here, we already do that (in a more
complete way) in load_addons test, this is only adding overhead and
doubling code to maintain).
Also do not try to load-as-modules add-ons that are not 2.8-ready, and
some other misc fix.
load_py_modules test should be passing again now.
Thanks to @sergey who did part of the work here as well.
This way float and multilayer images can now be packed without data loss. This
removes the as_png option and always uses the appropriate file format depending
on the image contents.
The `AMD Radeon HD 7600M` compiler is usually buggy for Geometry Shaders.
In this case, indexing was causing problems in `gl_in[i].gl_ClipDistance[0]`.
Reviewers: fclem
Differential Revision: https://developer.blender.org/D4700
Allowing direct import of text blocks isn't especially useful,
instead add `text.as_module()` script authors can do this
explicitly if it's needed.
Now the text "Register" option executes
instead of loading as a module.
This removes the need to keep track of the current Main,
and C code to override Python's import & reload.
Expanding enum items in a small area would render the names with
dots, what is a bad solution. When you use expand directly on a
`grid_flow` or `column_flow`, it will render the items on multiple
lines, giving more control to the developer.
Reviewers: campbellbarton, brecht, mont29
Differential Revision: https://developer.blender.org/D4698
Each space had separate operators, duplicating logic.
Use RNA properties instead so adding the ability to toggle other
region types (floating redo region for eg) doesn't need to have an
extra operator per space type.
It's also nicer to show a check-box for something which can be toggled.
This is an odd-ball: it's a library which has own style and
guidelines, and just happened to be developed by Blender developers
and also happened to rely on some functionality of intern/ for its
C-API.
Might consider using Google's clang-format in the future (this is
what the style is supposed to be in this library).
For some reason the buildinfo header was not re-generated. The root
reason is not really clear to me, so simply remove the header similar
to the CMake cache.
When in Axis alligned orthographic view a grid was always displayed.
With this change the user can enable/disable this grid.
The Grid is always visible and editable, but only rendered active when user is in quad view, or axis aligned ortho view.
Reviewers: brecht, fclem
Maniphest Tasks: T63517
Differential Revision: https://developer.blender.org/D4699
This removes the large edges and instead use colors to hint in which
selection mode the user is.
The component in each individual selection mode is more prominent to add
more emphasis. The other components are less prominent and dimmed.
A minor default theme changes needed to be made to keep clarity in all
combinations.
Forcing old behavior (no selection mode hit) could be done quite easily
by just making the 2 booleans true (selectFaces and selectEdges).
Reviewers: campbellbarton, billreynish
Reviewed By: campbellbarton, billreynish
Subscribers: ThinkingPolygons
Maniphest Tasks: T1234
Differential Revision: https://developer.blender.org/D4526
Previously, when a fcurve modifier used storage,
many heap allocations were done.
This caused major slowdowns as described in T63656.
Furthermore, the storage usage was a special case only
used by the Cycles modifier. This refactor makes
storage usage the "normal" case.
That reduces the overall complexity.
The storage is stack allocated now.
The framerate on the provided test scene went up from ~5 fps to ~16 fps.
Reviewers: angavrilov
Differential Revision: https://developer.blender.org/D4701
When the same .blend file was open in two instances of Blender,
the autosaved files would overwrite each other.
The solution is to put the pid into the filename.
The exact pattern is the result of a discussion with @pablovazquez.
Ripping edges would in some cases rip verts on unselected edges. This is now fixed so that we only rip verts from the actually selected edges.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4646
While fixing it, also reproducing multi-object selection logic we have
in the viewport. If you select one bone from an armature, it deselects all
bones from the other armatures currently in edit mode.
Inspired by proposed fix by Philipp Oeser (lichtwerk).
Code explanation
================
CTX_data_edit_object(C) may be NULL here. In the few cases where it is
not we use it for the notifier (in the other cases we call the notifier
with NULL, so no harm done).
This code is called during outliner drawing, and it is called, for
instance every time you mouse hover a bone.
If you have a mesh object in edit mode and mouse hover the bones of an armature
it is plain obvious we can't rely on CTX_data_edit_object() for anything
armature related. And that the original assert was asking for troubles.
Previously Spline IK provided only two choices: either scale the
length of the bone chain to fit the length of the curve, or don't
scale the bone in the Y dimension at all (ignoring effects of
actually fitting to the curve due to curvature and curve object
scale).
This patch adds a new option to use the pre-IK Y scale of the
bones to adjust their length when fitted to the curve, allowing
individual posing control over the length of the segments.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4687
Outline shaders were only selected when no checker depth was available.
Reviewed By: fclem
Maniphest Tasks: T63312
Differential Revision: https://developer.blender.org/D4661
Creating a fully functional custom property requires also setting up
its limits, description, default value and static override status.
It is complex enough to warrant providing an official utility.
Currently boolean properties are technically int, but the utility
pretends they are separate in case that eventually they are.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D4620
There are couple of caviats currently:
- The script requires system-wide Python 3 available in the current
search PATH as python.exe.
This will get addressed soon by distributing unpacked Python binary
in our libraries.
- Since the libraries folder is to be known, this requires to have
MSVC detected. Not too bad, since formatting is still way slower
than detection, but still doesn't feel ideal.
This also fixes alignment issues in many editors.
E.g. the frame numbers were badly aligned to the grid lines
in the timeline.
Checkout the images in D4681 for more examples.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4681
Add back the ability to choose the select operation for random select.
Now we allow the operator to specify if the select operation should be
visible in the GUI or not.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4665
Use CMake's target_link_libraries instead of manually maintaining
library dependencies in a single list.
In practice adding new libraries often ended up being guess-work,
now each library lists the libraries it uses.
This was used for the game player executable so libraries
could optionally link to stubs.
If we need this functionality it can be done using target-properties
as described in T46725.
Following user suggestion
- Similar to some apps
- Works better on laptops and desktops without numpads
- Also fix some small inconsistencies with select more/less in Dopesheet
Blender
Crash was introduced in rB48e48653c297, rB90f4ff95ed3a
`image_sample_invoke` is used for both `IMAGE_OT_sample` and
`IMAGE_OT_curves_point_set`. IMAGE_OT_curves_point_set didnt have
property `size` defined...
Now defines `size` for `IMAGE_OT_curves_point_set` as well
Reviewers: brecht
Maniphest Tasks: T63492
Differential Revision: https://developer.blender.org/D4675
Having this in blenkernel caused bad level calls to bf_editors_anim,
causing tests that use 'blenkernel' to require almost all libraries.
(complicating gtest linking & causing large binaries).
Mainly, scaling needs to be continuously faded out in these cases.
Plus, in the case of end roll-off, it seems better to integrate into
this general fading system, instead of ad-hoc interpolation of tail.
Although technically evaluating the proxy copy nodes doesn't require
the parent bone to be copied, other nodes that depend on the bones
sometimes expect the parent to be ready. To meet this expectation
it's necessary to add the dependency to the graph.
I'm not sur how it ever worked before. There was no texture bound to these
sampdensity and sampflame when no volume simulation was happening.
This fixes the issue using 1x1x1 dummy textures.
- Use C for Cursor tool
- Use D for Annotate tool
- Use tool shortcuts in UV Editor (although it's weak here still - waiting for Move, Rotate & Scale tools here)
This commit adds support for drawing Dots strokes with a gradient factor to get artistic effects like watercolor.
Currently, the option is only supported by Dots materials, and in the future will be added to line strokes, but now there is a limitation on drawing engine and we will keep disabled on Line materials.
Also, added the option to align Dots and Boxes strokes textures aligned with the drawing path to get more fluid strokes.
As reported in T63582, it can cause chan_array to be not ready.
To reliably avoid crashing, the only easy way seems to be to
create the index during COW -- maybe @sergey has a better idea.
* EEVEE support through irradiance volume and light probe.
* New shader ball shape (designed by Robin Marin).
* New cloth and liquid shapes, removed monkey.
* Replace world sphere by toggle to use world for any shape.
* Slight bevel on cube.
* More subdivision for displacement preview.
* Fixed and improved UV mapping for all shapes.
* Material icon / asset preview now uses specified shape instead of always
a sphere. So for example hair material can be displayed as hair.
Ref T57683
The bug is caused by problems in the dependency graph. Unfortunately,
fixing is not just a matter of fixing the graph, because correct
dependencies would cause a cycle here and in other reasonable use
cases. The real fix thus requires refactoring Spline IK to require
curve data only in the actual evaluation phase, and not in POSE_INIT_IK.
In addition, this separates the normal bone evaluation loop from
Spline IK computations for two reasons:
- That still needs to be done even if spline IK can't evaluate
due to missing curve data.
- It should reduce issues with induced shearing, as Spline IK now
controls how parent-child relations are handled in the chain, and
can take care to only carry over rotation and location.
No functional change, this adds LIB definition and args to cmake files.
Without this it's difficult to migrate away from 'BLENDER_SORTED_LIBS'
since there are many platforms/configurations that could break when
changing linking order.
Manually add and enable WITHOUT_SORTED_LIBS to try building
without sorted libs (currently fails since all variables are empty).
This check will eventually be removed.
See T46725.
Make the option enabled by default, and enable it in old files,
unless Show Extremes is enabled.
However, suppress it for the Timeline, because its View menu
doesn't include the option.
Reviewers: billreynish, brecht
Differential Revision: https://developer.blender.org/D4678
Just saying "parent transformation added" doesn't really explain
what this coordinate space does. Thus it seems almost nobody used
it and serious bugs like T61981 went unnoticed.
When a tool is activated in many cases the gizmo is needed so toggling
them isn't so useful.
Change behavior to use the active object instead so users who prefer
to enable gizmos in the popover can toggle them.
Also remove duplicate keymap item.
The popover for displaying gizmos now has options for
empty, light & camera gizmos.
Transform toggles are now checkboxes, which has the advantage that it's
more obvious they can be mixed however changing from one to another now
uses 3x clicks instead of a single click-drag action.
We might need a faster way to switch between transform types.
Based on feedback from animators, this is useful to keep as a view
option (as in 2.7x).
Now the transform gizmos can be enabled from the popover,
the tools still work for location/scale/rotation.
The transform tool has been removed.
See T63518
Previously B-Bone deformation mapped every vertex to just one
B-Bone segment. This results in abrupt transformation differences
between the sides of each threshold plane, reducing the quality
of B-Bone deformation and making the use of shape keys impractical.
This commit replaces this approach with a linear blend between
the two closest segment transformations, effectively representing
the B-Bone as two weight-blended plain bones for each vertex.
In order to distribute the interpolation more evenly along the
bone, segment matrices for deformation are now computed at points
between the segments and at the ends of the B-Bone. The computation
also uses the true tangents of the Bezier curve for the orientation.
The nodes at the end of the bone require some special handling to
deal with zero-length Bezier handles caused by a zero ease value.
The Copy Transforms constraint now also smoothly interpolates
rotation and scaling along the bone shape when enabled.
The initial version of the patch was submitted by @Sam200.
Differential Revision: https://developer.blender.org/D4635
The check of DUPLI object was used, but as the original object was not in the scene, the VBO data was not available and the draw was empty.
Now, the function checks if the original data is in the scene, and set the object as not Dupli to generate the VBO data.
The Alembic importer can now deal with XForm-less objects. Apparently
MeshLab exports point clouds without parent transformation matrix (XForm in
Alembic-speak). We shouldn't warn about this, but handle it properly
instead.
Draco py binding needs to be installed somewhere, when not installing
Python itself it's breaking the installation (since it creates a fake
empty py install, which will crash when trying to start Blender).
We could fix that in some smarter way maybe, but for now it's simpler to
just not care about Draco when we are not installing Python.
- Group tabs together based on the data-flow
- Puts Physics and Particles next to Modifiers, with which they interrelate
See D4612
Reviewers: pablovazquez
Do not gray out shading items if the xray alpha value is 1.
They are rendered correctly in viewport so there is no reason to mark
them as disabled in the shading menu.
Edit mode shows the mesh elements in X-ray mode even if alpha is set to
1. Now the code takes this into account so that you can still select
visible mesh elements in X-ray edit mode.
view3d_draw_legacy need to be updated with the new XRAY flag macros to
avoid crashes. Additional cleanup of the XRAY macro flags were done.
The `AbcMetaballWriter` now subclasses `AbcGenericMeshWriter` instead of
wrapping an `AbcMeshWriter`. `AbcGenericMeshWriter` was created for this
purpose (writing non-mesh objects as mesh to Alembic) and performs the work
in a cleaner and, more importantly, not crashing way.
The metaball 'mesh' is animated whenever any of its components are animated.
However, it's not easily possible to visit all its components (AFAIK it
requires a full scan of all objects in the scene and comparing names), so
due to the blobby nature of the metaballs I now simply assume they'll be
animated.
This fixes an issue where a metaball was considered static when the base
ball was static. For example, MBall static but MBall.001 animated would
be considered 'not animated'.
Draco is added as a library under extern/ and builds a shared library that is
installed into the Python site-packages. This is then loaded by the glTF add-on
to do mesh compression.
Differential Revision: https://developer.blender.org/D4501
this was very similar to T60684 (and the UV case was actually reported
back then as well), so the fix goes hand in hand with rBbd8039399575
Reviewers: fclem
Maniphest Tasks: T63464
Differential Revision: https://developer.blender.org/D4672
Before, there was a hacky code for Image editor when the image was empty.
Now, all 2D editors are locked to view and when you Move or Zoom, the stroke moves.
This is a fix of T63402.
Still pending if we add more options to 2D editors, but at least now, we have something consistent.
When using subsurf (and other modifiers) the edit flags are not
propagated correctly. Currently we assume to read the edit flags
from the original object which is kind off hinding the real issue.
Modifiers use `mesh_new_nomain_from_template_ex` to create a copy
from an existing mesh. this method is only used by modifiers. So
by placing this we will make sure that editmesh is propagated.
Reviewed By: fclem, sergey
Maniphest Tasks: T62449
Differential Revision: https://developer.blender.org/D4666
Unfortunately we can't use insert/delete_keyframe operators in keymaps
because the enums aren't known at the time of keymap registration
and the keying sets are dynamic and use a poll function.
Add a version of insert/delete operators that takes a string
instead of an enum. Needed for D4626.
Also extract int to keying-set into a utility function.
Was mainly visible with high-res image. Not entirely clear why it is
only happening on macOS. While the entire function should be re-written
to make float images displayed faster, still nice to fix the drawing.
Reviewers: fclem
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D4664
Now the depth order of objects when trying to select them is only used when not in X-ray mode.
Before, this was only the case in wireframe mode regardless of X-ray settings.
I've also unified the usage of V3D_XRAY and XRAY_FLAG as they were basically copies of each other.
Reviewed By: Clément
Differential Revision: http://developer.blender.org/D4504
- The ambiguous term 'handle' was used where 'ensure'
is typically used (get or add when missing).
- Rename `current` to `active`, all `current` functions which were
also ensuring.
- Clarify what is being operated on, using `BKE_gpencil_object_*` for
objects, `BKE_gpencil_brush_*` for brushes.
replaced check for (obsolete in 2.8) objects 'SELECT' flag with check for
bases 'BASE_SELECTED' flag
Reviewers: brecht
Maniphest Tasks: T60821
Differential Revision: https://developer.blender.org/D4606
If the Brush had a pinned material with texture, but the material slot was in a different material, the texture was not working.
The material was not recovered from brush, but from object.
GCC 9 started implementing the OpenMP 4.0 and later behavior. When not using
default clause or when using default(shared), this makes no difference, but
if using default(none), previously the choice was not specify the const
qualified variables on the construct at all, or specify in firstprivate
clause. In GCC 9 as well as for OpenMP 4.0 compliance, those variables need
to be specified on constructs in which they are used, either in shared or
in firstprivate clause. Specifying them in firstprivate clause is one way to
achieve compatibility with both older GCC versions and GCC 9,
another option is to drop the default(none) clause.
This patch thus drops the default(none) clause.
See https://gcc.gnu.org/gcc-9/porting_to.html#ompdatasharing
Signed-off-by: Robert-André Mauchin <zebob.m@gmail.com>
The hotspot was generated by mesh_cd_layers_type_overlap who was testing
way more data than it should have.
Here we reduce the whole CD layer mask to a 32bit bitflag that is easily
or'ed and tested.
Bonus point: We use atomic operation to merge the mask in order to allow
future multi-threading. (this was a TODO)
In the scene attached to T58188 this removes 5% of CPU time.
release_texture_slots() and release_ubo_slots() were one hotspot when
drawing taking ~9% of total CPU counters for no reason.
This was because of the loops using GPU_max_textures that was overkill and
slow.
Replace those by a simple 64bit bitwise OR operation.
Move free callback call to GPU_batch_discard to prevent the crash.
The issue was that clearing can happen after referencing to an instance
buffer and that's perfectly legal.
This is something where there is no single correct behavior,
sometimes it's needed to ignore the crease to make mesh more
smooth. But sometimes crease is to be considered after first
subdivision surface: for example, when adding extra subdivisions
for render-time displacement.
Made it an option whether modifier needs to take crease into
account or not.
Existing files should be openable in the 2.7 compatible way,
to re-create an old behavior the options is to be manually
disabled in the modifier settings.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4652
In the case of the report a GL_PROXY_TEXTURE_2D_ARRAY of 2509x2509x1 failed to be allocated.
This is a work around as the GL_PROXY_TEXTURE_* is not reliable.
Reviewed By: brecht, fclem
Maniphest Tasks: T63223
Differential Revision: https://developer.blender.org/D4651
* It can be hidden by dragging it up/down.
* It can be at the top or bottom, independent of the header.
* It uses the color theme from the header.
* It does not change its color, when the area becomes active.
Currently, it is used in the text editor to display the file path.
Differential Revision: https://developer.blender.org/D4601
When drawing in big files, the first points of the stroke were not smooth because the system was doing a copy of the depsgraph datablock.
Now, the depsgraph is not updated at the beginning and the feeling is far better, especially for big files.
To avoid the copy, the original datablock is used while drawing, because it's faster the lookup of the original data, than a full datablock copy.
Also some cleanup of the code.
When texture drawing was enabled the fallback materials were not transparent.
The reason was that the solid material was requested. This patch will requesting
in this case the material from the forward renderer.
Fix: T63077
Reviewed By: fclem
Maniphest Tasks: T63077
Differential Revision: https://developer.blender.org/D4641
This was not working well, because the search text was removed
after pressing this button. Finding the item that was inserted
was not easy.
Removing the option seems to be the best solution for now.
In final rendering mode the display type was tested for
visibility flag. This should only occus when doing viewport
rendering
Reviewed By: fclem
Maniphest Tasks: T63056
Differential Revision: https://developer.blender.org/D4643
In a separate step, all new functions should be added.
However, the best way to do this, is not clear yet.
A list of functions, that have to be added is available
in D4280.
Differential Revision: https://developer.blender.org/D4280
A regression since 64c8d72ef1.
The solution is to force modifier evaluation for an evaluated
object, and let it to copy binding data back to original when
is being evaluated for binding.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4642
The collection instance field is currently using flow.column(), which makes the name of the collection instance unreadably small. {F6898084}
I get that we want to save vertical space to avoid scrolling, but in this case it generates a new problem, so I'd rather avoid flow here.
Patch by sebastian_k
Reviewed By: billreynish
Differential Revision: https://developer.blender.org/D4614
Missing dependency, which made collission to access object transform
prior it was evaluated.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4636
The problem was that tangent layers (being computed on demand in the batch
cache generation) were generated but if MR_DATATYPE_LOOPTRI was not present
(in some cases) and the looptri_len was 0, leaving the CDATA layer cleared.
The new implementation uses WM_POINTERDOWN, WM_POINTERUP and WM_POINTERUPDATE
and the pointer API to process stylus events. This avoids the delays that comes
with the WM_MOUSE and WM_xBUTTON events. The implementation should work on
Windows 8, and Windows 10 with both legacy and new pen interaction.
It also changes how the pressure is reset when the Windows Ink implementation
is enabled. The previous version reset the pressure to full when the pen left
the screen, however for some hardware implementations this allowed a small
window where Blender may process the final move event and read the pressure
as full leaving a dot on the last event.
Differential Revision: https://developer.blender.org/D4314
Previously Automatic tablet API mode would handle both Windows Ink and
Wintab events. This is unpredictable and causes problems with the fix
coming in the next commit.
Instead assume that in most cases where Windows Ink is desired there
will be no Wintab. If that's not the case, it can be adjusted under
Preferences > Input > Tablet.
This fixes popups and preferences appearing behind the main window, as
well as flickering artifacts when Blender is in fullscreen.
Patch contributed by phocomelus.
Differential Revision: https://developer.blender.org/D4634
This way you can't accidentally close Blender when toggling the console
window to be visible. When Blender is started from the command promt the
close button remains.
Differential Revision: https://developer.blender.org/D4627
Allows to preserve topology cache even when copy-on-write is
fully re-copying the object. For example, toggling edit mode
in and out does no longer causes full topology to be re-created.
* When resizing sidebars, don't collapse when the region becomes too big but
instead clamp the region size to the available space.
* Fix clicking the tab to expand sidebars no working if the sidebar is too
wide to fit. Instead make it less wide so it does fit.
* Fix incorrect limit on tool properties region height, for example in the
file browser.
Differential Revision: https://developer.blender.org/D4611
Random place in the modifier stack can not be referenced,
so it doesn't make sense to sue GEOMETRY component as a
FROM operation.
So now drivers on modifiers are driving GEOMETRY component,
but are using PARAMETERS as a source for variables.
A new parameter in the layer adjustment panel allows to define the color of the channel in Dopesheet.
This is needed when there are a lot of layers.
See D4623 for more details.
Caused by own recent rB17c15798c35f33e (already a fix in that code).
We cannot erase immediately master_collection's childrn list, as it is
used in sub-code to check in how many scenes an object is instanciated.
Further more, we only want to do the remove old/add new children
collections in case we are actually duplicating them.
Makes me even more eager to nuke that whole piece of code and rethink
from scratch that kind of ID handling. Some day...
We were already getting the designated output node in
'ntreeGPUMaterialNodes()' but this wasnt used in 'ntreeExecGPUNodes()',
instead whatever node was tagged NODE_DO_OUTPUT was executed.
note: this is just the bare minimum to fix the bug, other improvements
previously done in D4482 might follow as a separate commit.
Reviewers: brecht, fclem
Maniphest Tasks: T62434
Differential Revision: https://developer.blender.org/D4630
'Linked objects' option was not behaving correctly before, effectively
linking in collections, so this one has been renamed to just 'Linked Copy',
and gives a fully shallow copy of current scene.
'Linked Obdata' was not really useful, kind of confusing, and was
painful to maintain, so dropping it now.
Some alignment rules from Visual Studion on 32 bit are not taken into
account, this fixes the structs to confirm. We now have static asserts
to catch when such breakage happens in the future.
This is just not practical to do for the code as a whole, and having it as an
exception for one specific data structure is not that helpful. This has only
been in the way for me when refactoring code.
The issue is that Eevee directly evaluates animation on a datablock which
is a part of active dependency graph.
This is a broken logic by design and requires a complete revamp to support
more real life cases when camera is parented to a camera rig, but it is
beyond of what i can do with a simple bugfix.
Issue was that (deep) duplication code of scene ended up leaving
children collections of new master one without any parent.
Note that even though I think that fix is OK for now, we should really
make 'deep' duplication of IDs part of the generic ID management code.
Am less and less happy with current handling of this, done half from
/editors code, half from some semi-specialized helpers from /blenkernel,
with sometimes nearly the same logic replicated several times for
slightly different needs, etc. Unfortunately this would not be a small
refactor, so it will have to wait...
In extrude operator when the point was added, the weight data pointer was wrongly connected to old pointer.
Now, when move the data, the pointer is moved, but when a new point is added, the memory is duplicated to keep separated copies of the pointer.
This is related T62872
Thanks to @sergey for his help fixing this bug.
- Passing original object with apply_modifiers=false will give a
non-modified non-deformed mesh.
The result mesh will point to datablocks from the original "domain". For
example, materials will be original.
- Passing original object with apply_modifiers=true will give a mesh which
has all modifiers applied.
The result mesh will point to datablocks from the original "domain". For
example, materials will be original.
- Passing evaluated object will ignore apply_modifiers argument, and the
result always contains all modifiers applied.
The result mesh will point to an evaluated datablocks. For example,
materials will be an evaluated IDs from the dependency graph.
Fixes T62916: Applying boolean modifier does not set material properly
Differential Revision: https://developer.blender.org/D4604
This was wrong and violating design to force modifiers to query
evaluated objects and IDs. It is up to the caller to make sure
the object is properly evaluatable.
Effectively, reverting changes from de491abf99 (and possibly
other related changes).
Since there is a flush of evaluated values back to the original
for an active dependency graph we don't need this lookup anymore.
Not only it slows interface drawing down, but also is becoming
in a way of the upcoming fix.
Some iterations in bevel were over a hash table, which leads
to possibly different results run-to-run, especially when
loop_slide is enabled. Changed those iters to go over all verts
of BMesh, which leads to consistent order run-to-run.
The old logic was working if the eraser was moved towards the end of the stroke, but got ugly results when the eraser was done towards the start of the stroke.
Dynamic callbacks generating lists of tools should accept None context,
not crash on it.
Similar to what we do with dynamic RNA enums, when NULL/None context is
given, assume we are in 'introspection' mode and return as many things
as possible.
This is currently essentially used by i18n messages extraction tool
(where getting all possible entries is kind of mandatory ;) ).
Also add some initial missing tranlations for complex cases that cannot
be automated, there'll likely be more of that kind...
The issue was the manipulated depth for images would be used when doing distance depth checks.
Now we will use the actual depth of the image object when doing such tests.
Reviewed By: Clément Foucault
Differential Revision: http://developer.blender.org/D4616
New icons from Andrzej Ambroż / jendrzych:
-Adds separate icons for 3D Cursor orientation and pivot
Also, many dozens of tweaks and updates - too many to mention here.
Full list is on Devtalk
This makes it easier to exclude a surface from casting shadows and can be
used to manipulate the shadows even for opaque surfaces.
Versionning ensure that old behavior is transfered to new rendering logic.
This was due to some shading resources not being bound for the shadow
shader. This commit completely remove the shading part when in a shadow
shader. Thus making it a bit more lightweight and removing the needed
sources.
Remove remapping to new IDs from `BKE_object_duplicate()` itself, doing
it here will break later remapping done after all desired objects have
been duplicated, preventing complete remap of ID pointers used in obdata
and other sub-data of duplicated objects.
Checked, the only two usages of this function were already doing
separated later step for this anyway.
Do it only for meshes/curves since those are potentially slow and need user
feedback to see things are not stuck. For object instances and lights assume
it's fast enough.
Printing too much can have a performance impact on slow Windows command
prompt or when logging complex scene renders.
This is to fix the slowdown issue experienced on windows when rendering
from command line.
Fix T59649 Eevee in command-line batch mode is slow with particles/duplis
The wrong transformation was used. Add a new matrix specially for this case.
This also fix the Node texture coordinate that was suffering the same issue.
Thanks to Gabor Fekete for helping finding the issue.
Was caused by uninitialized variable. Also took the oportunity to use comp
swizzling instead of multiple assignment.
Until now, Orcos were computed by the gpu (GLSL) and were not taking into
account the modifier stack (breaking orco for deformed mesh).
Now Orco is now computed on CPU but only if a modifier stack is present.
Tagging that an ORCO layer is present is done via a 4th component, which is
a waste of memory/bandwidth. Best would be to do the same as auto attrib
color space and save a bool uniform somewhere but for now it's too
disruptive.
Now the entire node gets greyed out, instead of using a red header that is
hard to distinguish from other nodes with similar header colors.
Differential Revision: https://developer.blender.org/D4597
BKE_mesh_new_nomain automatically added a CD_ORIGINDEX layer initialized to 0,
which was never filled in correctly. In 2.7 the equivalent function used to
modify the source derivedmesh and add valid original indices to it, but this
is no longer possible in the new design and was quite unpredictable anyway.
Now instead rely on mesh_calc_modifiers and the depsgraph to determine when
CD_ORIGINDEX should be added.
This should not use the original index as a unique instance index, with for
example an array modifier the same original index may be used multiple times.
Previous selected text color was almost invisible against the dark background.
Use same selected/highlight color as in other places in the default theme.
Issue reported on Devtalk by user Symstract
This operator was used only by annotations and it was part of the old legacy code.
Now, the operator is limited to annotations and cannot be used with grease pencil objects.
Also changed the tooltip to clarify it's only for annotations.
Switching shader modes to material or render during a sculpting session
showed the model when the sculptsession was started. The user needed
to click that forced an update of the sculptsession.
In `sculpt_flush_update` the eevee drawing mesh is not updated when
workbench engine is used.
When texure shading is enabled, the transparency accum shaders were not set/updated.
This resulted into a crash when transparent objects were rendered when shading mode
was set to texture.
This patch will add the V3D_SHADING_TEXTURE_COLOR to the modes for OIT so the
shaders are set.
Reviewed By: fclem
Maniphest Tasks: T62874
Differential Revision: https://developer.blender.org/D4592
Objects that internally uses DispList do not cast shadow in the workbench.
Their outline is also not visible in object mode. The reason for this is
that edge detection was not implemented for Display Lists. This patch will
implement the edge detection.
Reviewed By: fclem
Maniphest Tasks: T62479
Differential Revision: https://developer.blender.org/D4605
The root point of a bone is now always drawn unless it is connected to an other bone.
Reviewed By: Clément Foucault
Differential Revision: http://developer.blender.org/D4194
Now island margin acts as a percentage instead of being the actual distance.
So the margin distance will always be the same regardless of the actual size of the geometry.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4514
By design the annotation parameters must not be animatable.
Annotations are designed to take notes, not to create animations, so any animatable parameters have been flagged as non-animatable.
Note: As some properties are shared with grease pencil, I had to duplicate one property (adviced by @mont29) to keep grease pencil animatable but annotations don't.
Was a mistake in the recent change.
The idea here is: manual should invalidate point cache, but
rigid body simulation should only happen after point cache
was ensured to be in the correct state.
Behavior of source/destination options was rather flaky in reversed
mode...
Note that even though this mode is a bit annoying and cumbersome, the
only alternative (defining a specific transfer weight operator) would be
much more verbose, so think that for now we'd rather keep what we have
here.
This was originally added in 6feddb8b61, but the warning is quite clumsy
and we also don't limit editing object transforms that way. The add constraint
button at the top now also indicates if it's for objects or bones, which was
not the case before and probably was part of the confusion.
If it does still cause issues we can always change back or hide the tab.
for objects other than Empties
In 2.8 support for dupli-group instancing for non-empty objects was
removed (rB2eca054e14b1), so better prevent operators from setting
this...
Reviewers: brecht
Maniphest Tasks: T63004
Differential Revision: https://developer.blender.org/D4599
Use this for the save confirmation dialog so it has a default action
when pressing enter which draws with a highlight so it's clear what the
default action is (the dialog was just closing before).
Resolves T57686
This reverts commit b24a255ca6.
The warning wasn't being printed, so many add-ons weren't yet updated.
Re-enable now warnings are printed to allow addon developers to update.
Many OpenGL functions take NULL pointers, passing those was quite complicated
with some addons even using ctypes to manipulate internal bgl.Buffer pointers.
Eevee and Cycles don't use these clipping distances the same as Blender
Internal did, or at all in most cases. Just remove this since it makes no
sense anymore.
Afaik all the panels in the scene settings are render
engine independent. So it does not make sense to have
this extra complexity in every panel class.
Reviewers: brecht, billreynish
Differential Revision: https://developer.blender.org/D4594
Issue would happen in case affected object is linked to collections that
are shared between different scenes' ViewLayers. When switching back to
another scene after making single user, you'd get immediate crash.
Code was still 2.7x one here, Object's IDtemplate in properties editor
is now based on current view layer's collection of objects, not scene's
one anymore.
We do not support special color fonts (like colored emoji fonts), so
don't crash when trying to create a preview for unsupported fonts.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4590
This removes `VIEW3D_OT_select_or_deselect_all`, adding a
deselect_all option to the `VIEW3D_OT_select` operator.
- Add utility functions to simplify de-selecting all.
- Return true from selection functions when they change the selection
to avoid redundant updates.
- Use arrays of bases when passing objects between selection utility
functions since some users require bases.
- Fix logical error in box selection that updated all objects after
the first hit.
Some draw code did not respect Show Overlays option. These were:
* All mode based drawing engines (edit mode)
* Wireframe drawing
This change make them respect the Show Overlays Option.
Reviewed By: fclem, billreynish
Maniphest Tasks: T62774
Differential Revision: https://developer.blender.org/D4572
Not sure why that was asserted on instead of handling the flag properly,
if base is not selectable, then just do not select it...
Have the feeling this code handling sync of flags between bases and
objects could use some cleanup, but that will be for another day.
Using GP_BRUSH_MATERIAL_PINNED to switch between active material and brush material, instead of updating all brushes on active material changes. This will allow brushes to have no material and therefore to not inflate the user count.
This fix T62465.
Patch contributed by @matc
Reviewers: @brecht@antoniov @billreynish @mendio
Selected status was not properly moved from new object to new created base.
This prevented next steps (making obdata, materials etc. local) to ever
happen.
Most of the source tarballs are retrieved via http, but a few remain
that are still downloaded via ftp. This causes some pain with corporate
firewalls, so moving the last two URIs to http helps ease the build process.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4192
For custom path selected during 'install_deps.sh' using '--source'/'--install', paths for blosc, embree and opencollada are not printed/inclued into BUILD_NOTES.txt file.
As '/opt/lib/<package>' paths are hardcoded into CMakes's Find* modules, this error is not noticeable, but for custom paths it is.
This patch includes those fixes/prints for those packages.
Reviewers: mont29
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D4574
A version for menu's is going to be added next and we already have
UILayout.prop_menu_enum.
This name indicates the popover is added behavior instead of a
different kind of widget.
This adds entries to copy/paste (selected) IDs in the Outliner, as well
as usual ctrl-C/ctrl-V shortcuts.
Note that the clipboard is shared with other IDs copying (currently, the
one for objects in 3DView).
Reviewers: brecht, dfelinto, billreynish, pablovazquez
Differential Revision: https://developer.blender.org/D4568
This commit does not add anything new from user perspective, but make it
possible to paste any kind of IDs, not only objects/collections.
Will be used by new copy/paste in the outliner in next commit.
Support for RNA enum buttons that activate popovers when clicked.
This means we get useful tooltips, shortcuts and Ctrl-Wheel cycling
over enum items.
It also avoids inconvenient & slow access of enum values currently done
via RNA type lookups on the type to get the name & icon to use for a
regular popover button.
Resolves T57738
When two stops had the exact same color, 'Far' interpolation would
behave like the three others and generate constant colors, instead of
generating expected 'full turn' in Hue space.
Note that this will break the 'constant' color between two stops with
same value in the 'Far' interpolation mode, but think that change is
OK/expected.
This fix also exposed that 'early out' case when we are after last stop
was not working properly, still triggering actual interpolation
computation in all cases, thus the refactor in `BKE_colorband_evaluate()`
itself. We also now avoid computing factors in constant case, and get
out early in all interpolation cases but the 'spline' ones (where color
ramping can start before first stop, and end after last one).
Work initiated by patch from Charlie Jolly (@charlie) in T49979, thanks.
Reviewers: brecht, sergey
Subscribers: charlie
Differential Revision: https://developer.blender.org/D4556
This forces bone transparency / wireframe display when Xray is enabled.
This makes sense as the bone transparency is a kind of Xray and should
be linked to the Xray display.
Instead of doing some fancy stencil buffer tricks, just clear the depth
buffer before the "in front" meshes.
Fix T58841 Hidden Wire when in Edit Mode disables In Front Option
While I broke the non-stereo camera frame in rBf1fef41863b I came to
realize after further tests that the stereo camera frame was NEVER
correct.
It is now, regardless of the camera shift, the camera pivot, if you are
looking through the left, right or stereo eyes.
In the end all that was missing was to multiply the shift we needed to
offset the frame, by the frame width.
I also took the opportunity to perform some cleanup/refactor regarding
names. It was too confusing even for me.
Note: Non-uniformally scaled cameras still draw slightly wrong.
SpaceSeq main stereo drawing is for ar->regiontype RGN_TYPE_PREVIEW, but
the drawing code was assuming RGN_TYPE_WINDOW was the only one to be
considered for all the regions.
SpaceSeq still needs stereo drawing in RGN_TYPE_WINDOW but only when
backdrop is enabled.
Bug introduced on rBe01cadd657c76267.
Patch by Shinsuke Irie.
* * *
Note from reviewer/committer (dfelinto):
OpenGL render from the VSE preview window is working, while the VSE
preview window itself does not show stereo.
That said the patch is ok, and I was able to test that the VSE preview
OpenGL still works. I will tackle VSE preview itself separately myself.
Request from the Spring team. When compositing the CTRL-Shift Click on a
node cycles the viewnode with an output socket. When you have many
render layers you spend time cycling to the desired socket. This patch
allows the user to CTRL-Shift Click on a socket to connect directly to
that socket.
Reviewed By: brecht
Maniphest Tasks: T62785
Differential Revision: https://developer.blender.org/D4564
It updates only the main window you edit and the non-main window
children of this main one.
Reviewers: sergey, brecht
Differential Revision: https://developer.blender.org/D4528
This is a full revert of b6c61945ae.
But ammended to include the object.hide_collection() operator to the
pose keymap as well.
While trying to expose the shortcuts to pose mode, I ended up exposing
them for all the modes. It is now only available for object and mode
modes.
- Rename ui_but_find_active_in_region -> ui_region_find_active_but
- Remove ui_but_is_active (use ui_region_find_active_but instead)
- Re-order ui_but_* functions take the button as their first arg.
Stencil mask drawing was not implemented yet. This commit will implement this for texture painting.
It brings the state back to how it was for B279.
Reviewed By: fclem
Maniphest Tasks: T58727
Differential Revision: https://developer.blender.org/D4570
Basically the framebuffer size was different from the glViewport size.
This made the depth read in glReadPixel not corresponding to the center of the pixel.
Another thing that reduced precision compared to blender 2.79 is the `GPU_matrix_unproject` that now computes using `float`s instead of `double`s.
But this may be for another commit.
There was missing flush from transform update to the point cache
reset. Caused by the fact that when update happens in the "middle"
of component all the component operations will be tagged for update
(since the intermediate state is not stored), but that will not
flush updates to other operations since that would cause too much
of updates.
This now we tag point cache for reset after evaluation operation
but before final transform and before rigid body world.
Due to recent changes the AA was not rendered in the viewport when not
interacting. When interacting it was. This was due to incorrect
detemination to perform TAA.
Even though that one is not really useful just to build Blender, we can
as well explicitely include it here, since all 'default' Blender builds
will include full clang/llvm stack anyway (for Cycles and deps)...
- Replace Window context menu with File menu.
This was recently removed however artists in the studio use
link/append often.
- Add preferences to this menu so `F4, PKey` can still be used as a
quick way to access preferences.
This was inherited from 2.7x behavior with groups, but in 2.8
collections are also used as 2.7x layers, which turns to be a problem
when pasting back clipboard content, since it would instantiate
collections instead of objects, quickly leading to tens of new
collections in the viewlayer...
Instead, we only copy selected objects from the 3DView now. On paste
time, those will be added to the active collection.
Last part of fixes related to T61670.
Mouse wheel on menus changes the selected item item which doesn't work
well for popovers since buttons aren't always stacked vertically.
Resolves T62777
Create a new pass for them and draw them separately.
Beware of the double multisample resolve when multi-editing a mix of curves
with and without the "in front" option enabled.
The issue is a feedback loop with point cache reset operation.
Solved by introducing extra node which ensures object's transformation
is ready for simulator input. Allows to route relations without causing
a fake dependency cycle now.
This patch adds a new "Use Alpha" option on image empties to avoid ordering
issue of reference images.
If ordering is not important, "Use Alpha" can be enabled to provide
transparency and alpha blending support.
The issue was caused by mesh needed for dynamic paint brush being
stored in the modifier. That make it to be freed and set to NULL
when running copy-on-write.
Solved by moving the mesh to be stored in modifier's runtime data.
Missing dependency graph tag. Is only happening for scripts, since
the interface is handling this from do_constraint_panels().
Fix is suggested by Edmund Kapusniak (@edmundmk), thanks!
While it looks more longer, but also contains more comments
about what's going on.
Surely, this function almost never breaks and investing time
into maintaining its tests is not that important, but we should
have a good, clean, understandable tests so they act as a nice
example of how they are to be written. Especially important to
show correct language usage, without old school macros magic.
Doing this at a lunch breaks, so will be series of some updates
in the area.
When the area of the crop node is zero the hotspot margin becomes infinite.
This makes the gizmo by default think the translate mode hotspot is everywhere.
This fix will return a state if the INFINITY is detected so we can
return the hotspot drawing.
The cage2d gizmo is still not usable for sizes of 0 by 0 as now it won't
draw anything. the issues of the gizmo should be re-designed so we can
mitigate these limitations. But that is out of scope for this fix.
Reviewed By: campbellbarton
Maniphest Tasks: T62248
Differential Revision: https://developer.blender.org/D4516
Instead of making the entire body of the function conditional, it now
returns early, unindenting the entire function and preventing the reader
from searching for a non-existent `else` clause.
No semantic changes.
The `BLI_path_frame_strip` function was completely broken, unless the
number of digits in the sequence number was the same as the length of
the extension. In other words, it would work fine for `file.0001.abc` (4
digit `0001` and 4 char `.abc`), but other combinations would truncate
to the shortest (`file.001.abc` would become `file.###.ab` and
`file.00001.a` would become `file.##.a`). The dependency between the
sequence number and the file extension is now removed.
The behaviour has changed a little bit in the case where there are no
numbers in the filename. Previously, `path="filename.abc"` would result
in `path="filename.abc"` and `ext=""`, but now it results in
`path="filename"` and `ext=".abc"`. This way `ext` always contains the
extension, and the behaviour is consistent regardless of whether there
were any numbers found.
Furthermore, I've removed the `bool set_frame_char` parameter, because
it was unclear, probably also buggy, and most importantly, never used.
I've also added a unit test for the `BLI_path_frame_strip` function.
In 2.79 there was an "Item" panel with only to give access to rename
from the 3D view.
Add this functionality for context sensitive renaming.
Currently this works for objects/bones/sequence strips & nodes.
This uses F2 for a window level shortcut, replacing the file-io menu.
See: T61480
A hard coded threshold was used to ignore cursor motion,
make this a preference since tablet users may want to increase it
since a pen hovering isn't as easy to keep still as a mouse.
Resolves T56278
The first part of this patch is just a way to improve performance a bit
by not calling BKE_scene_object_base_flag_sync_from_base() twice.
The second is the proper fix for the reported issue.
That said, the report can be re-opened since there is still discussion
about whether or not to bring the collections with the objects.
Reviewers: mont29
Subscribers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4558
When copy/pasting an object in same .blend file, orig object would be
instantiated into the active collection when pasting, along with the
content of the paste. Was missing a clear of LIB_TAG_DOIT on objects...
Found while investigating T61670.
Some view operations auto-keyed when locking the camera to view
but some NDOF operators didn't yet support it.
Auto key during animation playback for recording camera motion.
Unlike most transform operators that use an orientation and values,
Mirror uses the constraint axes to define which axes are mirrored.
Now constraint axes are shown in this case.
Other changes:
- Use 'EXEC_DEFAULT' for mirror menu items.
- Show mirror on local axis when not in edit-mode since this is useful
in object mode too.
- Remove mirror vertex groups since this is in the vertex group menu and
this menu isn't currently used for general symmetry operators which
are currently in other menus.
If a subfolder was specified which didn't exist, logic would fallback
to get_path_user (instead of get_path_environment).
Now always use the from the environment variable if it's set and exists.
Finding visible connected elements is often a direct lookup
when they're not connected to hidden geometry.
Add inline wrappers that avoid a function call,
gives minor speedup creating GPU edit-mesh data.
All edit mode related operators are now "Hide Selected / Reveal Hidden".
Before we had different variatons of them:
Hide Selection:
* MESH_OT_hide
* GPENCIL_OT_selection_opacity_toggle
Hide Selected Bones:
* ARMATURE_OT_hide
Hide:
* MBALL_OT_hide_metaelems
Hide Selected:
* CURVE_OT_hide
* PARTICLE_OT_hide
* POSE_OT_hide
* UV_OT_hide
Reveal Bones:
* ARMATURE_OT_reveal
Reveal:
* MBALL_OT_reveal_metaelems
Operators not renamed:
* Hide Curves (GRAPH_OT_hide)
* Hide (OUTLINER_OT_hide)
Note we can do a step further and mass rename them to match their UI
names. Where Reveal Hidden is Show Hidden. But for now at least they are
all unified.
This is only available through the API, mainly intended for render farms to
combine rendered multilayer EXR Files with different samples. The images are
currently expected to have the exact same render layers and passes, just with
different samples.
Variance passes are still simply a weighted average, ideally these should be
merged more intelligently.
Differential Revision: https://developer.blender.org/D4554
- Consolidate each brush section (Color, Palette, Gradient) and make them distinct
- Remove the lock icons and move these items into an Options sub-panel, together with other toggles
- They now have more descriptive names
- Use an enum for view vs scene brush unit
- Use Property Split layout and sub-panels in line with the rest of 2.8
- Rename Curve panel to Falloff
Reviewed by: campbellbarton, pablovazquez
Maniphest Tasks: D4529
Differential Revision: https://developer.blender.org/D4529
@sergey wrote: "The reason it was failing is because md points to an original armature, which doesn't have runtime data needed for deformation. That data is to come from evaluated armature."
Thanks to @sergey for his help.
When Specular Transparent materials are used the world clipping
did not work on the transparent materials. The reason was that the
accum shaders did not vary on the existance of the clipping planes
This patch will variate the accum shaders when clipping planes are
active.
Reviewed By: fclem
Maniphest Tasks: T61023
Some users of the 3D versions were storing 2D data in it.
Using a 3D tree for 2D data adds a spatially redundant branch
every 3rd level, as well as some extra memory use, best avoid this.
Workbench render engine did not work when an odd number of AA samples
were used. A user could enter these values by disabling AA or set the
number of AA samples in the Render/Film panel to an odd number.
This commit will not perform TAA passes when AA is disabled.
For supporting the setting of 5 or 11 samples the bitmask was replaced
by an if statement as this was making the odd number not render
correctly.
As extra introduce the jitter samples of 5 and 11 so the images will
be more clean. a jitter sample of 11 used to read outside the allocated
space of the jitter samples.
Fix T60820
Reviewed By: fclem
Maniphest Tasks: T60820
Differential Revision: https://developer.blender.org/D4546
Group nodes should not allow to add IO sockets to themselves directly,
in that case we actually want to add IO sockets to their underlying
node tree. Fairly straioght forward to support actually.
This moves logic into kdtree_impl.h which is included in a source
file that defines the number of dimensions - so we can easily support
different numbers of dimensions as needed
(currently 3D and 4D are supported).
Macro use isn't so nice but avoids a lot of duplicate code.
When using LookDev in Quad view the cache was only updated for the first
Q-view. The evaluation data of the last cache was not stored with the
lightprobe textures. As the evaluation data was reused between the
Q-views only the lightprobe textures of the first view was updated.
With this patch the evaluated data is stored amongside the actual
lightprobe textures.
Fix T59046
Reviewed By: fclem
Maniphest Tasks: T59046
When the grease pencil was active no other objects were taken into
account. There is no need for a special case here since objects
bounding boxes are used.
Existing code was not really handling well cases were only edges or
faces were selected (with match select modes).
In those cases, we can assume user want to only affect loop normals of
selected faces/edges, not all lnors of all (indirectly) selected verts.
Also refactored the code a bit to move whole 'loop to edit' selection
process into its own single function.
Regression in 358e07f447 for ms-windows since off_t is an int32_t
even on 64bit systems causing files over 2gb not to load.
Poison off_t so this doesn't happen again.
- Quaternions weren't normalized before interpolating
causing incorrect results & assert.
- Make the newly calculated quaternion compatible with the previous
to avoid axis-flipping & setting values with large changes compared
to existing key-frames.
Was due to incorrect output size. Use engine->resolution_x/y instead of
computing the size on the fly from the scene specific settings (which are
not overwritten by the rendering scene).
-Use property split
-Use sub-panels for Options panel
-Slightly re-organize Options panel
-Remove redundant Deflect Emitter from Options panel (it was already in the tool settings for the Comb tool)
Scale waveforms in sound strips by volume.
If any drawn line exceeds value 1 or -1 it is drawn by red color so user can see point, where clipping occurs.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4515
Add movieclip fallback render option, for case, when proxies are not enabled or built
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4219
Instead of only `ymin` and `ymax`, `ANIM_channel_draw_widgets`
has a rectangle as input now.
This allows the caller to set a custom width for the channel.
Some space types need the extra space for the scrollbar
(drivers, graph), but the other don't have a scrollbar.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4543
The general idea of this change is to have a runtime data pointer
in the ModifierData, so it can be preserved through copy-on-write
updates by the dependency graph.
This is where subdivision surface modifier can store its topology
cache, so it is not getting trashed on every copy-on-write which
is happening when moving a vertex.
Similar mechanism should be used by multiresolution, dynamic paint
and some other modifiers which cache evaluated data.
This fixes T61746.
Thing to keep in mind, that there are more reports about slow
subdivision surface in the tracker, but that boils down to the
fact that those have a lot of extraordinary vertices, and hence
a lot slower to evaluated topology.
Other thing is, this speeds up oeprations which doesn't change
topology (i.e. moving vertices).
Reviewers: brecht
Reviewed By: brecht
Maniphest Tasks: T61746
Differential Revision: https://developer.blender.org/D4541
libquery code has some specific handling for IDs tagged as 'no_main',
among which to never consider them as refcounted/refcounting other IDs.
This is fine, but it also means we have to be careful when moving an ID
from main to out-of-main status, to do all id remapping we need
//before// we tag it as no_main.
That was a bit tedious to track down, we'll have to be careful that all
the corner cases we have to take care of, do not end up in a giant soup
of expections to exceptions, where nobody can find its way anymore...
Always use native function since this was already the case due to
__CL_USE_NATIVE__ not being defined in time, and seems to have caused no
known issues.
A deform-only modifier that needs access to normals need a copy of
evaluated mesh with those normals updated, when it is not the first one
in the stack.
That issue had been partially fixed in Object mode a long time ago
(see T23673), but it was still broken for deform-only stacks cases.
And it was also completely missing from the Edit mode code
(`editbmesh_calc_modifiers()` function).
We are in a totally out-of-main context here, so no refcounting of any ID...
Note that this whole 'render preview' area could use some refactor with
modern ID management API, but that would go way beyond a mere bugfix,
and it is not the time to do such things.
This replaces the global `PEData->dist` with a thread local variable
that is passed into the callback functions separately.
There are two new function types which have `Hit` in the name.
Those take the distance to the mouse as an additional parameter.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4538
We don't want to use flow control like `break` statement into the basic
`FOREACH_MAIN_ID` macro, as this is a nested loop.
When refined behavior is needed (like breaking whole iteration, or just
skipping to next ID type), FOREACH_MAIN_LISTBASE and
FOREACH_MAIN_LISTBASE_ID macros should be used instead.
Based on D4382 by @campbellbarton
(Other potential solution, using flow control macros: D4384).
This couldn't be accessed because the poll function was checking the
active button.
Add a "color_ramp" context pointer which is set by the menu.
Alternative fix to D3984
Logic to for boids to avoid head-on collisions was in BLI_kdtree.
Move this into a callback which is now defined in boids.c
so the kdtree code can be kept generic.
This commit groups several options that were tested in grease pencil branch:
- Changes to fill algorithms and improves, specially in small areas and stroke corners.
New options has been added in order to define how the fill is working and internally there are optimizations in detect the small areas in the extremes.
Kudos to @charlie for coding this fill improvements.
- New 3D cursor view plane option.
Now it's possible to lock the drawing plane to the 3D cursor and use the 3D cursor orientation. This allows more flexibility when you are drawing and reduce the need to create geometry to draw over surfaces.
- Canvas Grid now can be locked to 3D cursor.
- New option to reproject stroke using 3D cursor.
- Small tweaks and fixes.
Changes reviewed by @pepeland and @mendio
T62570 and T61796 show that we need two slightly different behaviors in
post-linking collection process:
* For linking, we never want to instantiate indirectly-linked
collections or objects.
* For appending however, since all collections and objects will become
local and hence need instantiation, we want to 'link to scene' all
collections first, better than instantiating the objects in the master
collection opf current scene.
This makes degenerate triangles be ignored, as they cause artifacts and
don't contribute to the result anyway. There is an extra cost for
all triangles, to check if they are degenerate. Though not ideal, this
seems to be the safest solution at the moment (see code comments).
This commit also changes the clamp value for the bleed len_fact from 5
to 10, to give good corner results with some sharper angles, considering
they will be rounded later.
A lot of the texture paint initialization code could be optimized
by caching the results throughout the session, instead of rerunning it
for each stroke. Stroke initialization is way too expensive at the
moment, and does many redundant computations.
There are two issues at play here.
First, BLF_width computed a width that was not wide
enough to actually hold the text.
Second, blf_glyph_calc_rect_test computed an incorrect
rect->xmax when the glyph was moved to the left a bit.
We ignore the overlap on the left, but the right side
should still be adjusted accordingly.
The cause of this issue goes deeper inside of the custom nature of
the event loop. In short, when not using [NSApp run] closing "About"
window does not make previous key window a key again.
The solution is quite similar to other projects, but we only force
key window from handler when closing one appears to be "About" window.
In all other cases we leave it up to Blender's window manager to make
decision and do not interfere with it.
Test plan:
- Open Blender application
- Go to Blender -> About Blender menu item
- Close About window
- Note that Blender's window does not become active again.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4534
Actually, when you press G/R/S in Draw mode, the transformations can be done, but this is annoying in draw mode and must be captured an cancel.
This patch capture the transformation and cancel it.
Reviewed by @brecht
It's effectively always enabled, only not on some unsupported OpenCL devices.
For testing those it's not useful to disable these features. This is replaced
by the more fine grained feature toggles that we have now.
This was already supported for Cycles shader nodes, but now also works for
Eevee and compositing nodes. Instead of a generic NodeCustomGroup, now
there is ShaderNodeCustomGroup and CompositorNodeCustomGroup that can be
subclassed and registered.
Differential Revision: https://developer.blender.org/D4370
User counting now happens before init() and after free() methods, so that
the ID users are in a valid state when Python might modify them. ID user
counting was moved into node.c and simplified.
Patch by Miguel with further refactoring by Brecht. Ref D4370.
X3D has been (basically) ported in rBAe8da70ab73d2dd5ff46, and 3DS has
been downgraded to 'community' support, so we can get rid of that ugly
'slience warning' hack.
The billboard particles were only used by Blender Internal. So until it is
supported by Cycles of Eevee there is no reason to keep it in the code and
UI.
Fix T61695 Billboard particles not displaying in Eevee viewport, render
The shaders are: `GPU_SHADER_3D_FLAT_SELECT_ID` and `GPU_SHADER_3D_UNIFORM_SELECT_ID`.
This commit allows the drawing of the mesh select ids to be done on a 32UI format texture.
This simplifies the shader that previously acted on the backbuffer and had to do an uint to rgba conversion.
Differential Revision: https://developer.blender.org/D4350
We already have dedicated outliner visibility operators that not only
take precedent over those ones, but also unhide collections as well.
No change for the end user. That said, since I added the outliner
hide/unhide operators I assumed that newly visible objects are to be
selected. That can change in a separate commit, but it should be done
by changing OUTLINER_OT_unhide_all.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4527
The select option in the operator was never working because we were
trying to select an object considered unselectable (since it was hidden).
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4527
This introduces a new iterator, FOREACH_COLLECTION, that unlike the
FOREACH_SCENE_COLLECTION it iterates over all the Blender file
collections, including the scene master collection, as well the database
ones (bmain).
Reviewers: brecht
Similar to T60623 / rB2894e75121d7.
Issue (when parenting with 'Follow Curve') is that the curves
CU_FOLLOW flag is set on the original, but not the
evaluated curve (yet), leading to misbehaviour in 'ob_parcurve()'.
Setting this on both now.
Reviewers: sergey
Maniphest Tasks: T62601
Differential Revision: https://developer.blender.org/D4524
This version fixes various bugs, and there is no need anymore to use both
9.1 and 10.0 for different cards.
There is a bug related to WITH_CYCLES_CUBIN_COMPILER and bump mapping in the
regression tests, so that remains disabled same as it was for CUDA 10.0.
Fix T59286: CUDA bake failing on some cards.
Fix T56858: CUDA 9.2 and 10 issues.
It only returned those for the active device type. For backwards compatibility
return them all again, but still avoid enumerating them from our own code on
startup or opening preferences.
The main goals of this change is faster starting when using foreground
rendering.
This patch will build kernels in parallel to the update process of
the scene. When these optimized kernels are not available (yet) an AO
kernel will be used.
These AO kernels are fast to compile (3-7 seconds) and can be
reused by all scenes. When the final kernels become available we
will switch to these kernels.
In background mode the AO kernels will not be used.
Some kernels are being used during Scene update (displace, background
light). When these kernels are being used the process can halt until
these become available.
Reviewed By: brecht, #cycles
Maniphest Tasks: T61752
Differential Revision: https://developer.blender.org/D4428
Issue introduced on d227c58e3e.
This was affecting WorkSpaces as well as background scene and probably
other places. We use the operator poll to decide whether to show it.
Reviewers: brecht, mont29
Differential Revision: https://developer.blender.org/D4488
The functions that determine the program name + filename of kernels
were missing some base kernels like denoising and base. For completeness
I added those kernels so the function returns the correct results.
This mimics RNA style class registration,
keeping the same internal data types.
Currently there is a template which shows an example of adding a tool
group with a keymap.
Icon generation still needs to be exposed for general use.
Prepare for exposing tool registration to the Python API.
- Generated tools can use their own prefix so naming collisions
won't happen between hard coded & generated tools.
- Add-on authors can use the add-on name as a prefix.
Currently the names match, renaming will happen next.
Enable depth picking by default.
This adds new 'gpu_flag' since it's not so relevant to add GPU drawing
options into uiflag & uiflag2.
This resets the recently added smooth edge flag.
Invert the texture of the stroke if the random UV is activated. Before, this parameter only affected to dot mode, but not to line strokes.
The texture is inverted vertically.
This fix saves the camera matrices in order to not call
BKE_animsys_evaluate_animdata during each draw loop. This function tags
the view as dirty even if the camera does not move.
This effectivly, avoids the constant reset of TAA.
The main problem was that the character `J` has "negative kerning".
It was cut off because it started outside of the clipping rectangle.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4513
Need to ensure all dopesheet are properly tagged for update.
The tricky part here is that due to the animation we need to keep
movie clips covered by copy-on-write, but the interface uses an
original clip to draw dopesheet.
Use same idea of an active dependency graph as is done in other
similar cases.
This patch puts threads that render the same pixel closer together,
as opposed to threads that render the same sample. Thus threads
within a warp are more coherent in memory access and control flow,
leading to performance improvements.
Example benchmarks on a Quadro RTX4000 (WDDM) on Windows 10:
Koro: 4:23 -> 3:46
BMW: 1:18 -> 1:25
Barbershop Interior: 17:52 -> 14:55
Classroom: 4:37 -> 3:45
Performance differences on OpenCL/AMD were hit and miss, some scenes
became faster, others lost significantly. Therefore, this is kept as
CUDA only change for now.
Stupid mistake in previous 'fixing' commit, I put the call computing
extra required source cddata masks into the callback used to get
destination (current object) cddata masks...
The NODE_GROUP_LEVEL of the Geometry node should be bumped to 1
when Backface is connected. Backface uses `NODE_LIGHT_PATH` that
is part of NODE_GROUP_LEVEL1, the rest of the geometry ndoe is
NODE_GROUP_LEVEL_0.
For full keymaps, ensure keymap items are lists to allow predictable
manipulation at runtime.
When calling `keymap_init_from_data` directly, allow any sequence type
to allow tools to define their keymaps as tuples.
In this case, the generic vertex attribute is {0,0,0,1}. So we look for
this case.
This fixes black text objects with a normal map applied. Also this could
help porting sculpt mode drawing to Eevee without supporting normal
mapping.
Note that will just fix black meshes due to T61870 but objects will not
show their normal map. So it's not a fix for this issue.
Left and right aligned regions get hidden when there is not enough space,
and the main center region was empty. Now remove the left region and put
its contents in the center region, so that it always remains.
Draw manager was checking for particles being enabled in viewport, but
actual evaluation was happening for render. Now the check in draw manager
properly follows current evaluation mode.
Additionally, de-duplicated some check, by moving everything needed to
a single place.
Annoying part is the const-cast, that could be cleaned up later by
modifying some underlying functions.
The core was hooking up relation to a wrong operation, in an
attempt to cause proper full update of the pose (since there
is no intermediate state stored anywhere). This is now done
as a part of flush_handle_component_node(), so can properly
route more granular relation.
This fixes the crash which was caused by an access of partially
evaluated data by solving the fake cycle. However, if there
is a real cycle involved the code will still not survive, but
this is more generic issue (which also includes modifiers
for meshes and curves).
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4509
Before the cursor was only visible after the cursor moves over side areas, but not in viewport area.
Now, as soon the drawing is completed, the cursor is reset in order to get again the icon in the viewport.
The option can be changed in the Preferences
in `Interface -> Editors -> Factor Display Type`.
Reviewers: campbellbarton, brecht, billreynish
Differential Revision: https://developer.blender.org/D4462
Use dedicated flag to tag animation just for copy-on-write synchronization,
which makes it so copies of the original data blocks gets in sync with the
original ID. This will not flush the animation update to all objects which
depend on that animation.
If such flush is required, use ID_RECALC_ANIMATION.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4508
When having loops in a kinematic chain and also a IK constrain,
the solve_cycle function removes graph relations necessary to ensure
the cleanup operation was running at the very end.
Due to his Blender was crashing when some operations (the bone
constraints) accessed a pointer that was already freed.
Solved by adding more relations between crucial nodes.
Broke uv-sphere creation, further it might be a problem
for script authors expecting matching edge order for duplicated content.
Now only apply this when duplicating via the operator.
Confusing to call a menu a panel when subclasses need to define
panel specific variables.
Avoid having bl_ui depend on bl_operator module too.
Since this isn't an operator, add utils modules for shared types.
staticmethod shouldn't be used unless it's necessary,
in that case a comment should be included especially when
subclassing non-staticmethods from RNA.
This is a hacky fix. We just convert the int as a float and use it as such.
This works ok for small int but will not be correct for numbers greater
than 4194303.
Correct support would require deeper change for UBO creation and socket
conversion.
Add Object.local_view_get/set
Alternate fix which supports removing from local view & checking
if an object is in local view.
Also avoids redundant refresh.
Matches 2.7x Object.layer_local_view capabilities more closely,
without exposing the flag directly.
This reverts commit 9054b39b27.
Using the region data to find the screen & space-data doesn't make sense
since local-view is space-data there is no reason to pass in a region
argument, then lookup the space-data.
There was no way to remove an object or check it's local view state
as there was in 2.7x.
This can be made closer to 2.7x API and simplified.
Allow dynamic defines to have defaults which don't exist in the
static array, this is an exception that's only allowed when the item
array is set to DummyRNA_NULL_items.
The poll function accepts hidden objects, so they are not stuck in pose
mode, but the operator itself expects a valid base, which we don't have.
When called from OBJECT_OT_mode_set() it will fallback to object mode.
It is the same that happens when in Edit Mesh mode with a hidden active
object and trying to change to Vertex Painting mode.
This was due to environement not being rendered with alpha blending. So
color was still written and contributed to the final render color. Now
we multiply by background alpha so that it removes any background pixels
intensity.
For this reason this made the (incorrect) final premult unecessary.
If you try to erase empty annotations, the operator is canceled.
The message is necessary to avoid the keymap is captured by the next keymap in the stack.
These new functions control the per view layer object hiding state, similar to
the selection state. All these object state checking functions now also optionally
take a view layer to use instead of the active view layer.
Fixes T62062.
To make triaging a little easier these batchfiles generate the
debug log and sysinfo files and instruct the user how to attach
them to their bug report
Differential Revision: https://developer.blender.org/D4505
Reviewers: brecht, zeddb
After recent changes BASE_VISIBLE was not always disabled properly when the
object is hidden. This refactors the code a bit to hopefully be more clear.
The `--render-format` CLI option takes `EXR` and `MULTILAYER`, whereas
the DNA image format render setting uses `OPEN_EXR` and
`OPEN_EXR_MULTILAYER`. This commit adds the DNA values to the CLI
argument, so that it is possible to take the selected value from DNA and
pass it as-is to the CLI. This is used in Flamenco, for example.
The `OPEN_EXR` and `OPEN_EXR_MULTILAYER` are now aliases, so both the
existing and the new options keep working.
Reviewers: campbellbarton, brecht
Reviewed By: brecht
Subscribers: fsiddi
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D4502
This was the cause of some issue with normal mapping. This way is cleaner
since it does not modify the state of the drawcalls and other ad-hoc
solutions to fix the problems down the road. Unfortunately, it does require
to fix every sampling coordinate for this texture.
Fix T62215: flipped normals in reflection plane
In other words, Duplicate Collection could link local ID into a linked
one... Nasty. ;)
Add checks that found parent is not a linked data-block (and try to
find a fall-back one if this is the case).
As exposed in T62406, we can have some rare cases of crashes due to
memory re-allocation happening outside of expected scenarii.
Ideally this should be re-designed, but at least keep track of those
known exceptions to general rules...
This patch will reduce the number of times that we need to
recompile kernels. It does this by (en/dis)abling features
by default. So when the user needs them that the kernels are
already available.
Other features are enabled by default for background and foreground
rendering. When in background rendering the user wants the best
render performance. When in foreground rendering the user wants
the least amount of recompilations.
Enabling volumetrics or subdivision evaluation will still trigger
a recompilation during foreground rendering.
Reviewed By: #cycles, brecht
Differential Revision: https://developer.blender.org/D4485
Use the term "AFTER" instead of "OFS" since it wasn't obvious these
macros operate on everything after the struct member passed.
Avoid casting to non-const types when only reading.
Now when the text argument is omitted, use the default name
matching how regular properties work.
Avoids passing in the same name which RNA has,
matches UILayout.prop behavior.
Also use keyword only for optional arguments.
In keeping with convention to match code & UI naming.
- No user visible changes.
- Include 'menu' in the name since context is an overloaded term.
- While a few of these are panels, from a user perspective they are
still context menus.
Am not sure whether we want to support that at some point, but at least
if we don’t, make it safe and get out of copy code early, previously
trying to duplicate workspace through BKE_id_copy() would have given a
very bad a broken result...
2D artists have requested a way to see in viewport the name of the material assigned to a stroke. This is a special request for 2D animation and required to manage complex drawings with multiple materials on it. We don't need add a separate option for this in the panel.
Now, when enable Name option in the Viewport Display panel, when you select a stroke in edit mode, the name of the material is displayed near the first point selected.
Design reviewed by @mendio and @pepeland
Recent changes to blendfile reading adding deferred reading of actual
data broke it, we cannot use the nifty `bhead + 1` to access data
anymore, since there is no guaranty that that block hass been fully
read.
Note that there is still one case in `read_file_thumbnail()`, however
loading of blendfile preview itself seems to be working fine... Maybe
@campbellbarton can double check that point (or knows of hands whether
it is OK there)?
Some uninitialized colormanagement data was breaking RNA acces for them,
exposed in batch preview management when generating previews for scenes
because we have to backup/restore scene and all its settings
before/after rendering it...
Do proper tagging for material changes, and avoid grease-pencil hack
in the relations builder.
The dependency graph code was forcing animation to be run for any
copy-on-write change of material. However, animation is not supposed
to be run on copy-on-write changes.
This commit possibly fixes the T58938.
The crash happens when a shader that is created in a rendering context is deleted and another shader with the same name/program of the deleted one is created in the same context and used in another context.
This is obviously not a perfect solution. However, to do proper
highlighting a more advanced Python parser would be necessary.
I think this patch implements a good heuristic to differentiate
between the cases when `@` is used for a decorator vs for
multiplication.
When `@` is directly followed by an identifier, it is interpreted
as decorated. Otherwise not.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4495
Need to use original mesh to get virtual modifiers list, otherwise
key datablocks will not be properly taken into account (since evaluated
mesh doesn't have key datablock).
This allows to have drivers on node trees.
Probably now it will be better to simply add parameters for any
ID node, to avoid such amount of manual work.
Initial idea was to not have nodes if it's not needed to avoid
any possible overhead. Having more robust system is probably more
valuable. We can always optimize overhead in one way or another.
When manually selecting a different image, this image will become pinned and
continue to show. The material node is not automatically modified as it was in
old Blender versions, only the image displayed in the UV editor.
Fixes T61239: confusing behavior when unwrapping non-square images. By showing
the relevant image by default it's more clear why it does aspect correction.
Make sure we don't reallocate arrays in the pointcache when not needed, the
size of a memory allocation can be slightly bigger than the requested size.
Also, use consistent check for shared cached in copy and free functions.
Main change are:
- the fresnel LUT is separated from the main GGX LUT.
- LUTs use sqrt(1.0 - NV) as roughness remapping. Improving precision and
removes needs for acos().
- LTC LUT is normalized by matrix middle component. Improving precision.
This is something what is caused by OCIO library. The patch
has been submitted there:
https://github.com/imageworks/OpenColorIO/pull/638
For until it is refined and checked we do workaround from
our side.
Solves weird situation when default display name is queried
from OCIO, but Default view being assumed to be set for it.
Now view is initialized to a default view of that display.
`id_single_user()` was still code from older ID management, now we can
use modern code to handle animdata duplication for us (which allows to
keep handling of nasty sub-data ID horrors like nodtrees in a single
place, much safer and simpler).
If we were not in x-ray/wireframe mode, the snapping would treat all
objects like they were in solid mode. This were not desirable if you
had objects that were in wire or bounding box display mode.
Now objects that are in bounding box mode is ignored (unless the object
is in edit mode) and objects with wire display mode is treated as if
they were in x-ray mode.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4493
This change updates the brush material in the topbar when a new material is created using copy button.
Related to task T62384
Thanks to @matc for suggesting the fix.
Still pending a problem whith user number.
This patch fixes T60171 by adding a dummy read from the `dummy` vertex attribute to `hair_get_pos_tan_binor_time` in `common_hair_lib.glsl`. Confirmed to work on my machine (macOS 10.14.4 Beta, Radeon R9 M295X).
According to my experiments regarding this issue, the problem is triggered when all of the following conditions are met: (a) the shader has no vertex reads; (b) the index buffer is ≥ 256KiB. I can't really give an explanation of this misbehavior because of the video driver's closed-source nature.
Reviewers: fclem
Reviewed By: fclem
Subscribers: zeddb
Maniphest Tasks: T60171
Differential Revision: https://developer.blender.org/D4490
This fix is a hack and will slow down update performance. But at least,
all shading features are supported. This is only in "normal" sculpt mode
(no dyntopo, no multires).
Steps to reproduce were:
* Maximize area (Shift+Spacebar in 2.7, Ctrl+Spacebar in 2.8)
* Open temp file browser (Ctrl+O)
* Cancel file browser (Esc) - should return to previous full-screen
* Press "Return to Previous" button
The previously maximized area would turn into a file-browser.
Note that the issue will still happen when opening old files saved while
in maximized area full-screen.
This bug was introduced by previous Pin fix commit. Maybe it's not the best solution, but this bug is critical and need a quick fix.
This commit fix the problem, but not totally sure this is working fine with pin.
@dfelinto Could you review and adapt if required.
During screen to workspace/workspace-layout conversion, the first layout
of each workspace would be activated. For temporary full-screens, this
used to be the full-screen layout (IIRC), apparently it's now the
the layout to return to on "Back to Previous". So the 'previous' layout
would be activated, not the full-screen one.
When actually pressing "Back to Previous" now, the operator would first
free the active layout (which it assumes to be the temp full-screen -
wrongly in this case) and then try to activate the non-full-screen one,
causing use-after-free.
When adding and removing weights from the python api, no depsgraph
updates were triggered. This would cause the mesh geometry to be out of
sync in the viewport if there were any modifiers that used the vertex
groups in question.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4439
When using a transform.select_orientation keyboard shortcut, the
orientation change would not update the viewport.
So if you had the move gizmo visible, it would not change transform
orientation.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4386
Bone selectability in edit mode now matches the behavior of pose mode
(can no longer be selected).
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D4469
Delete workspaces via the menu was not refreshing the workspace tabs
drawing. This way if you deleted the non-active workspace with the "e" shortcut
from the workspace tab context menu and clicked on the workspace tabs again, it
would crash.
A few notes:
* Deleting a non-active workspace is changing which workspace is then active,
which is really strange.
* Even when deleting the active workspace which workspace then becomes active
seems random.
* Using notifiers (ND_WORKSPACE_DELETE) to delete the workspace seems rather
abusing notifiers in my humble opinion.
This is not an important bugfix anyways, people probably would rarely
run into this. I just ran into it while investigating another bug.
This is a fixup for my own: 92d185faeb.
I'm also fixing the poll of the EEVEE_MATERIAL_PT_context_material
which would fail when we had no context.material available.
This partially reverts commit a77feabb51,
removing the shallow 'duplicate hierarchy' option from outliner.
Core changes from that commit in BKE_collections are kept.
Remove cddata layers which are not supposed to be written in .blend file,
when reading it.
This is a cheap and simple check, which also gives us forward compatibility
in case we decide some day to make a currently un-written layer type
to be written in .blend files.
Mistake in rBab0bc65c24bd, in conversion of static 'common' sets of
masks, BWEIGHT slipped in CD_MASK_MESH (probably from copy/paste from
CD_MASK BAREMESH).
Note that having it in CD_MASK_BAREMESH but not CD_MASK_MESH looks rather
counter-intuitive, but this is not the time for that kind of cleanup.
Went over whole conversion again, and found two other mistakes, should
really have recheck that before committing...
Source (i.e. other) mesh should not be modified in any case in modifier
evaluation case (this is forbidden by design and can lead to all kind of
threaded locks and crashes), and doing so even in operator case was
never a good idea either.
Now that we can specifically request needed data (poly and/or loop
normals) from evaluation code, we can finally get rid of those
computations inside data transfer/mesh remapping area.
This is hopefully the last remaining bit of this 'bad crashing code' in
datatransfer area.
In some cases (currently, only when using avanced mapping of loops),
code needs access to some cddata of the source mesh (CD_NORMAL...).
We need a way to inform calling code about that (actual issue requiring
this change is fixed in next commit).
This is merely making behaviors of this function a bit more explicit,
and avoid re-adding another CD_NORMAL layer to polys in the (unlikely)
case it would already have one.
It also handles CD_MASK_NORMAL in cd_dirty_poly, but this is more like
future-proof thing, this is not used anywhere currently afaik.
Part of the cleanup of the OpenCL codebase.
Single program is not effective when using OpenCL, it is slower
to compile and slower during rendering (when used in for example
`barbershop` or `victor`).
Reviewers: brecht, #cycles
Maniphest Tasks: T62267
Differential Revision: https://developer.blender.org/D4481
This change displays the stroke in the deformed position, but the edition is done in rest position.
This is no the final solution, but at least we have some visual feedback in edit mode.
The real solution is to create a full system to manage derived strokes in Edit mode, but this needs a lot of work and we haven't resources now to do it.
* Rename "Use Smoke Highres" to "Use High-resolution Smoke"
* Use 'Display' instead of 'Draw', following the naming conventions in 2.8
* Drop the 'Use' prefix in the UI, a "High-resolution Smoke' checkbox sufficiently communicates what it does.
llvm generates some header files at build time that differ between
debug/release causing linker errors when you used the release headers
for a debug build.
llvm generates some header files at build time that differ between
debug/release causing linker errors when you used the release headers
for a debug build.
This removes special dependency graph code that was intended to avoid GPU
shader recompiles by preserving the node tree and GPU material in specific
cases. This is no longer needed now that we have a general shader pass
cache that compares the generated shader code. The GPU material is already
being freed in material and world eval as well, so there's no point.
Note also that GPU materials are now safe to free from threads, actual
OpenGL buffer freeing happens delayed.
-Group gradient with the other paint tools
-Nest the two Sample tools together
-Keep painting tools first, secondary tools after
-Make sure toolbar doesn't jump when selection is enabled
Commit 208fafb to fix bevel tool units included a change to always
start modal in offset mode. This was may or may not be right but
broke a way in which users could set percent mode and then subsequently
start a modal that bevels in that mode. So going back to old way.
GHOST event handling does not need a window in the context, and restoring
it to the previous value is problematic if the window was deleted in the
meantime.
Convention was not to but after discussion on 918941483f we agree its
best to change the convention.
Names now mostly follow RNA.
Some exceptions:
- Use 'nodetrees' instead of 'nodegroups'
since the struct is called NodeTree.
- Use 'gpencils' instead of 'grease_pencil'
since 'gpencil' is a common abbreviation in the C code.
Other exceptions:
- Leave 'wm' as it's a list of one.
- Leave 'ipo' as is for versioning.
The UI was trying to use screen_context.c for its poll and draw
functions. So the active object and active object data and active layer
was used in the UI, instead of the context one.
Besides, for the material, the wrong context path was used altogether
when the active object was a greasepencil.
This would lead to all sort of pinning problems:
* A Mesh panel is pinned, but the active object is a grease pencil, the
grease pencil panels would show.
* If a Grease Pencil (data) panel is pinned, but the active object is not
the one pinned, nothing would show.
* Material panels and pinning were totally broken, showing the material
context for pinned mesh data panels even.
I also sanitized the name of the panels, their inheritance and poll
functions.
Reviewers: antoniov, brecht
Subscribers: billrey
Differential Revision: https://developer.blender.org/D4470
The issue was that `bNodeSocketValueVector` and `bNodeSocketValueRGBA`
don't store the value at the same location in the struct.
I kept the cases for `SOCK_VECTOR` and `SOCK_RGBA` completely separate
for now, because they only share code by coincidence and not because
they are actually the same. Eventually there could be a "Vector Input"
node similar to the "RGB" node.
Reviewers: fclem
Differential Revision: https://developer.blender.org/D4472
The fix was move the grease pencil engine before overlays
This also fix the origin not visible bug T62100
There is a limitation of this fix, but we cannot fix both problems. If you have object behind a transparent graese pencil objects, this objects will not be visible.
Thanks @fclem for his advices to fix this bug.
The SubdivCCG structure was re-allocated, but the cached one in
the sculpt session was still pointing to an original one.
This is rather a workaround, but is least evil from what i can
currently think. Remember something similar was also required in
2.7 as well.
Definitely need to revision this change.
Evaluate the dependency graph a bit later still. This kind of worked before
because it was using the drive namespace from the previously loaded .blend
file but that's of course wrong.
We already have different storages for cddata of verts, edges etc.,
'simply' do the same for the mask flags we use all around Blender code
to request some data, or limit some operation to some layers, etc.
Reason we need this is that some cddata types (like Normals) are
actually shared between verts/polys/loops, and we don’t want to generate
clnors everytime we request vnors!
As a side note, this also does final fix to T59338, which was the
trigger for this patch (need to request computed loop normals for
another mesh than evaluated one).
Reviewers: brecht, campbellbarton, sergey
Differential Revision: https://developer.blender.org/D4407
This is still is a dirty workaround.
Note that we are drawing the whole sets of point multiple times. While this
is ineficient, the main bottleneck is CPU transformation.
Empty object skipping now only skips the bounds, since you can select
the empty by it's center point and want to navigate to it.
Also share object skipping check between view all/selected.
This behavior matches back-face culled mesh objects,
where the wire outline doesn't depend on the viewing angle.
Applying this before empty visibility check for view framing,
since it's strange if viewing all gives different results
depending on back-face culling.
Note this was broken even in 2.7x.
We had a different logic for the plane wire, as for the plane itself.
And they were both wrong when changing the camera shift or the stereo
pivot.
Both of their logic is now unified and correct.
Also I had to create a new gpu batch for the quad wires, since there is
no state that allows me to filter out the geometry, and the square gpu
batch is quite different than the quad one (2d x 3d and orientation).
In accordance with T56648.
- Render Border -> Render Region
- Clipping Border -> Clipping Region
- Instancing 'Scale' -> 'Scale to Face Size'
- Instancing 'Rotation' -> 'Align to Vertex Normal'
Python API is left as-is, for now.
DIfferential revision: D4435
Volume itself was working, but buggy. The issue was that transparent shapes
does not write to depth buffer. But the custom MSAA resolve shader discard
pixels that did not write to depth making them transparent.
Transparent passes should not be inside the MSAA passes, this is also
a matter of performance.
Point is, you may want to duplicate your set of collections, but not
duplicate all their objects.
Some notes:
* Am rather skeptical about the practice of using operator name to
define behavior of some common exec code. Imho, that should be a single
operator with an enum to refine its behavior (@cambpellbarton may also
have an opinion here?). Left it as-is for now, because this seems to be
used by other operators too in that code. :/
* @dfelinto, @pablovazquez, @billreynish am not so happy with current
names, but cannot find really good short ones either... Also, shouldn't
we move those into a dedicated `Duplicate` sub-menu?
That one was an empty place-holder, BKE_collection_duplicate() is now
doing that. And its call from full scene copying is not needed
currently, 'deep copy' in that case is handled in editor's
`ED_scene_add()`.
Note that at some point, we might want to move that logic into BKE, but
for now let's keep thing as they are - working.
The mouse movement scale needed adjusting according to object scale,
since the amount is on the unscaled model but the viewport shows the
scaled one.
Also fixed proper units for amounts in bevel tool, as was already
done for modifier. Percent should be comfortably adjustable by mouse.
This is unreported. When using volume opacity 0.0, the wires were not visible
(nor were they in other cases, but this was the most obvious one).
The volume itself is still glitchy, and slightly worse than 2.79. I
still don't know how to fix this though. It seems a common problem
(see T62262).
Use annotations inside grease pencil drawing mode is something incompatible by design. Actually, the annotations are disabled in overlay panel for 2D template and the tool icon is not in the toolbar.
The unique way to get annotations was using D key, but this is wrong.
If you are inside drawing mode, all the events are captured by paint operator and to capture annotations, the operator must be canceled and the mode changed, but this change breaks several things.
It's not logic add annotation inside darwing mode, because you can simply add a new layer and write the text you want.
This change checks the mode and cancel the annotations if the mode is not thje right one.
State tracking works in pretty much all cases but calling the clear command
does change the write mask outside the draw manager. For now we just reset
the write mask before each pass.
Fix T62203 The selected bone is not highlighted inside the other bone.
`ED_mesh_uv_texture_add()`/`ED_mesh_color_add()` would always either
copy data from current active one, or (for UVs), generate default
'valid' UVs for every face.
This commit adds an option to not do that, just keeping default values
from raw CDLayer creation. It is only used/exposed from RNA API
currently.
This is especially useful for importer add-ons, since some formats
support multiple layers of those kind, as well as 'partial' dataset not
explicitely defining values for all mesh items.
Preliminary step to fix T62224.
We really do not need two 256 chars variables to hanlde renaming, a mere
pair of flags can handle the situation just as well.
Also, scroll to newly renamed item, will help when one want to find
again the directory they just added and rename.
At some point we'll probably want to refactor scrolling further (to make
it fully out of rename code/context e.g.), but for now think this will do.
This was changed (accidentally?) in d192d72312, now go back to behavior more
similar to 2.7. Python scripts should be able to set up drivers or do other
changes that will then be taken into account when evaluating the scene for the
first time.
The more tricky thing is that Python scripts now run before ED_editors_init,
but given what happens in there it seems safe.
Works as expected and mimics Cycles behavior.
The patch is a bit hacky: In order to not touch the lower level function,
we search for the active output inside groups (recursively) and the first
valid one is then copied (or extracted if you want) in the previous parent
nodetree. So we recursively extract the output node back to the main
nodetree while preserving the links through the nodegroups interfaces.
This way everything works as expected in gpu tree evaluation and bsdf
tagging.
Fix T61869 Material Output Node Inside Node Group Renders Pink in Eevee
Avoid allocation for each string,
improves redraw speed for text heavy views.
A contrived test showed FPS ~18.5% speedup but this doesn't represent
typical usage.
The problem was that `balance` expected that all node children
are set to `KD_NODE_UNSET` by default.
However, this might not be the case when `balance` is called
more than once.
The balance function might change the order of nodes even
when no new point has been inserted.
There were two problems:
1. `mesh_get_eval_final` has to be called with the evaluated object.
2. Particle systems have to have unique names within an object.
The depsgraph seems to use the particle system name as identifier.
This issue is actually independent of duplication.
The old code used a small hack to create unique names.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4451
Float2 are now a new type for attributes in Cycles. Before, the choices
for attribute storage were float and float3, the latter padded to
float4. This meant that UV maps were inflated to twice the size
necessary.
Reviewers: brecht, sergey
Reviewed By: brecht
Subscribers: #cycles
Tags: #cycles
Differential Revision: https://developer.blender.org/D4409
Do not force to link indirectly linked collections into current scene,
that is usually not desired. Note that user can always add this link
manually if they want.
All this 'implicit instantiation' post-linking process is rather hairy
to get it correct, hope this time it's not breaking something else...
The Lamp data was not always set. When using CUDA or CPU it was, but when using OpenCL
without `OBJECT_MOTION` `sd->lamp` not updated to the actual lamp. This made the
TextureCoordinate output the wrong normal when used in a light shader.
As the normal was incorrect it made the IES node render incorrectly.
(what is the default for the IES node).
By setting the lamp data when no `__OBJECT_MOTION__` compile directive is present makes
sure that the normal is correctly calculated.
Fix D4450
Reviewed By: Brecht van Lommel
ScreenArea->type is NULL-ified on read, and need to be initialized
(usually by `ED_area_initialize()`), but RNA can also access it before
it happens, so need to do it itself...
This reverts part of rBbc5482337669.
Problem with above commit is that the evaluated object seems to not have
partype, par1, par2, par3 copied from the original (yet). Using original
object instead now.
Second issue (when parenting to 'Bone Relative') is that the bones
BONE_RELATIVE_PARENTING flag is set on the original, but not the
evaluated bone (yet), setting this on both now.
Fixes T60623 (and part of T59352)
Reviewers: brecht, sergey
Maniphest Tasks: T60623
Differential Revision: https://developer.blender.org/D4309
This reverts the changes from ce927e1 to put the main and job threads on
node 0. The problem is that all threads created as children from these
threads will inherit the NUMA node and so will end up on the same node.
This can be fixed case-by-case by assigning the NUMA node for every child
thread, however this is difficult for external libraries and OpenMP, and
out of our control for plugins like external renderers.
The crash was detected in draw_manager, but the real problem was a wrong pointer that corrupted the struct, so the free function failed.
Redesign this area to keep pointer correct all the time.
since header regions can possibly overlap, accommodate by getting main
regions visible rect first.
Reviewers: sergey, brecht
Maniphest Tasks: T62190
Differential Revision: https://developer.blender.org/D4449
The screen identifier was set to the "layout_name" which isn't
guaranteed to be unique or even exist.
This meant `bpy.data.screens.keys()` could have duplicate keys,
even when not duplicated, these keys couldn't be used for
`bpy.data.screen[key]` access because internally ID's use names for
collection look-ups.
Make screens match other ID's, using the ID.name as the identifier.
Note: this change was from the initial workspace commit 7f564d74f9
Since then the layout selector has been removed so this isn't needed.
Keymap: E - Extrude Gizmo supported.
If the extrude points are the first or last of the stroke, the stroke is extended.
If the extrude points are in the middle of the stroke, a new stroke is created because the grease pencil strokes can be only with 2 extremes and fold the stroke to get a new point gets very bad results.
Still pending define a new icon. Also, it could be good to set by default XYZ axis in the gizmo.
Note: There is a change in the transform_gizmo_extrude_3d.c gizmo for OB_MODE_EDIT_GPENCIL. This change must be undo when the mode will be integrated into OB_MODE_EDIT, but while we have both modes, we need to keep this code in order to keep running the gizmo.
Requested by some users who prefer old wireframe precision.
Smooth wires are still enabled by defaults as they don't have a noticeable
perf impact.
Application restart is needed for changes to take effects.
Cache the result of DNA_elem_array_size which was being called
for each element when reconstructing each struct.
Also skip padding members during reconstruction.
Gives ~18% overall speedup loading 10_010_A.anim.blend (r3154)
from the spring project.# Please enter the commit message for your changes. Lines starting
Note, this was reverted recently because it caused T62181
Changes to DNA_sdna_patch_struct_member_nr were needed which
caused issues loading old 2.8x files.
There was a conflict between Object "In Front" and how the strokes were managed in Stroke panel for grease pencil datablock.
In order to keep consistency, the order of the object has been moved to Object level and now it's using the standard "In Front" flag. As all it is 3D, the option "Back" has been removed.
This also fixes the problem with task T62137.
Also related to commit 54ffc4e19d and D4405
This fixes the issue of long spikes on sharp corners by rounding them,
guaranteeing that the bleed will never exceed the bleed distance
set by the user in any part of the mesh.
See: D4437
Two aspects are addressed:
- Correct computation of dilation distance,
so that dilated boundaries remain parallel to the original boundaries
(and with the actual distance specified as bleed distance).
- Dilate with regard to adjacent seams
instead of adjacent triangle edges, for a more correct result.
This is especially important in the case of concave shapes,
where spikes could overlap with actual geometry.
See: D4436
Some old AMD drivers crash when a vbo with stride 1 is used a few times.
I have not found a real solution to this problem. So the solution was to use a vbo with stride 4 (which in theory is less efficient and takes up more memory space).
- paintcurves were hitting an assert in outliner_add_element()
- missing outliner update when adding a paintcurve
- paintcurves were not showing an icon [they dont have a dedicated icon,
took the one used elsewhere]
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4445
Added all 'Extended Cyrillic' chars I found in DejaVuSans (unfortunately
that one does not have the complete set).
Note: At some point we might want to switch to some more modern solution
than our Frankenstein font (thinking about noto one)...
Previously metadata was only shown for images which are saved
to disk.
This change makes it so that it's possible to see metadata right
after pressing F12.
This is fundamentally different from Shrink/Fatten for meshes, because the control points are not moved, only the radius.
Issue raised by Ronan Ducluzeau, name suggested by Campbell Barton
Not sure where that piece of code originates from, but trying to remap
usages of deleted scene to newly active scene in *whole* bmain is
really, really not the thing to do! Just use generic ID deletion code
here, no reason it could not handle the task properly. ;)
Looks like something forgotten at some point, define in callback flags
was there, just not used at all.
Added most obvious cases now (we are probably still missing some), and
proper handling for it in remapping code (when we hit such case, we
'convert' new_id to NULL and go on with it).
Related (as in, exposed by) T62076, but not the actual cause of that bug.
When move the mouse/pen very fast, the events were not enough to create smooth strokes.
I have used a simplified version of the same logic used with grease pencil objects to smooth annotations.
Cache the result of DNA_elem_array_size which was being called
for each element when reconstructing each struct.
Also skip padding members during reconstruction.
Gives ~18% overall speedup loading 10_010_A.anim.blend (r3154)
from the spring project.
- Avoid using white with low alpha to make grey.
Note that this is the second time we've run into this problem: T59626.
- Use a light backdrop when the icons are dark
so they're visible over a black background.
* Fix incorrect handling of children collections being linked more than
once in the hierarchy (previous code would make a new copy for each
link, instead of just re-linking the first copy for each extra link).
* Simplify some aspects of it (we do not need a GHash for new objects,
we can use ID->newid pointer instead, and some iterations can be done
directly on existing linked lists of old collection, instead of making
temp local copies of them).
* Move all copy logic into a single private recursive function (it was a
bit odd/disturbing to see calling function being indirectly called again
by the recursive helper one - not wrong, but that kind of code path can
quickly become problematic in recursive patterns).
* Added some comments about expected behavior of
`BKE_collection_duplicate()` depending on its booleans options.
Use 'size' instead of 'len' to represent the size of data in bytes,
'len' is used for the result of 'strlen' or the length of an array
in some parts of 'makesdna.c' & 'dna_genfile.c'.
Also clarify comments and some variable names, no functional changes.
As the 2D viewport color is white, if the object default color is white, when wireframe is enabled the object is invisible.
Now, the grease pencil object default viewport color is black.
There was a bug when selected Solid mode with Material or Texture mode. The textures were not visible.
Now, the mode is passed to shaders to decide if use the solid color or the result texture color. The mode is passed using an array with shading type and mode.
Also ensure elements fit evenly into the chunk size
causing allocations to be slightly smaller in some cases.
In own tests reduces overall memory use by about ~4.5%
for high poly meshes in edit-mode.
This enables static linking of libstdc++ by default when building using
`WITH_STATIC_LIBS`. This makes builds more portable for anyone making
static builds (in particular for older systems).
Reviewed By: brecht, campbellbarton, sergey
Differential Revision: https://developer.blender.org/D4393
When annotations enabled, image borders were drawn around whole area instead around preview image.
Reviewed by: Brecht
Differential revision: https://developer.blender.org/D4430
Revert "Outliner: Enable new faster 'Delete Hierarchy' code by default."
This reverts commit 491a98ca44.
It fails in the most basic of tests (see report). No point in leaving
this commit around until it passes the easy to test cases.
-Wireframe use Background color for X-Ray off
- Added support to Solid mode.
- Solid mode shows fill or not depending X-Ray.
- Solid can use Single, Material, etc.
- Wireframe and Solid mode don't show FXs.
Bug introduced on 012483b6e4.
Since we notify similar things when changing active and selected
objects, I believe we didn't notice this was missing a ND_OB_SELECT
notification before the small refactor to use the messenging system
exposed that bug.
We are mixing bool and fancy 3-in-1 func-set buttons in the outliner.
So they would return different pushed state in
ui_drag_toggle_but_pushed_state().
We now have a callback function that allows the button to set its own
pushed_button_state callback function.
Note: This is a bit of overkill since we are planning to change the
3-in-1 outliner buttons. That said, it may be nice to have, since in the
future we can mix those buttons for other things.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4434
The issue was discovered only after recent changes, but roots back
to much older changes.
What was happening is scene's ID recalc flags where never cleared,
which caused ensure_view_layer() to always run copy-on-write on the
scene. This resulted in certain runtime data being cleared, without
proper flag stored in the dependency graph.
This was caused by ID recalc clear function checking whether any ID
was tagged for recalc in that graph or not. This was happening due
to all areas using DEG_id_type_tag() which can only set flags on the
graph from viewport scenes, and could not inform render dependency
graph.
Now ID tyoe tagging is happening on per-graph level, which avoids
possibility of flags running out of sync.
In a bit longer term we also need to get rid of two functions which
are clearing flags: DEG_id_type_tag() and deg_graph_clear_tags().
Object of evaluated base is not yet copied, so we can not know whether
it has animation on visibility or not.
This issue was reported in T56635#630383.
Now that we have better options (duplicate collection and duplicate linked) there is no
longer need for the original dupli operator.
In fact, as it was it was of little use if you ever had nested collections.
As per the suggestion on T57064, this introduces two new options to duplicate collections.
We then have:
* Duplicate > Collection (New collection with linked content).
* Duplicate > Hierachy (Duplicate entire hierarchy and make all contents single user).
* Duplicate > Linked Hierarchy (Duplicate entire hierarchy keeping content linked with original).
Development TODO: `single_object_users` can/should use the new functions.
Reviewers: brecht, mont29
Subscribers: pablovazquez, billreynish, JulienKaspar
Differential Revision: https://developer.blender.org/D4394
In short, the settings to expand/collapse the Particles Animation Dopesheet expander
were no longer getting exposed, so the F-Curves attached to the particle settings
were not showing up in the channels list as that section was collapsed and couldn't
be opened from the UI.
Early on during the development of 2.8, we originally wanted to completely remove
the Particle System. Eventually that decision got walked back, and so particles
were reinstated. Well... most of the relevant code was! One of the areas that was
the most messed up during this process was the animation editor support for these
channels. It seems that there was almost a two-step removal process here -
the first pass tried to keep the channel definitions while removing all references
to particle stuff, while the second pass tried to remove the definitions completely
and/or re-added them in the wrong places, etc. To say the removal/reverting history is
here is "colourful" is an understatement...
This refreshes on cursor motion so it's worth avoiding redundant
updates, especially for multi-object edit-modes where many objects
aren't even near the object being selected.
This commit also moves to passing eSelectOp to circle select functions
in preparation for adding a select mode tool option.
This reverts commit b104b3cdcf and 04baefcc2f. Changes to core UI like
this should go through review, and doing them during Beta development is
not generally the right moment unless they fix an important problem.
Was caused by another fix in the area, and root to the wrong though that
transformation is only initialized from a fully evaluated dependency graph.
The latter one is not a case when changing transformation mode.
Solved by copying transform to an evaluated object.
Two changes:
Removed the explicit version for the macOS SDK, recent
versions of Xcode have a symlink to the newest SDK.
Fixed the build script for OpenMP by removing extra ' marks that
install_name_tool took literally and replaced INSTALL_PATH with
INSTALL_DIR.
Previously a modifier key-map type only worked when the same key was
enabled as a modifier as well.
This allows for users to assign an action to double-tap-shift for eg.
Workaround to tranform feedback not working on mac.
On some system it crashes (see T58489) and on some other it outputs
garbage (see T60171).
So instead of using transform feedback we render to a texture,
readback the result to system memory and reupload as VBO data.
It is really not ideal performance wise, but it is the simplest
and the most local workaround that still uses the power of the GPU.
This should fix T59426, T60171 and T58489.
Now it will be simpler for code jsut wanting to preserve custom normals
around to set them, without having to add same boiler plate code all the
time around actual code.
Currently modifier stack assumes there are no poly normals data passed
around, so in case a modifier generates such data, it has to clean it up
after usage.
Whole handling of normals is a bit annoying and week currently, we can
probably enhance that once fully per-mesh item type cddata masks are in?
Not a bug, but supporting preservation of custom normals in that
specific modifier makes sense, in game pipeline contexts.
Could also ease work of IO add-ons that want to export
triangulated geometry...
* Move Revert, Recover Last Session, Recover Auto Save to its own sub-menu.
Had three entries of the same category, and solves user reports on "Revert" being dangerously accessible under Open.
* Move up Link, Append, Import, Export as they are used more often than e.g. Save Startup File.
Now always refresh when the material changes. Depsgraph tag moved out
of the refresh function since that gets called on depsgraph update,
which should not trigger a second depsgraph update.
* Remove single item Armature sub-menu. Add Armature straight away, unless the menu is expanded (like with Rigify enabled)
* Group Light and Light Probe between separators
* Move the lesser used Speaker item below Camera
Original optimization idea was wrong: it is possible that some other
ID would reference an object which is also used by a base.
Rolled back to a bit more fragile solution.
In the future would be nice to make it somewhat less duplicated with
the builder itself.
Fixes assert failure (and possibly crashes) when adding grease pencil
object and switching to a draw mode.
While it's kind of common to use camel case in C++ this is not
currently agreed style for C++ in Blender.
Got confused by working on other areas with 3rd party libraries.
-Rename 'Specials' menus to 'Context' menus for Grease Pencil
-Make Grease Pencil contextual menus follow the design of the regular contextual menus more
-Add more useful operators to the contextual menus in the paint modes
Almost every pulldown menu and popover has a little dropdown arrow shape.
Unfortunately it is a bit wonky. The top of the right side of it is wider than the top of the left side. And both sides are narrower at the bottom than the top. It might be hard to see, but this image should help:
{F6728281}
The patch fixes the symmetry of the shape while keeping the weight as similar as possible. In the following image you can see the outline of the current version in red and this new version in green.
{F6728298}
With patch applied the arrow looks perfect:
{F6728302}
Reviewers: brecht, billreynish
Reviewed By: billreynish
Subscribers: pablovazquez
Tags: #bf_blender, #bf_blender_2.8, #user_interface
Differential Revision: https://developer.blender.org/D4424
This allows to update base flags to a proper state then object's restriction
flags are changed, without requiring to re-evaluate an entire tree of flags.
Some old unused flags are were removed by this change, and also disabling
menu items might not work the same as before. This is something we can bring
back if it's really needed (the way how flags are handled did change since
that interface code was done anyway, so code was looking weird anyway).
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4420
This allows dependency graph to evaluate drivers of those objects
and put them to a correct state. It will increase memory usage
since now we can no longer save it by skipping copy-on-write for
such objects. It will also currently make things slower, because
we do not have granular enough visibility update of components in
the dependency graph. Can do it later when the rest of the changes
are finished.
This commit does not update restriction flags on the base, since
that is somewhat tricky to do currently: need to somehow see whether
object is disabled due to flags on collection or due to own flags.
Differential Revision: https://developer.blender.org/D4419
The idea is to keep bases which are known for sure to be in the
dependency graph. Previously, this code was duplicating logic
around checking restriction flags, which becomes more hard to
maintain once we are moving towards to more comprehensive checks
about whether base is a part of evaluated scene or not.
As discussed with @billreynish this makes little sense now that we don't
have a dedicated textured mode. We don't have a superior texture or shaded
mode anymore and we also cannot mix different engines together (workbench
with eevee/lookdev).
The only feature it removes is the possibility to hide textures for certain
object in solid mode.
This actually makes more sense than removing them in the
load-post callback. During load, the file might register
timers that would be removed immediately.
Removed in aa7b013bd5 for performance reasons, however highlights
can't always be seen against specular shading, see: T55456#510873
Instead of having a highlighted inner-edge, use the active edge color.
Remove confirmation popup menu, just exit.
Note that this option is mainly for developers or people reviewing
blend files, see D4406 for discussion on reason for keeping this feature
while simplifying how it works.
Fix T61241
- Changing preview size does not affect drawn image size, only resolution.
-- Consistent behavior, when changing full-size / proxy / scene render size
-- Scopes are rendered in *same size* as source image
-- Over all, user does not have to readjust preview zoom.
Reviewed by: Brecht
Differential revision: https://developer.blender.org/D4315
It was being drawn when any Gizmo was highlighted.
There are several ways to solve this (creating a new parameter to the operator, checking the gizmos in the invoke or creating a global context).
Also, for a micro-optimization, all of those conditions in the `drawDial3d` could be done once in the invoke.
But since all of these changes involve changing the customdata (`TransInfo`), which is already a bit confusing with so many members, I thought it best to make minimal changes.
This makes the bones transparent when the object or the viewport display
type is Wireframe. This is in order to make things consistent.
In object mode all bones are fully transparent to not create more visual
noise if the scene is complex.
Another small addition is that the Bounding Box draw mode now works as
expected on armatures.
After a previous commit to use unique identifiers for Ghash key, I had missed to free the memory of the name key.
Thanks to Jacques Lucke for detecting the leak.
No need to have iterator loop in the view layer evaluation,
this only makes it more difficult to have base flags covered
by the dependency graph.
Other good thing is that we don't need to worry about whether
base has been removed from the evaluated view layer or not.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4414
Shear gizmo now uses a single orientation and sets
different ortho axis instead of constructing a different matrix for
each handle - allowing the redo panel to select orientations.
ApplySnapResize did not take into account invalid distances. Added check for this.
Reviewed By: Campbell Barton
Differential Revision: https://developer.blender.org/D4417
Transform orientation was previously related to constraints,
recent changes meant it was used even when not constraining to an axis.
Now transform orientation is separate from axis constraints.
Avoids mixing these in with regular variables in code-completion.
Use char for pad members except for 'void *', to make size clearer.
Removed/shrink a few redundant padding vars which were >= 8 bytes.
Note: Things were working fine if you were to pin the Grease Pencil
object, but not if you were pinning the GP data.
In too many poll functions context.object was being requested when
a simple context.gpencil_data would suffit.
Panels that are still not showing in pinning:
* DATA_PT_gpencil_display.
The panel needs to be split in sub-panels, leaving all object-dependent
properties in its own panel so we can poll it away, while showing the
rest.
* * *
This commit doesn't handle greasepencil material. In this case I
recommend we do as properties_material.py and have a generous poll(),
followed by different drawing logics whether or not we have an object.
In Blender face is for tesselated faces, that kind of miss-naming is a
pretty good way to shoot yourself in the foot at some point or another
(see T61979)...
Broken logic in rB181356edba04, thanks most likely to stupid naming
(face in Blender is for tessellated faces, use poly for BMesh polygons).
Cleanup to follow in next commit...
The problem was not only for instances, but for particles too, and produced segment fault.
For some reason due any internal modification of how duplicated objects are generated, the duplicated object are not available when the draw manager try to use runtime data.
Now, before drawing the particle or the instance, the pointers of the duplicated objects are reassigned to the original "real object" to get full access to runtime data.
-The pivot point and orientation of any transform are strongly related
-It matches the comma-key and period-key on the keyboard who are neighbours
-We get slightly nicer grouping this way, with the two axis-related options on the left and the two toggles on the right
Reviewers: pablovazquez, campbellbarton
Differential Revision: https://developer.blender.org/D4413
New icons from Andrzej Ambroż / jendrzych:
-New trash icon for deleting ID's and other data (currently unused)
-New icon for the Grease Pencil select between strokes mode
-New icon for Proportional Editing Root Falloff curve
Also adjustments for Jump to Next / Prev. Keyframe, Camera ObData, Point Light ObData, Light Probe Object and ObData, Collection & Save icons.
This also includes fixed/slighly refactored drawing code for marker lines.
The old code used the wrong height.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4411
Now it's possible use the different Wire modes (Single, Object & Random)
Also support for x-ray mode.
For random colors, the name of the object and the name of the layer is used.
Also some parameters cleanup.
New edit mode operator and post-processing brush option.
Trim works on a single GP stroke. It removes trailing points before and after the first intersection (or loop) nearest to the start of the stroke.
When there was no rotation the axis was zerod,
while not exactly a bug, it means changing the angle does nothing
and all axis-angle values are initialized with Y=1,
use this convention when resetting the axis too.
Displacement and Background kernels are selectively used, but always compiled. This patch will not compile these kernels when they are not needed.
Displacement kernel is only used for true displacement.
Background kernel is only used when there is a (Cycles)Light of type `LIGHT_BACKGROUND`.
Reviewed By: brecht, #cycles
Tags: #cycles
Maniphest Tasks: T61971
Differential Revision: https://developer.blender.org/D4412
The goal of this patch is to have limit the number of times
kernels needs to be compiled and are reused as kernels with
different compile directives can lead to identical same
binaries.
The implementation does this by stripping the compile directives.
and reshuffling kernels so the output is more likely to be the
same.
We focussed on the kernels where it was easy to detect and maintain
(bundle, bake, displace, do_volume and background). More optimizations
could be done but they are probably less obvious.
Merged the data_init and state_buffer_size kernels to split_bundle.
This patch will also remove empty kernels for do_volume and bake
when their features are not enabled.
When using the benchmark files there are less background, bake and
do_volume kernels compiled.
Fix: T61576, T61501, T61466
Reviewed By: brecht, #cycles
Differential Revision: https://developer.blender.org/D4390
- Add XYZ option.
- Orientation now works as expected.
Now a redo for rotation works logically,
setting the axis to Z & the orientation to view.
Resolves T57205
- Add a new panel to differentiate between viewport display and stroke options
- Use clearer naming for depth ordering and stroke thickness properties
Reviewers: antoniov
Differential Revision: https://developer.blender.org/D4405
When enable Wireframe mode in the shading type, all strokes are displayed as simple 1 pixel lines.
The color of the line is equal to the stroke color or the fill color if the fill is enabled and the stroke is disabled or has invisible alpha value.
In wireframe mode, all FX are disabled because sometimes the effects can make the lines invisible.
The modifiers are not disabled.
Still pending to decide if we must add support for Random colors, but not sure if this is useful in 2D.
BLI should always comes first, before DNA, BKE etc. And
`BLI_utildefines.h` should come before any other BLI (since it's some
sort of system include really, among other things...).
Thisi should help to reduce the noise in patches when adding stuff
like uint64_t members to DNA structs... ;)
While Buffering output is useful for file writing and memfile
compression it's redundant when the output is already buffered.
It doesn't make a significant difference for ZLIB
however it makes a moderate improvement for LZ4, see T56162.
Glyph cache is cleared by UI_view2d_zoom_cache_reset, when zooming V2D, but is required to calculate text height in UI_view2d_text_cache_draw
This caused text in strips to "jump around"
There was a comment in UI_view2d_zoom_cache_reset:
While scaling we can accumulate fonts at many sizes (~20 or so).
Not an issue with embedded font, but can use over 500Mb with i18n ones! See [#38244].
This fix sets the Base color in the principled BSDF Shader and in
the Material->r,g,b,a values. So the transparency and color are the
same when switching the "use nodes" option for the material
- The Collada exporter did not take care of
material transparency when nodes are turned off.
- recent change to use ma->alpha_threshold seems to have
been wrong. transparency is now taken from ma->a when
nodes are turned off.
This caused text in strips to "jump around"
There was a comment in UI_view2d_zoom_cache_reset:
While scaling we can accumulate fonts at many sizes (~20 or so).
Not an issue with embedded font, but can use over 500Mb with i18n ones! See [#38244].
Reviewed by: Brecht
Differential revision: https://developer.blender.org/D4389
Add a new Draw Mode to display panel in order to define the z-.depth order of the strokes using the real 3D position and not the 2D layer position.
This change makes possible to use VR with grease pencil drawings because the depth of the strokes change with camera position. Also, provide an alternative solution to tasks: T57859, T60325,
The parameter only works with 3D space depth ordering. The Back and Front depths are incompatible with 3D Space mode.
Options are:
- Back
- Front
- 3D Space->2D Layers (default)
-3D Space->3D Location (new mode)
VS2019 is binary compatible with the existing vc14 libraries and no
new libraries libs are required in svn.
VS2019 support requires cmake 3.14.
VS2019 is still in pre-release state, you are required to explicitly
select the pre-release version by using:
make full 2019pre
VS2019 is binary compatible with the existing vc14 libraries and no
new libraries libs are required in svn.
VS2019 support requires cmake 3.14.
VS2019 is still in pre-release state, you are required to explicitly
select the pre-release version by using:
make full 2019pre
On CCG side it is done similar to displacement, where we have
a dedicated functor which evaluates displacement. Might be seemed
as an overkill, but allows to decouple SubdivCCG from mesh entirely,
and maybe even free up coarse mesh in order to save some memory.
Some weak-looking aspect is the call to update normals from the
draw manager. Ideally, the manager will only draw what is already
evaluated. But it's a bit tricky to find a best place for this since
we avoid dependency graph updates during sculpt as much as possible.
The new code mimics the old code, this is how it was in 2.7.
Fix shading part of T58307.
It is always dangerous to add more shortcuts those days. But this way it
is consistent with 2.79 to a point.
When in edit mode (mesh, greasepencil, ...) 1-3 to change submode still
has priority.
When in posemode or greasepencil draw mode however, 1-9 can still be
used to temporarily get some collections out of the way.
The option is separated from the solid mode color option.
Random color uses the same method as solid mode.
Selection state is indicated by a brighter color that is outside the
brightness range of the unselected state colors. The active state is
indicated by the outlines that is, now, still drawn in wireframe mode.
Coloring of the selection / active outline is not optimal because it
can look ugly in some cases of color combination. But the outline color
is using index range coloring so it's not trivial to change the color of
the outline per object. For now we use the same outline color used in solid
mode for consistency and also still add an emphasis on the selected objects.
The Single color option uses the theme color. Maybe it would be nice to
change the name of it in a latter commit to avoid confusion.
The change in outliner and viewport visibility (897e047374) was made
assuming the bases of the render and viewport depsgraph were
independent. Thus we were deliberately setting base visibility when
rendering:
```
/* When rendering, visibility is controlled by the enable/disable option. */
if (mode == DAG_EVAL_RENDER) {
base->flag |= BASE_VISIBLE;
}
```
However, we were syncing data back to the original depsgraph, leading to
hidden viewport objects to re-appear.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4391
The goal is to make it easy to know which exact blender version
and built was used for a job on a farm. This includes but not
exclusively render farms. But same is handy for simulation tasks
as well.
This problem existed in 2.79 as well. The rigid body setting is related
to the scene the object was created.
We now clear all the rigid body properties of the appended objects to
prevent them from lingering in this state where they have settings yet
cannot be used in the simulation.
Reviewers: mont29, brecht
Differential Revision: https://developer.blender.org/D4380
There is still work needed to be done from multires side to fully
support smooth shading. So can't just always have smooth shading.
Roll back to a proper code in GPU side, the rest will be handled
from CCG side.
Delay loading all DATA sections of the blend file until they're needed.
Loading all data-blocks caused high peak memory usage especially with
libraries - since a lot of data may exist which isn't used directly.
In one test (spring project: 10_010_A.anim.blend),
peaked at ~12.5gig, dropping back to ~2.5gig once loaded.
With this change peaks memory usage reaches ~2.7gig while loading.
Besides this there are some minor gains from not having to read data
from the file-system and we can skip an alloc + memcpy reading data
written with the same version of Blender.
Removes the flat shader variant since the attrib is specified for each vert
loop in flat shaded mode. It was something leftover from the previous
implementation.
The multires sculpt drawing was a not working in smooth mode.
Also hidding was not supported by the wireframe overlay and flat shaded
faces.
Codewise it is cleaner and index buffers are only updated if the
smoothing changes.
Before 1bfbfa2810 this wasn't essential because the constraints
prevented the axes from being applied.
Now redo ignores constraints - the input values must be constrained.
When the line width was larger than the UI scale, there was not enough
space for thicker widget outlines to draw properly. Now widgets are made
a little larger to accommodate the thicker outlines.
Differential Revision: https://developer.blender.org/D4368
* Two cursors for horizontal and vertical split.
* Four cursors for each join direction.
* One cursor to indicate when splitting is not possible.
Differential Revision: https://developer.blender.org/D4264
Made it so that generated coordinate is always calculated.
Ideally, it will only be done depending on a current shading,
but code is quite deep, and doing smarter thing here will end
up in way bigger refactor.
First, make things working, and then make them fast if they
pop up in a profiles.
When using preview rendering through a camera or final rendering
the `scene.render.use_motion_blur` was not respected when building
the compile directives.
This patch will when building the compile directives check if
motion blur is enabled at all. This should lead to more efficient
kernels when no motion blur is needed.
Tags: #cycles
Differential Revision: https://developer.blender.org/D4387
Transfomr init code called just after duplication (presumably before
regular depsgraph update is executed) would erase new objects'
transflags.
This is more like a hack than a real fix, but since that transform piece
of code is already a hack... Other solution would have been to force DEG
to run after object duplication, think it's better to go with that
solution for now.
Not to mention to fact that dupli flags are put into transflag... ;)
Issue was a concurrent modification of an evaluated mesh by two
other meshes using it as source for custom normals data transfer.
Note that this fixes the crash (modifiers are strictly forbidden to modify
any data besides their own!), but now will have to add a new CD type to
be able to specifically request 'computed' clnors data layer, and not
only 'encoded' one, for source mesh...
This commit makes it so both Subdivision Surface and Multiresolution
modifiers are caching OpenSubdiv topology. This cuts down evaluation
time quite a bit, especially for meshes which don't have many extra
ordinary verticies.
Only working for animation. Other modifications like edit mode needs
more work to make topology cache preserved by copy-on-write.
It was copying the alpha from the foreground instead of background image,
which is not usually what is needed and inconsistent with the compositor.
Differential Revision: https://developer.blender.org/D4371
Constraint options had confusing behavior:
- When non were pressed, the orientation was ignored.
- When any were pressed, the orientation was used,
but only unconstrained axed could be adjusted.
Now constraining is only used for modal execution
so there is no need to show these in the interface.
When an orientation is selected, the XYZ values always transform
using that space.
Note, transform system should be refactored to support different
orientations w/o having to use constraints.
Addresses T57204
The bake kernels are also used during mesh displacement and light
importance sampling. We disabled the implementation of these kernels
when baking was not enabled.
These are place-holders with only a few items in each, as with the rest
of the context menus they need to be populated & organized.
Weight Paint 'weight' shortcut has been changed from W to Ctrl-F,
to co-exist w/ the context menu shortcut.
Was happening when looking for all intersections for transparent shadow rays
in the case the ray is degenerate.
Still quesitonable whether we should consider this a transparent or opaque
configuraiton. Ideally, we should prevent such rays from happening, but that
is another vector of debugging.
-Use factor for flame_vorticity, slice_depth, density & volume_density
-Use distance for surface_distance
-Use factor for mix factor in Data Transfer modifier
-Use prop_translation for pivot constraint offset
-Use distance for Shrink/Fatten Distance
-Use factor for Smooth Factor
-Use Factor for Randomize Uniform and Normal values
-Use Distance for Randomize distance amount
-Randomize Transform Scale was wrongly using distance
Default behavior is unchanged still, but can be changed in the keymap.
From testing I think this needs better visual feedback to indicate that
you are in local view, if the view does not move it's not as clear.
To keep running these tests relatively fast and practical to run often,
running it on all .blend files is a bit much. So now we only run it on
files from this directory.
Additionally this adds supports for following symlinks, so that you can
easily symlinks to other directories if you want to tests extra files
which may have linked libraries.
Using OpenCL MegaKernel has been slow and therefore not usefull.
This patch will remove the mega kernel from the OpenCL codebase
and the OpenCLDeviceBase class.
T61736: removal of mega kernel
T61703: baking does not work with mega kernel
Tags: #cycles
Differential Revision: https://developer.blender.org/D4383
Don't disable the save over popup through the keymap, just remove it entirely
from the code so that the file browser interprets the property correctly.
Was once again caused by an ambiguity of the entry/exit operations.
Only did for objects since those are the only one who needs this.
The rest types of IDs needs to be checked and only added extra
operations if needed (adding operations and relations causes some
overhead for evaluation, so need to be careful).
Add getter callback support for 'WM_HANDLER_TYPE_KEYMAP' type handlers
this is needed for key-maps which change based on the active tool.
Replaces 'sneaky_handler' hack which temporarily inserted a handler.
- Many factor properties were set to PROP_NONE,
even properties that had 'Factor' in the name!
- Some time properties were not set to PROP_TIME,
especially in Particles.
- Changed motion_blur_shutter to use a soft max value of 1 instead of 2.
Anything > 1 here is not physically correct
and makes no real logical sense.
- Changed display name of Dynamic Paint dissolve_speed to Dissolve Time,
since it's a time property, not speed.
We are still ditching the specular intensity of SSR (ssr_data.xyz).
But at least now there is some comment about it.
See T61704 for user reports on that matter.
Comments with the blessing of Clément Foucault.
The reported case was with the render output filename,
however the same was happening for file open.
Bug introduced on c20c203b82.
I can't find in the original commit any reasoning for the change
that introduced this bug.
If all axis and grid options were turned off, the grid in the main ortho views would not be rendered.
Now we force rendering of the grid regardless of the settings when in one of the main ortho views.
Reviewed By: Brecht Van Lommel
Differential Revision: https://developer.blender.org/D4378
This is a regression from Blender 2.79 where the usage
of <triangles> was already implemented, but unintentionally
removed in Blender 2.80
Also renamed variables for better reading.
There is a generic function to retrieve float and float3 attributes
`primitive_attribute_float` and primitive_attribute_float3`. Inside
these functions an prioritised if-else construction checked where
the attribute is stored and then retrieved from that location.
Actually the calling function most of the time already knows where
the data is stored. So we could simplify this by splitting these
functions and remove the check logic.
This patch splits the `primitive_attribute_float?` functions into
`primitive_surface_attribute_float?` and `primitive_volume_attribute_float?`.
What leads to less branching and more optimum kernels.
The original function is still being used by OSL and `svm_node_attr`.
This will reduce the compilation time and render time for kernels.
Especially in production scenes there is a lot of benefit.
Impact in compilation times
job | scene_name | previous | new | percentage
-------+-----------------+----------+-------+------------
t61513 | empty | 10.63 | 10.66 | 0%
t61513 | bmw | 17.91 | 17.65 | 1%
t61513 | fishycat | 19.57 | 17.68 | 10%
t61513 | barbershop | 54.10 | 24.41 | 55%
t61513 | classroom | 17.55 | 16.29 | 7%
t61513 | koro | 18.92 | 18.05 | 5%
t61513 | pavillion | 17.43 | 16.52 | 5%
t61513 | splash279 | 16.48 | 14.91 | 10%
t61513 | volume_emission | 36.22 | 21.60 | 40%
Impact in render times
job | scene_name | previous | new | percentage
-------+-----------------+----------+--------+------------
61513 | empty | 21.06 | 20.35 | 3%
61513 | bmw | 198.44 | 190.05 | 4%
61513 | fishycat | 394.20 | 401.25 | -2%
61513 | barbershop | 1188.16 | 912.39 | 23%
61513 | classroom | 341.08 | 340.38 | 0%
61513 | koro | 472.43 | 471.80 | 0%
61513 | pavillion | 905.77 | 899.80 | 1%
61513 | splash279 | 55.26 | 54.86 | 1%
61513 | volume_emission | 62.59 | 61.70 | 1%
There is also a possitive impact when using CPU and CUDA, but they are small.
I didn't split the hair logic from the surface logic due to:
* Hair and surface use same attribute types. It was not clear if it could be
splitted when looking at the code only.
* Hair and surface are quick to compile and to read. So the benefit is quite
small.
Differential Revision: https://developer.blender.org/D4375
There is a generic function to retrieve float and float3 attributes
`primitive_attribute_float` and primitive_attribute_float3`. Inside
these functions an prioritised if-else construction checked where
the attribute is stored and then retrieved from that location.
Actually the calling function most of the time already knows where
the data is stored. So we could simplify this by splitting these
functions and remove the check logic.
This patch splits the `primitive_attribute_float?` functions into
`primitive_surface_attribute_float?` and `primitive_volume_attribute_float?`.
What leads to less branching and more optimum kernels.
The original function is still being used by OSL and `svm_node_attr`.
This will reduce the compilation time and render time for kernels.
Especially in production scenes there is a lot of benefit.
Impact in compilation times
job | scene_name | previous | new | percentage
-------+-----------------+----------+-------+------------
t61513 | empty | 10.63 | 10.66 | 0%
t61513 | bmw | 17.91 | 17.65 | 1%
t61513 | fishycat | 19.57 | 17.68 | 10%
t61513 | barbershop | 54.10 | 24.41 | 55%
t61513 | classroom | 17.55 | 16.29 | 7%
t61513 | koro | 18.92 | 18.05 | 5%
t61513 | pavillion | 17.43 | 16.52 | 5%
t61513 | splash279 | 16.48 | 14.91 | 10%
t61513 | volume_emission | 36.22 | 21.60 | 40%
Impact in render times
job | scene_name | previous | new | percentage
-------+-----------------+----------+--------+------------
61513 | empty | 21.06 | 20.35 | 3%
61513 | bmw | 198.44 | 190.05 | 4%
61513 | fishycat | 394.20 | 401.25 | -2%
61513 | barbershop | 1188.16 | 912.39 | 23%
61513 | classroom | 341.08 | 340.38 | 0%
61513 | koro | 472.43 | 471.80 | 0%
61513 | pavillion | 905.77 | 899.80 | 1%
61513 | splash279 | 55.26 | 54.86 | 1%
61513 | volume_emission | 62.59 | 61.70 | 1%
There is also a possitive impact when using CPU and CUDA, but they are small.
I didn't split the hair logic from the surface logic due to:
* Hair and surface use same attribute types. It was not clear if it could be
splitted when looking at the code only.
* Hair and surface are quick to compile and to read. So the benefit is quite
small.
Differential Revision: https://developer.blender.org/D4375
When hiding the curve handles/points previously, the control points would still be drawn (loose verts).
Now we hide everything related to the handle if it is hidden.
Reviewed By: Clément Foucault
Differential Revision: https://developer.blender.org/D4373
Now that we are looping over all image users that were previously ignored,
it shows some scene pointers are invalid. Always clear them on load, and
don't keep scene permanently in the image user except for the image editor.
Otherwise the pointer can go out of date.
This commit only contains some of the changes in the diff.
Some require more discussion/work.
Differential Revision: https://developer.blender.org/D4337
Do not instance linked object immediately in scene, this was never a
good idea and is doomed to fail nowadays, with complex relations between
objects, collections and scenes.
Instead, this commit refactors a bit linking code to add loose objects
to current scene *after* everything has been imported, and ID pointers
have been properly remapped to new ones - i.e. once new linked data is
supposed to be fully valid, just like we were already doing with
collections.
As a bonus, it means we do not have to pass around scene, view3d etc. to
`BLO_library_link_named_part_ex()` and co.
Currently it's effectively a boolean for file-select handlers.
Prepare for refactoring event handlers into their own types (keymap,
operator, gizmo, ui & dropbox) to help make logic easier to follow.
Committing this since it does fix broken logic (previously in that
condition obdata would always be set to NULL, since
`BKE_object_runtime_reset()` is called before).
However, this has presumably been broken that way since 05/2018, so
maybe that whole condition is not needed anymore? Or NULL pointer was
working as well here?
@sergey eyes are required here I guess ;)
Caused by rBae2b677dcb5a70f5, Object.runtime has lot of weird specific
handlings in depsgraph...
For now modified `deg_backup_object_runtime()` and
`deg_restore_object_runtime()` to mimic previous behavior regarding
Object bbox (i.e. pass it around, instead of wiping it clean).
Reported in T61660.
The dependency graph now handles updating image users to point to the current
frame, and tags images to be refreshed on the GPU. The image editor user is
still updated outside of the dependency graph.
We still do not support multiple image users using a different current frame
in the same image, same as 2.7. This may require adding a GPU image texture
cache to keep memory usage under control. Things like rendering an animation
while the viewport stays fixed at the current frame works though.
The initial idea of using char pointer was to save some
memory since the dependency graph was kind of the one
with the main database.
Nowadays dependency graph should be separatable from the
main database and being self-sustainable.
Other issue which was caused by this pointer is the
re-tagging of operations during relations update: it is
possible to have node which as tagged for update but had
the owner of the name removed (i.e. driver or bone was
removed).
Patch by @sergey.
Note that this is really a bad thing actually, ideally we should never
get that situation (IDs in Main referencing temp IDs outside of it).
That can lead to many possible similar cases...
Fixing that is not trivial though, so for now we'll have to live with
it, until we have migrated *all* of our temp datablocks generation
outside of Main's.
Previously, hair strands of zero length of too few control
points would have been ignored. This is fine for a render
without motion blur. But once motion blur is enabled it is
becoming more tricky to match topology.
Even more, it was causing access (and possibly writes) past
the array boundaries in case when time step 0 ignored some
strands and steps around it did not.
If this is becoming problematic for BVH to do reliable
intersections this is to be solved on the BVH builder side.
The export from Blender to Cycles shouldn't really make
decisions there.
Was happening when image buffer had cryptomatte pass, which can easily
exceed 530 bytes used by the buffer.
Now default buffer is bumped to 1K, and also allowed to be heap-allocated
when really need bigger buffer.
Possible optimization is to allocate buffer once, but in practice those
re-allocations will not happen often, so keeping code simpler is not an
issue. Just something for a rainy day.
* Added a TL;DR first paragraph summarizing that one shall not keep any
reference to Blender data when modifying its container.
* Added some info about fact that adding items to some data containers
(like Collection) can also invalidate existing items (due to array
re-allocation).
* Added a Do/Don't example which shows a crash after adding some items
to a collection.
Related to T61297.
This introduce the wireframe batches. Creating the indices buffer does
not seems to slow down the sculpt in my testing (but it is kind of hard to
test reliably)
This includes a bit of cleanup in gpu_buffers.c.
This will effectively make the AA passes thicker in some cases but it is
required for better AA on wireframes. The trick is to occlude the wire
passes so that they do not output fragment that could be behind actual
geometry.
- Makes it possible to show a vertical line for every marker in the graph editor.
- Makes the marker line visiblity optional in the sequencer and graph editor.
Request from @hjalti.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4348
All the controls were just really thrown in there without any proper
organization.
This gives it more structure.
- Correct use of sub-panels to communicate hierarchy and sections.
- Use flow layout for toggles.
- Use consistent names for "Bake Type".
Smaller adjustments to the Light Probe properties layout.
- Correctly use alignment for multi-property values.
- Correctly use sub-panels.
- Correctly use PROP_FACTOR for visibility_bleed_bias and
visibility_blur.
Some properties were accidentally hidden for particle fluids.
- Made sure we show the Forces and Integration
sub-panels for particle fluids.
- Slightly re-ordered the sub-panels here, so that the same sub-panels
are at the top for Newtonian and Fluid particles.
- Separated the Fluid Interaction sub-panel so we can give it a unique
name.
- Removed lingering unnecessary 'Keys' label in the Keyed physics.
Some RNA errors are quite similar, use clog for consistent logging that
always includes the file, function and line number - making errors
quicker to troubleshoot.
This affects point, spot and area lights. Sun light strength remains without
a unit. This change does not affect .blend file compatibility in any way, as
with the rest of the unit system it's purely a display and editing feature.
Not used for Cycles yet, that will be done after unifying the settings with
Eevee.
The code assumed all datablocks were read from .blend files saved with the
same version. This restructures the Cycles versioning code to take into
account libraries.
For llvm 6 the visual studio integration was 'not great' and we had
our own, which broke when llvm 7.0.1 came out. llvm now has properly
supported integration available on the VS market place hence we can
retire our custom support.
When modal map was introduced, left out handling of what
happens when bevel is made active tool in toolbar and user
starts bevel by clicking and dragging.
Fail to build on errors in new names - without this renamed values
would be written to DNA breaking backwards & forwards compatibility.
Note that errors in old names aren't detected.
This allows us to rename struct & struct members in the source code
without changing the file format.
This is useful because the code becomes increasingly confusing when
names such as oops, ipo & dupli aren't used anywhere except DNA headers.
dna_rename_defs.h is used to define renaming operations.
The renaming it's self will be done separately.
In this case we simply create a new screen area that copies the currently
fullscreened area.
Note: At the moment there is no indicative in the non-main window that we are in
fullscreen. That happens because this information is part of the bar and we have
no topbar in this window.
The problem
===========
For armature, if the active object was in pose mode and the newly
selected armature data (not the pose, but the edit armature) we would
get a crash.
For mesh objects, the issue would happen with the active object in object mode.
Then the new selected object would switch to edit mode, however the overall
mode would still be object mode, leading to unsynced mode across the objects.
The solution
============
Using shift to extend selection makes current selected (compatible)
objects to go to edit mode as well. Otherwise only the newly selected
object will switch to edit mode.
This also works if you are in edit mode for a curve, and click in a mesh icon.
This also changes the rules for multi-object editing (or rather, how we
put objects in and out of it). Now shirt is also taking into
consideration there. So if you simply click in another mesh object's
data, it will have only the newly selected object in edit mode.
To reproduce the old behaviour you need to use shift to include the
newly selected object in the multi-edit party.
Reviewers: campbellbarton
Subscribers: brecht
Differential Revision: https://developer.blender.org/D4344
was caused by rBc6e3a20ab60b, copied node was actually added to the
nodetree, resulting in an endless loop.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4360
At some I unified the "move to collection" with the remove from all collections
functionality. That meant that even when we were still to keep the object in one
of the collections we would clear its rigid body data.
Now why to even remove the rigidbody data when removing an object from all
collections? That mimics the 2.79 behaviour when we were to unlink an
object from a scene. I suspect it has to do with the rigid body data
being tied to the scene rigid body. Which is a strange design anyways
(add to the list?) since an object can be in more than one scene.
Is mainly used by drivers variables. The slow part was about
iterating over all pose channels to find the one which has a
given constraint.
Now we build a look up table, so this operation is way cheaper,
Brings down relations update time from 0.7sec to 0.4 with Spring
production file.
This triggered an "almost parallel" case in setting the
offset meet points, which is OK but code needed improvement
put the meet point in a more accurate place.
This ia fix for part of the report T61214.
This partially reverts bf2c5217 and makes it so animation is evaluated
for datablocks which were never evaluated within the dependency graph.
Not ideal, but safest way currently.
Animation for already evaluated datablocks will only be evaluated on
manual edits, so the initial bugfix is still valid.
This is unreliable for cases when multiple dependency graphs
are to be updated.
The only reason why it was attempted to be made is to deal
with cases when ID appears in the dependency graph for the
first time. But even then it should be smart enough bring
itself to an up-to-date state without any extra tricks.
This patch implements a workaround to get the multithreaded compilation from D2231 working.
So far, it only works for Blender, not for Cycles Standalone. Also, I have only tested the Linux codepath in the helper function.
Depends on D2231.
Patch by lukasstockner97, jbakker, brecht
job | scene_name | compilation_time
----------+-----------------+------------------
Baseline | empty | 22.73
D2264 | empty | 13.94
Baseline | bmw | 56.44
D2264 | bmw | 41.32
Baseline | fishycat | 59.50
D2264 | fishycat | 45.19
Baseline | barbershop | 212.28
D2264 | barbershop | 169.81
Baseline | victor | 67.51
D2264 | victor | 53.60
Baseline | classroom | 51.46
D2264 | classroom | 39.02
Baseline | koro | 62.48
D2264 | koro | 49.03
Baseline | pavillion | 54.37
D2264 | pavillion | 38.82
Baseline | splash279 | 47.43
D2264 | splash279 | 37.94
Baseline | volume_emission | 145.22
D2264 | volume_emission | 121.10
This patch reduced compilation time as the split kernels and base
kernels are compiled in parallel. In cycles debug mode (256) you can set
unmark the opencl single program file, what reduces the compilation time
even further (bmw 17 seconds, barbershop 53 seconds).
Reviewers: brecht, dingto, sergey, juicyfruit, lukasstockner97
Reviewed By: brecht
Subscribers: Loner, jbakker, candreacchio, 3dLuver, LazyDodo, bliblubli
Differential Revision: https://developer.blender.org/D2264
This patch implements a workaround to get the multithreaded compilation from D2231 working.
So far, it only works for Blender, not for Cycles Standalone. Also, I have only tested the Linux codepath in the helper function.
Depends on D2231.
Reviewers: brecht, dingto, sergey, juicyfruit, lukasstockner97
Reviewed By: brecht
Subscribers: Loner, jbakker, candreacchio, 3dLuver, LazyDodo, bliblubli
Differential Revision: https://developer.blender.org/D2264
Right clicking on a menu item now closes it's sub-menus and opens
the button's context menu.
This is needed for adding them to the quick favourites menu.
Resolves T58729, T61015.
Also capture event to avoid Move transform.
Note: Now it's using a report message. Maybe this can be removed, but without the message, the event is captured by move transform.
Previously, the curve self snapping would only snap to points that were
earlier in the curve structure. This was because of a simple coding
snafu of using break when meaning to use continue.
This was caused by curves pointing to each other
creating a cyclic dependency.
While the dependency graph detects this, generating a mesh for render
recursively generates data which cashes in this case.
Add in a check to detect cyclic links.
Note, this bug exists in 2.7x too - but only crashes on render
since 2.7x didn't use 'for_render' when converting data.
By default wire would z-fight against the surface.
Increase the bias, also don't adjust the 'w' component
since it causes bias that depends on the view direction.
This allows secondary keys on tap.
Currently Z-key to toggle wireframe and tilde for navigation.
This is currently experimental, if users like this the preference
can be kept and used where appropriate.
Add shortcut since this has been removed from the context menu,
now it's in the mesh normals menu which isn't so convenient to access.
Shift-N is already used to recalculate normals,
this fits the convention of Alt removing/reversing.
There is no reason not to duplicate Actions too here, especially when
Materials' Actions are pretty much impossible to edit from current UI
(afaik, DopeSheet editor does not has any way to change them?).
Values outside the 0..1 range produce negative colors, so now clamp to that
range everywhere. Also fixes improper handling of hue > 2.0 in some places.
That one is utterly ugly fix really, but unfortunately a proper one
would require some changes to our RNA (or more precisely, pyrna) code,
so that when we subscript a dynamically generated RNA collection, the
item is somehow duplicated (and probably 'assigned' to its py object?),
before the temp RNA array memory is freed...
Usual legacy/history crap in NodeTree code... Datablocks's specific
freeing code should never, ever do refcounting management, this is
handled by higher-level code from BKE_library area.
Nuke away old nodeCopyNode(), much better to use new BKE_node_copy_ex(),
which behaves as expected for the various optional flags that can be passed.
This also removes the need to handle ID refcounting in calling code
(ugly!) and allows us to remove an even uglier name from our codebase! :D
Note that this fixes three related issues actually, that bug was also
affecting copy/paste of nodes, and 'Separate with copy' operator (the
latter being actually fully wrong, since it was not refcounting
anything, not even node->id pointer...).
More or less same code was being executed twice during ID copying.
Makes no sense to add yet another switch-by-ID-type to handle
specificaly runtime data during ID copying, we already have
BKE_xxx_copy_data() functions for that.
For OIIO 2.x we must use unique_ptr. This also required updating the
guarded allocator for std::move to work. Since C++11 construct/destroy
have a default implementation that also works this case, so we just
leave it out.
rB55c281415b67 removed 'BLENDER_RENDER' as a COMPAT_ENGINE but the
cycles addon checks for this in its get_panels() function.
Adding this back for now.
Reviewers: brecht, billreynish
Maniphest Tasks: T61499
Differential Revision: https://developer.blender.org/D4346
Was a use-after-free during relations update.
Now we do similar dependency graph tags, but without any
extra animation update logic, which was accessing various
pointers.
Was found when looking into a file from T56635.
Using ID_LIGHT or ID_ID for "Lamp" meaning, "Light" without context
being for 'not heavy'.
That rename of data-block was not really nice on that side of things :/
Related to T43295.
Tested on an `AMD Radeon HD 7570M`.
It seems that a VBO containing only `unsigned bytes` or `unsigned shorts` can't be read correctly in a shader.
Strange that if the index buffer repeats the drawing of the vertices (as was done before rBa04dd15193e6) the problem disappears.
The disadvantage of this solution is that the memory size for a selection VBO increases by about 4 times.
But the loss in optimization is negligible.
Thanks to @fclem for pointing out the possible source of the problem and reviewing the fix.
It is supposed to be already evaluated. If for some reason it's not
doing such direct evaluation will not be reliable anyway (indirect
dependencies for example).
This fixes an assert part of T61431.
Ensures that object which is set for instance-vert or instance-face
is evaluated prior to metaball. This is because metaball will request
list of instances during evaluation.
This should fix issue reported T61431 in release build. The assert is
still there and is to be addressed separately.
This keymap was used in the old grease pencil and now must be removed.
The keymaps for brush are:
F: Change Radius
Shift + F: Change strength
Ctrl+F -> Removed.
Make Difference a default value for boolean modifier operation property.
Currently operation property of the boolean modifier is set to Intersect, which is the least frequently used boolean operation of the three available. It is also goes out of sync with Intersect (Boolean) tool, where Difference is a default operation.
Reviewers: mont29, brecht, sergey
Reviewed By: mont29, brecht, sergey
Subscribers: mont29, brecht, campbellbarton, sergey, billreynish
Tags: #modifiers
Differential Revision: https://developer.blender.org/D4340
This makes it so modifiers are using object transform prior
to the rigid body simulation, and then result of modifier
stack is fed to the solver.
Solves dependency cycle which was happening when object's
modifier was dependent on the modifier transform.
While now it is not possible to change simulation, things
are somewhat more clear and reliable in other ways.
For example previously, solver was using derives mesh from
a previous step in time, which causes unfixable simulation
issues (with intersections and such)
Fixex T57589: 2.79 Rigid Body Sim. Does Not Behave The Same In 2.8
Fixex T61256: Compositing scenes causes crash, but rendering separately does not
Fixes T61262: Armature and rigid body crash
Fixes T61346: Rigid body with modifiers incorrect work
This is what modifiers are to use to indicate that they depend
on a transformation of the object itself.
Currently should be no functional changes, but in the future
this will allow to easily change transform operation depending
on whether there is a simulation associated with the object.
There is an issue of hair being completely messed up when
switching to a simulation view layer for Autumn.
Restoring back the code which was re-setting particles on
file load. This will re-set unbacked particles on file load
but this appears to be happening in 2.7 as well.
Can not reproduce bugs which were fixed in this area recently,
so maybe it's finally tackled (fingers crossed!).
Currently only a single function was duplicated which isn't so bad,
this change is to allow DNA versioning code to be shared between
dna_genfile.c and makesdna.c.
Blender is typically used maximized or fullscreen,
load maximized instead of attempting to fill the screen bounds.
To load un-maximized use '--window-border' argument.
See D4332
`BKE_modifier_get_evaluated_mesh_from_evaluated_object()` used by
modifiers needing access to other objects' geometry probably skipped out
of the radar when cage and final evaluated meshes were added to
BMEditMesh? In any case, we do not need to duplicate (and then free!) a
temp mesh from editdata anymore, and we can even add instead a parameter
to get cage instead of final. Also makes modifiers code a bit simpler.
The render layer name is now always included. Best to keep these consistent,
so that animation denoising and sample merging works the same for both and
tests can be the same. Ref D4311.
It is now possible to adjust the group node background alpha.
The defaults are the same as before, but you can now adjust the alpha
level via the theme preferences (and the alpha value is no longer hard
coded).
Fix T61406: Particles don't render
Consider initial dependency graph evaluation as a file load.
Is still resetting too much, but that we can solve later.
This adds a cycles.denoise_animation operator, which denoises an animation
sequence or individual file. Renders must be saved as multilayer EXR files
with denoising data passes.
By default file path and frame range come from the current scene, and EXR
files are denoised in-place. Alternatively, a different input and/or output
file path can be provided.
Denoising settings come from the current view layer. Renders can be denoised
again with different settings, as the original noisy image is preserved along
with other passes and metadata.
There is no user interface yet for this feature, that comes later.
Code by Lukas with modifications by Brecht. This feature was originally
developed for Tangent Animation, thanks for the support!
Differential Revision: https://developer.blender.org/D3889
This adds a cycles.denoise_animation operator, which denoises an animation
sequence or individual file. Renders must be saved as multilayer EXR files
with denoising data passes.
By default file path and frame range come from the current scene, and EXR
files are denoised in-place. Alternatively, a different input and/or output
file path can be provided.
Denoising settings come from the current view layer. Renders can be denoised
again with different settings, as the original noisy image is preserved along
with other passes and metadata.
There is no user interface yet for this feature, that comes later.
Code by Lukas with modifications by Brecht. This feature was originally
developed for Tangent Animation, thanks for the support!
Also update relations when modifier texture changes.
Basically same as rB6e00415a85a9, rBca2680aaeb1 but this time for
VertexWeight modifiers
Reviewers: sergey
Maniphest Tasks: T61185
Differential Revision: https://developer.blender.org/D4305
We may want to use 'TEST' BCode in the future for including data
besides thumbnails. This allows negative values to be used w/o
attempting to load a thumbnail.
Currently the preferences have both tweak and drag threshold,
this is confusing because most actions users would consider
dragging use the 'tweak' setting.
Now one drag threshold is used for both, with a maximum limit of half
the button unit-size in case of dragging UI elements.
Also added keys for toggling harden normals,
and cycling through miter types.
Still to do: add some shortcuts for affecting the
spread value for arc miters.
Most artists agree that RGB by default is not as flexible as HSV.
It's just the first time it opens anyway, since it will remember whatever
was set last like it always does.
When one is indirectly linking collections, better add collection to the
scene, than instantiating its objects into master collection of the
scene. That is much cleaner.
Noted/related to T61141.
Issue is, ob->id.us is not relevant anymore here, since several
collection might be referencing it inside of a same scene, that is still
only one usage from user perspective...
Note that for now we are just counting scenes instantiating an object,
time will say wether we need more refined/complete check (as a reminder,
most [all?] other Object usages are *not* refcounting ones).
Yes, we do can undo an ID deletion now.
However, this requires extra care in UI 'remapping' to new IDs step
(when undoing, we do not fully reload the UI from saved .blend).
Otherwise, new UI (i.e. one from saved .blend file) might reference
IDs that where freed in old bmain (the one before the undo), we cannot
use those to get ID name then, that would be a nasty use-after-free!
To prevent this, we generate a GSet of all valid ID pointers at that
time (i.e. those found in both old and new Main's), and ensure any ID
we try to remap by its name is in that GSet. Otherwise, there is no
possible remapping, just return NULL.
While ideally we could have a complete detailed list of deleted IDs,
that would require more work UI wise, think for now we can live with
just a rough summary.
Related to T61209.
make_duplis_collection() depends on the collection object cache,
which was not freed upon object viewport disable change.
The best way to reproduce the bug was to instance the default
collection, disable the cube, save and re-open the file.
Now even if you set the original cube to be enabled, you wouldn't see
the instanced one until you forced collection cache to be freed (e.g.,
changing a collection disable state.
Fix T61289: Emitting particles from instances not working properly
The first issue has been re-introduced by a code which was dealing
with missing hair after opening the file. That was re-setting all
particle systems all the time because modifier flags were not copied
back to original. This made every modifier run to be seem as an
initial file open.
Now we copy flags back to an original modifier. But also we are
trying to not do any resets unless needed in that case. This way
we can preserve in-memory caches.
Other part of the change is related on re-setting particle system
if number of mesh elements changed. But we only do it if the
modifier has been already evaluated once.
Caused by an error in rBe65784a0519e.
And since we are going over loop triangles anyways, we can remove the
part quecking for quads [remainder of tessface era] entirely.
Reviewers: campbellbarton
Maniphest Tasks: T61309
Differential Revision: https://developer.blender.org/D4324
Mainly 'X' icon for Delete, which was already on modifiers and constraints,
but not for objects. Select icon for object selection and refresh for library reload.
Show One Level, Hide One Level, Show Active, Show Hierarchy were taking
four items on the context menu when they are not accessed that often
from the context menu (they all have shortcuts).
The "View" name is used to match other editors View menus.
Mainly the first of their category or when they need to be highlighted:
* Delete
* Enable Viewport/Render (match icons to make a visual connection)
* ViewLayers (it's used pretty often so it makes it easier to find)
Also group Show/Show All, Hide/Hide All together.
Originally, when transferring all source data layers to destination
meshes, code would abort in case destination did not have all needed
layers, and creating them was not allowed.
Now, it will instead transfer data to layers that exists, merely
skipping source ones for which it cannot find a matching destination.
Some developers were using undo for their scripts, this allows for undo
pushes in background mode, however - as with 2.7x, undo isn't
initialized at startup in background mode.
See replies to T60934
As far as we know this wasn't widely used, and relied no storing data
in the temp dir which may be cleared on reboot.
More generally, alternative behavior for a core area like file IO
is not something to keep if it has unresolved issues.
See D4310 for details.
That guy was still from the era where only way to remove an ID was to
save & reload the .blend file. Use modern code instead, should also be
much, much more efficient in big production files.
And that’s another nice occasion to add/test new batch ID deletion code, too. ;)
Related to T61276 Make Single User unlinks original object.
This avoid edges covering a part of vertices.
This comes at a (very minor) perf cost as vertices can cover some edges
pixels and early discard them with the depth test. But this only happens
in artificialy dense mesh and is not a real problem for common cases.
This make sure only one line is drawn per edge.
It makes the function mesh_create_edit_loops_points_lines() non-thread safe
but this is fine as of now because nothing is multithreaded at this point.
Also this is the only function use this flag so it might be OK.
The side effect is that we don't need to use depth test in edit mode
overlay so the masking artifact will not appear.
This make it (theoriticaly) compatible with all supported hardware with
consistent results.
Also we now draw the lines with analytic anti-aliasing instead of relying
on MSAA (which offers less benefits in our case).
The remaining aliasing comes from edges cut in half by the mesh which is
not rendered with MSAA. Hopefully this is not too much distracting and only
happen if the face is almost parallel to the view.
This is backporting a change from 2.8, which may help solve crashes when
activating a window. Previously bringTabletContextToFront() would call
tablet API functions with NULL tablet, which may crash on some drivers.
Ref T60811.
Is available when doing "View -> Show Metadata". Will draw all the
fields which are not part of the stamp at the bottom of the image.
Couple of hand-picked fields are ignored, since those are not very
useful to be seen.
Aimed to ease review of rendered shots.
Reviewers: brecht
Reviewed By: brecht
Subscribers: fsiddi
Differential Revision: https://developer.blender.org/D4316
Is a missing do-verisoning code in e3d31b8dfb.
Unfortunately, at this point it is rather tricky to tell old and new
hair dynamics modifiers apart. Probably easier to accept possible
breackage of the files which were created in 2.7 and saved during
2.8 which had incomplete do-version code.
- studiolights were not installed to their proper subfolder (thus not
recognized on blender restart)
- they were actually loaded with a wrong path which could lead to
deletion of the original source file when uninstalled again
This is a follow-up to rBb44e6f2b3d32, for some reason that issue was
not detected back then: in some cases, DEG_iterator_objects_next() will
free the temp list of dupli objects once it does not need it anymore,
henceforth freeing the dupli_object_current memory of the DEGObjectIterData
that we are storing in the RNA_Depsgraph_Instances_Iterator struct.
And yes, the uglyness of that hack is getting even better now...
Found while trying to export dupliobjects with FBX...
Note: We still change it to the collection we are directly isolating/making
visible and its parents (in the case of isolating). But no longer its children.
Feedback and discussion on D4011. The motivation is that if we don't keep those
locked the disable state becomes useless.
Was introduced by point cache reset on manual edits. Needed to
split evaluation and introduce an explicit init key, which allows
to hook up relations which are "monitoring" manual edits to the
channel.
Noticed while looking into T61190.
Was visible with certain configuration only, is a numeric
instability caused by degenerate ray direction.
Not sure the distribution is correct, just fixing crash
which was caused by usage of watertight intersection.
This is a request by the studio here to make it possible to see how
many samples were used to render a specific shot or a frame. It is a
bit more tricky than simply stamping number of samples from a scene
since rendering is happening in multiple ranges of samples.
This change makes it so Cycles saves configured number of samples for
the specific view layer, and also stores start sample and number of
samples when rendering only a subrange of all samples.
The format used is "cycles.<view_layer_name>.><field>", which allows
to have information about all layers in a multi-layer EXR file.
Ideally we can store simplified "cycles.<field>" if we know that there
is only one render layer in the file, but detecting this is somewhat
tricky since Cycles operates on an evaluated scene which always have
single view layer.
The metadata is shown in the Metadata panels for clip, image and
sequencer spaces.
Example screenshot which shows the metadata:
{F6527727}
Reviewers: brecht
Reviewed By: brecht
Subscribers: fsiddi
Differential Revision: https://developer.blender.org/D4311
This is the internal implementation, not available from the API or
interface yet. The algorithm takes into account past and future frames,
both to get more coherent animation and reduce noise.
Ref D3889.
Prefiltering of feature passes will happen during rendering, which can
then be used for denoising immediately or written as a render pass for
later (animation) denoising.
The number of denoising data passes written is reduced because of this,
leaving out the feature variance passes. The passes are now Normal,
Albedo, Depth, Shadowing, Variance and Intensity.
Ref D3889.
Most of this code is deprecated for many years already and does not
work at all in Blender 2.8.
Reviewers: brecht, aligorith
Differential Revision: https://developer.blender.org/D4271
The old function was numerically very unstable for 2 reasons:
computing the square and then subtracting the results.
In the example in T60935 all precision was lost and it returned the distance 0
for all points.
I also removed the `depth` parameter since it wasn't used and computing
it would have made the code more complicated.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4308
Need to synchronize simulated frame back to original object.
Solves the lag during transformation, but amount of floppyness is
lower for some reason. Final animated object behaves the same as
in older Blender though.
rBec3357e03ab1 introduced multi-object snapping.
Seems like this was done without mixed-mode selections in mind.
So code assumed that all selected objects are actually armatures [which
can fail].
In 2.7 this was not a problem, because code only took active object into
account, 2.8 was iterating over all selected_editable_objects.
Now just iterate over objects in posemode instead
Reviewers: brecht, dfelinto
Maniphest Tasks: T61051
Differential Revision: https://developer.blender.org/D4287
This complicated handling of undo steps in a generic way
especially switching between undo systems that stored data to ones
that accumulated changes.
Now each undo system must treat it's steps as check-point,
internally it can apply/rewind changes.
This commit also fixes projection paint where the object mode wasn't
following the undo steps.
Needed to fix T61196, supporting clipped back-buffer in the 3D view
which is done outside the draw module.
It was also inconvenient having DRW_shader_* versions of GPU_shader_*
API calls.
- Clipping distances are now supported as a shader configuration
for builtin shaders.
- Add shader config argument when accessing builtin shaders.
- Move GPU_shader_create_from_arrays() from DRW to GPU.
Now collection and objects can be either:
* Disabled for all the view layers.
* Hidden for a view layer but not necessarily for all others.
* Visible for a view layer but not necessarily for all others.
Regarding icons: Whatever we decide to use for the "Hidden for all view
layers" needs to be a toggle-like icon. Because when viewing "Scenes"
instead of "View Layer" in the outliner we should be able to edit the
collection "Hidden for all the view layers" as an on/off option.
The operators are accessible via a Visibility context menu or shortcuts:
* Ctrl + Click: Isolate collection (use shift to extend).
* Alt + Click: Disable collection.
* Shift + Click: Hide/Show collection and its children (objects and collections)
Things yet to be tackled:
* Object outliner context menu can also get a Visibility sub-menu.
* Get better icons for viewport enable/disable.
Note:
* When using emulate 3 button mouse alt+click is used for 2d panning.
In this case users have to use the operator from the menu.
See T57857 for discussion.
Patch: https://developer.blender.org/D4011
Reviewers: brecht and sergey
Thanks to the reviewers and William Reynish and Julien Kasper in
particular for the feedback.
[re-committing]
We still control this in the viewport collections visibility menu. But
now we are actually changing the visibility of the collections, not of
the objects.
If a collection is indirectly invisible (because one of its parents are
invisible) we gray it out.
Also if you click directly in the collection names, it "isolates" the
collection by hiding all collections, and showing the direct parents and
all the children of the selected collection.
Development Note:
Right now I'm excluding the hidden collections from the depsgraph.
Thus the need for tagging relations to update.
If this proves to be too slow, we can change.
This was deliberately disabled since I didn't get the drawing working
originally. It is fully working now.
Note: camera lens widget still needs to be fixed since it still draws it
wrongly.
This will help with upcoming outliner visibility icons with 3 states.
It is done by using the icon to identify the state. If that is not unique
there is no visible difference to users anyway.
When generating a mesh from a curve object, do not generate temp objects
and curves in main, but rather as 'localized' copies.
This is cleaner, and might add a marginal speed-up in some cases (like
rendering thousands of curve objects), since we save some processing.
Note that this is the function behind py API's `Object.to_mesh()` too.
Use first combined pass if possible. Is not ideal but better than
showing completely empty image.
Also, covers quite a lot of usecases when movie clip editor is
used to review animation render of single-layer renders but with
multiple passes.
This adds a new geometry shader (specific to edit mesh for now) that
reproduces the effect of glLineWidth > 1.0, since this is not supported on
all platform.
This fix could be generalized to other shaders later.
- Add manual depth offset to vertices and edges.
- Revert to plain edge decoration.
- Fix active edge coloring.
- Remove active face display if not in face selection mode.
- Add wide line support.
This is work in progress. Look is not final.
This align data VBO data structure used for edti cage drawing to the one
use for normal drawing.
We no longer use barycentric coords to draw the lines an just rasterize
line primitives for edge drawing. This is a bit slower than using the
previous fast method but faster than the "correct" (edge artifact free)
method. This also make the code way simpler.
This also makes it possible to reuse possible and normal vbos used for
shading if the edit cage matches the
This also touches the UV batch code to share as much render data as
possible. The code also prepare for edit cage "modified" drawing cage (with
modifier applied) but is not enabled since selection and operators does not
work with modified cage yet.
Quite straightforward: first, convert metadata from file to
stamp data which is stored in the render result, and then
for every requested layer/pass use that as a metadata.
Noted those as missing in XXX comments some time ago, running again on
that code I still see no reason for this missing feature, so now when
doing a full scene copy, including duplication of Freestyl's linestyles
and scene's greasepencil data, their potential Actions will also be
properly duplicated (like it was already the case for world, and scene
itself).
This is a slightly more risky commit, as it is very difficult to fathom
all that may happen when localazing IDs. Would not expect any issue
though.
Note that a big TODO remain to refactor fully that ID localization
process (for 'shading IDs'), it's still doing pretty much same thing as
regular out-of-main copies, but the infamous ntree topic makes it
delicate to handle...
Turns out most of our 'local working copy' cases can use same set of
flags.
Note that this commit adds LIB_ID_COPY_CACHES to all our local meshes
copying, however this is no-op since that flag is unused during mesh
copying... We may want to add another set of flags without that one at
some point, but for now it would not be useful imho.
No local work copy is expected to need preview data, at least it should
not. Part of copy flags cleanup, done in separate commit in case
something goes wrong here...
Those two first sets of flags should represent some common use cases.
The goal here is to reduce verbosity of calls to BKE_id_copy_ex, and
help make it more obvious the 'common behaviours' of ID copying across
codebase.
Previously, if double-quotes appeared in the KeyingSet.bl_description field,
these would cause a syntax error in the resulting .py script export of the
KeyingSet. Since single quotes are even more likely to appear
(e.g. as apostrophes), we now use triple quotes here.
Unreported bug, noticed earlier when investigating T61010.
This commit adds a datablock filtering option for cache files channels,
so that a shot with lots of these in addition to standard animation
(e.g. the Spring production files) don't become bogged down by these.
Furthermore, these channels also respect the "Only Selected" toggle too now.
This feature is intended only for testing,
to automate simulating user input.
- Enabled by '--enable-event-simulate'.
- Disables handling all real input events.
- Access by calling `Window.event_simulate(..)`
- Disabling `bpy.app.use_event_simulate`
to allow handling real events (can only disable).
Currently only mouse & keyboard events work well,
NDOF, IME... etc could be added as needed.
See D4286 for example usage.
* Use simple default view transform for color pickers, as Filmic does not work
well for all types of colors. We better handle this with an option and tagging
of colors as emissive or albedo like.
* For solid/workbench we also no longer use Filmic, as there is not enough contrast
and it's not really needed since this is not physically based lighting.
* For lookdev always take into account the view transform and look. Other view
settings like exposure are only taken into account if scene lighting is used,
since these are often dependent on scene light intensity.
Fixes T61022, T57649, T59363.
Now when remove points from a cyclic stroke, the last island is joined with first island in order to fill the gap of the cyclic.
This change affects not only to cutter, but to any delete process in cyclic strokes.
BF-admins agree to remove header information that isn't useful,
to reduce noise.
- BEGIN/END license blocks
Developers should add non license comments as separate comment blocks.
No need for separator text.
- Contributors
This is often invalid, outdated or misleading
especially when splitting files.
It's more useful to git-blame to find out who has developed the code.
See P901 for script to perform these edits.
For ears it was already how we evaluate modifiers. There is no
need to go more granular than is actually needed. And no need
to use some obscure prefix for operation.
BF-admins agree to remove header information that isn't useful,
to reduce noise.
- BEGIN/END license blocks
Developers should add non license comments as separate comment blocks.
No need for separator text.
- Contributors
This is often invalid, outdated or misleading
especially when splitting files.
It's more useful to git-blame to find out who has developed the code.
See P901 for script to perform these edits.
Was happening due to missing relation from geometry to
transform component. Did not happen in old dependency
graph because that one could never evaluate geometry
prior to transform.
Scopes were moved to properties area, so need to adjust
the optimization part of tagging.
Ideally, tagging will always happen (and happen for free)
and then drawing code will update scopes when they are
actually displayed. But this is outside of the scope of
this fix since requires some design changes.
The issue was caused by dependency graph resetting particles
when evaluating copy-on-write version of object. Solved by
only doing reset from dependency graph on user edits.
Other issue was caused by modifier itself trying to compare
topology and reset particles when number of vertices or faces
changed. This isn't reliable, since topology might change even
with same number of elements. But also, since copy-on-written
object initially always have those fields zero-ed the reset
was happening on every F12.
The latter issue is solved by moving reset from modifier stack
to places where we exit edit/paint modes which might be changing
topology.
There is still weird issue of particles generated at some
weird location after tapping tab twice, but this is not a new
issue in 2.8 branch and is to be looked separately.
Was missing synchronization of current frame to the original one,
which is one of the issues why point cache does not properly reset
on edits.
Also clear recalc flag on original particle system.
Ideally we need to get rid of recalc on a particle system, since
that is not really covered by tagging system of dependency graph.
Some summary of changes:
- Don't use DEG prefix for types and enumerator values:
the code is already inside DEG namespace.
- Put code where it locally belongs to: avoid having one
single header file with all sort of definitions in it.
- Take advantage of modern C++11 enabled by default.
Was happening when value of one shape key was driving property of
another shape key of same datablock.
Solved by making shape key blocks properties more granular.
Makes it more explicit whether RNA property is used as a source
dependency for something else, or whether some other dependency
is being hooked up to evaluate that property.
This is necessary when adding a new keyframe to a fcurve
that also has a driver.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4278
This removes a bunch of animation/driver evaluations and recalc flags that
should be redundant in the new depsgraph, and were incorrectly affecting
the evaluated scene in a permanent way.
Still two cases that could be removed if the depsgraph is improved, in
BKE_object_handle_data_update and BKE_cachefile_update_frame.
For physics subframe interpolation there are also still calls to
BKE_object_where_is_calc that should ideally be removed as well, though
they are not known to cause keyframing bugs.
Differential Revision: https://developer.blender.org/D4274
Some features are incompatible with multithreading and reliable evaluation
of dependencies. We are now removing them as part of a bigger cleanup to
fix bugs in keyframing and invalid animation evaluations.
* Dupliframes have been removed. This was a hack added before there were
more powerful features like the array modifier.
* Slow parent has been removed, never worked in 2.8. It was always
unreliable for use in production due to depending on whatever frame was
previously evaluated, which was not always the previous frame.
* Particle instanced objects used to have their transform evaluated at
the particle time. Now it always gets the current time transform.
* Boids can no longer do predictive avoidance of force field objects,
but still for other particles.
Differential Revision: https://developer.blender.org/D4274
While this is harmless, it did cause T55399 in the past.
Sculpt adds it's own undo steps, so don't request the operator type
to do it too.
This is consistent with other sculpt operators.
Issue actually exists since ages, probably 2.7x update system forced all
armature proxies to be fully refreshed after an undo?
In any case, proxy_from should only be reset for 'local' proxies (i.e.
directly linked datablocks), not for linked proxies...
Only keep this function when drawing to avoid COW overhead that reduce performance.
After some changes I did some time ago, the use of original ID was not required and this only added depsgraph overhead and problems.
This change solves the problems with updates in render mode.
Related to T57484 and the changes requested by Sergey.
Kind of funny to see that this has been missing presumably since the
first version of library linking in Blender, and only gets noticed now.
Then again, that was not really a critical issue, iirc write code
ensures all libraries directly used get properly written, even if flags
are incorrect.
When using `--cycles-resumable-num-chunks N` to render a subset of the
samples, having N close to the total number of samples causes rounding
issues.
For example, a file configured for 250 samples and 150 chunks should
have 1.6666 sample per chunk. The old code rounded this to 2 samples per
chunk, which would result in too many samples being rendered. When
rendering a single chunk this doesn't matter much, but when larger chunk
ranges are rendered with `--cycles-resumable-start-chunk` and
`--cycles-resumable-end-chunk` the rounding errors start to add up.
By multiplying with the number of chunks to render first, and only round
to integers after that, this issue is solved. In the above example,
rendering 3 chunks will correctly render 5 samples rather than 6.
When the requested number of chunks is larger than the number of samples
there will be duplicate samples (that is, sample N appearing both in
chunk M and M+1). In this case a warning is printed to stderr.
This is needed for T50977 Progressive render: use non-uniform sample
chunks.
Reviewed by: sergey
Differential Revision: https://developer.blender.org/D4282
Add the ability for undo steps to request memfile undo step added after
them, useful for mode switching, where we need the data to exist for
undo to enter the mode.
This is only partially working, because some bAnimListElem items do not
have any ID pointer set (for wome mysterious reason...), notably the
'group' ones.
Will re-assign to @aligorith for that, this code is rather complicated
and hard to follow (with all those macros ;) ).
Support the alpha channel use of the object color in solid mode.
The Transparency effect is still using the Xray algorithm and not
true Alpha blending.
We've had many reported crashes on Windows where we suspect there is a
corrupted OpenCL driver. The purpose here is to keep Blender generally
usable in such cases.
Now it always shows None / CUDA / OpenCL in the preferences, and only when
selecting one will it reveal if there are any GPUs available. This should
avoid crashes when opening the preferences or on startup.
Differential Revision: https://developer.blender.org/D4265
This adds the posibility of having certain materials transparent in solid
mode. The option is (for now) per material only and thus only shows in
material color mode.
This uses the same rendering technique as Xray mode.
Note that objects are not considered transparent for selection with this.
- Add noise to remove undersampling artifact
- Create 2 mipmaps to the scene color buffer in order to have bigger blurs
- Replace blur2 with a 3x3 median filter that doesn't dilate the highlights
- Use temporal accumulation to remove noise
For some reason all of this exacerbate some bleeding issues happening on
far foreground elements from near foreground elements. The actual problem
was already happening before but was not really noticeable. It needs some
more work to be fixed.
The inset operator uses 0.01 as default for the inset.
When the face is very small than this default is very confusing (see T60226).
The simplest fix seems to be to just use 0 as default.
This is similar to the extrude operator which uses 0 as default as well.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4273
The subdivision method for getting corner shapes has a fullness
parameter which had been set by eye before. This change uses fullness
as found by offline search process to best match the superellipsoid
octant in the cube corner case (except cube corner case is still handled
by other code). This somewhat improves the look of cube corners with
inner arc miters, however.
Currently names are used for edit-mode undo-steps,
any changes to Main ID names cause lookup failure (crashing).
This commit ensures any undo steps that use ID lookups have the same
mem-file undo state loaded that was used to encode the steps.
Renaming also has an undo push added (last commit).
Some more tests showed no issue, so now feeling reasonably confident.
Old, 'safer' one remains available through setting debug value to 666,
for a few more weeks.
This bring macOS on par with Windows and Linux. It uses the OpenMP library
added to our precompiled libraries.
Custom flags are set because FindOpenMP from CMake below 3.12 does not support
AppleClang, and more recent versions do not work with our custom directory
location either.
Differential Revision: https://developer.blender.org/D4257
This is a quick workaround to prevent the crashes with multi-view.
The ultimate solution can be plenty, and would turn around refactoring
Cycles to handle multi-view internally, so that depsgraph could be freed
before render with no problems.
Reviewers: brecht, sergey
For the complete discussion check: https://developer.blender.org/D4239
This was very simple to reproduce, just turn on Freestyle and press render.
Now to the truth of things. Most (if not all) of
~BlenderStrokeRenderer() can be removed. I believe this was done back
when freestyle was using G.main, and since we gave freestyle its own
main we can just leave the cleanup for later.
I will leave this for freestyle maintainers to think over though.
Note: There is a chance this was the issue reported on T57890. I will
wait for the reporter to confirm this as fixed though.
This commit adds another optional check (when `--debug-io` is set) on
write .blend process, to check and ensure all shape keys have their
'from' pointer properly set to their respective user ID.
This is intended to be used as debuging tool mostly (to try to detect
when/why some of those pointers can become NULL).
For now, it also systematically perform same checks/fixes when loading a
.blend file, to fix all broken ones laying around. Later we might move
that usage to a do_version instead, but for now think it's safer to
always perfom it (and it's rather cheap process anyway).
Does not make sense to keep that with BLO_writefile.h, this can also be
used by read code, and some other parts of Blender (like ed_undo.c
currently)...
Render-type panels are only shown when the relevant type is active anyway.
Saves a click especially when using object or collection as render, since
you _have_ to set an object or collection to use it.
- Compute samples positions on CPU.
- Use 3x3 Box blur instead of 2x2.
- Implement bokeh parameters.
With this commit, dof performance is almost negligeable.
The quality is a bit lower than before but can be improve. Also now big
Circle of confusion are supported (up to 200px).
Cost is ~1.25ms on AMD Vega with a 2560p viewport than full HD and
pretty shallow depth of field.
Coc downsampling and dilation is not used anymore for now (commented).
The algorithm used is borrowed from :
http://tuxedolabs.blogspot.com/2018/05/bokeh-depth-of-field-in-single-pass.html
This makes it possible to have a decent blur for foreground over defocused
background in one pass only.
The algorithm is using a gather approach that is much faster
than the scatter approach used in Eevee. This makes it possible to have
custom bokeh shapes (not implemented yet) which would be impossible with
a separable gaussian technique.
The blur is done in 2 steps. The first one define the shape of the bokeh
and the second that fill the undersampling.
A downsample max-CoC tile texture speed up the gathering process.
Was generating INVALID_FRAMEBUFFER here instead of failled texture alloc.
Add safety asserts in gpu_texture.c and clamp minimum size to 1 inside
GPU_offscreen_create.
Artists requested to show the stroke while drawing a new stroke using a material with fill color only, because it's very difficult to see the stroke.
Now the stroke shows always but using the fill color, not the stroke color because maybe is not set.
The issue was caused by the hair step checking whether
particle system needs to have path cache. This was done
in a way which was traversing an entire scene and was
checking every object for particle instance modifier.
Ideally, path cache should be an own operation in the
dependency graph. Or at least, this flag should be set
by dependency graph builder, similar to curve's path.
Since the code was broken already (it was only checking
first particle instance modifier), it is easier to
remove the buggy code, solve the crash and move on for
now.
If this causes an issue, simply set particle system to
be rendered as path.
Fixes crash with playback of Spring scenes.
This reverts commit 1d908bffdd.
Enough uses of repeat last expect skip-save properties to be set,
transform being the most obvious example T60777#605681.
I wanted to avoid operators having account for two kinds of 'skip-save'
but this may be unavoidable.
Originally I wanted to avoid adding draw manager specific ifdef's all
over generic shaders however this isn't needed in so many places.
Also there are shaders that are only used by the draw manager so
duplicating them only to have the original unused doesn't make sense.
The integrator maximum number of closures was not set properly for the CPU/mega
kernels to match the actual available memory. Before relatively recent code
refactoring we did not use this value in those kernels so it worked fine.
this can now be found in the sidebar View panel
- uses existing 'lock_camera_and_layers' but renames the property to
'use_local_camera'
- uses RNA_def_property_boolean_negative_sdna to flip the value
- remove the local view code in
rna_SpaceView3D_lock_camera_and_layers_set
- update Python code
- update Addons code will be separate commit
Fixes T60756
Reviewers: billreynish, brecht
Maniphest Tasks: T60756
Differential Revision: https://developer.blender.org/D4247
Do not see why flags from loaded file should be skipped when we do not
load UI, this is not related to UI...
Think we can keep flags from file in both cases, should this raise some
other issue we'll just have to fine tune masked flags in each case
separately.
Do not see why flags from loaded file should be skipped when we do not
load UI, this is not related to UI...
Think we can keep flags from file in both cases, should this raise some
other issue we'll just have to fine tune masked flags in each case
separately.
Even though it makes sense logically to have displacement actually displace
the mesh, this is causing a lot of confusion for existing users that are used
to the previous behavior. Further, since Eevee does not support displacement
yet and the discrepancy between the viewport and final render is problematic.
Operator relys on 3DView and was failing from Topbar and Properties
Editor. Now tries to find the biggest 3DView and uses that.
Reviewers: brecht
Maniphest Tasks: T60133
Differential Revision: https://developer.blender.org/D4215
Use zoom steps lower than 1. This allows to zoom out a high-res
image. For example, before it was not possible to maker 4K image
to fit on FullHD monitor.
Also, don't force zoom to be above 1. Not sure why that was done,
but this disallows zooming out.
It is still not possible to zoom in higher than the window size
allows. In order to support this the player needs to be refactored
in a way that allows to decouple zoom from window size.
Fixes T59177: Animplayer extreme zooms in when playing rendered animation
There were two problems:
1. The scopes were only updated when the "Scopes" category is active,
but this category has been removed in Blender 2.8.
2. The scopes moved from the TOOLS to the UI region.
However the update-code still searched for the "Scopes" category
in the TOOLS region.
Both problems are fixed with this commit:
1. Scopes have there own category again.
2. The update code is in the correct draw function now.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4245
The problem was related to cache data that was removed from memory before the FX finished. This could affect to any FX.
Now all the information is saved in the FX itself in runtime struct to keep memory safe when cache memory is released.
DRW_shader_get_builtin_shader can replace GPU_shader_get_builtin_shader
when we need to support clipping.
Use this for loose point & wire drawing in object mode,
clips edges in lattice edit mode.
PROP_SKIP_SAVE is often used as a way to detect the difference between
adjusting options from the redo panel and initial execution.
Repeat last operator was executing with skip-save properties set,
preventing operators from initializing them based on the context.
Fixes 60777.
This helps to generate cleaner topology and define sharp features for dynamic
topology. Best used on relatively low-poly meshes, it is not needed as much
for high detail areas and has a performance impact.
Differential Revision: https://developer.blender.org/D4189
It is possible that context does have selected_sequences but
it will be set to None. In this case getattr() will return
None, breaking the intended logic.
Depending on area size, the scrollbar covered the bottom of the text,
with the extra it will only cover the padding at worst.
Differential Revision: https://developer.blender.org/D4207
* Add threshold for minimum amount of mouse movement for dragging to
get activated.
* Limit angles at which dragging is considered an action, do nothing if
mouse does not clearly move up/down/left/right.
* Increase action zone size vertically.
Differential Revision: https://developer.blender.org/D4227
There are probably many more cases in which the menu looks a little different.
However, I don't know them all and it's too easy to break something accidentally here.
Maybe a user could try the different combinations of object types and check if there are entries that should not be there.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4240
For now this is not part of copy-on-write, and needs extra animation
evaluation.
Reviewers: sergey, brecht
Maniphest Tasks: T59939
Differential Revision: https://developer.blender.org/D4140
Allows users to select a font for text strips in the video sequence editor.
Related: 3610f1fc43 Sequencer: refactor clipboard copy to no longer increase user count.
Reviewed by: Brecht
Differential Revision: https://developer.blender.org/D3621
The clipboard is not a real user and should not be counted. Only on paste
should the user count increase.
This is part of D3621, and was implemented by Richard Antalik and me.
Cycles shows first the render, and then the viewport settings.
One could argue that EEVEE's main setting is the viewport one.
But that is silly. If we need an extra setting for the lookdev mode so be it.
But EEVEE should be treated as an engine just as Cycles.
Also, removed the " Samples" bit from their labels since they are under
the Sampling panel.
This is more like a band-aid than a real fix actually, real fix would be
to understand why rendering smoke requires auto texspace to be ON
(afaict, this was not the case in 2.7x)...
But I've already spent way too much time on this issue, at least now we
get better situation than before (i.e. smoke with adaptive domain works
well even when orig domain mesh has autospace flag disabled).
Not sure why that was that way (can't remember any good reason at least,
so assuming this is a dummy mistake from own rB33cbcd73448f), this
should be done in any case.
Also includes some minor refactoring: use guard clauses instead of nested conditionals.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D4238
Free the BVH tree immediately along with the mesh, otherwise we might access
invalid mesh data.
Differential Revision: https://developer.blender.org/D4201
After rename is done we need to make sure all copies of
corresponding datablocks are updated in all dependency
graphs: otherwise bone will have a new name, but animation
will still be using an old one.
Index files used by emacs, vim and others, for autocompletion and
searching; generated by etags, universal-ctags and others.
Differential Revision: https://developer.blender.org/D4208
The issue was caused by intermediate DerivedMesh being created with
scene's Simplify settings taken into account. This is what happens
when one area makes implicit decisions based on whether passed Scene
pointer is not NULL.
Made it so ignoring simplification serttings is an explicit flag,
which makes it easier to follow what's going on.
This is actually a workaround for the crash in OpenSubdiv.
Topology refiner will have a crash when special conditions
are met:
- Refiner is configured to use infinitely sharp patches.
- Refinement happens for the level 1 (which we call Quality 1 on
Blender side).
- Mesh has non-quad faces.
The workaround is to force refinement to happen to level 2 (or
quality 2 on Blender side) when those conditions are met.
Later on with the next OpenSubdiv update we can remove this
workaround, since there was work done on OpenSubdiv side to
deal better with such configurations.
The modifier will now be somewhat slower, but this will be
compensated with upcoming topology cache enabled by default.
The workaround is done when initializing settings, so the
comparison of topology refiner settings is happening without
any extra workarounds there.
Add 'G_draw' for all draw manager globals,
avoids adding extern to each file.
Connection between `ts` and `globals_ubo` wasn't obvious,
now called `G_draw.block` & `G_draw.block_ubo`.
While verbose, this is a more flexible way to construct shaders.
Libs & defines can be optionally included for each shader type
which was previously done with inline string creation.
Not only were those often making doublons with already existing
BLI_math's stuff, but they were also used to hide implicit type
conversions...
As usual this adds some more exotic inlined vector functions (one of
the rare cases where I really miss C++ and its templates... ;) ).
Will document the new options in release notes, then in manual.
Still a bit of work to do on the bulging shape that appears
on cube corners if using arc inner miters, but will do that later.
Also need to do something smarter in clamp overlap.
Reported by @pepeland.
Adding missing events on the first point was breaking the guide behaviour.
Also, updated Ckey so it always defaults to Circular mode when guides are off.
Instead of doing manual ray-plane intersection we use normalized positions
of the grid mesh and apply scaling after interpolation so that we keep
good precision even at really far distances.
Precision is now two order of magnitude better and does not produce the
same kind of artifact at lower clip start values.
This commit also cleanup the implementation.
Fixes T58918 Grid not appearing correctly at low clip start in 2.8
Memfile undo isn't compatible with sculpt or edit-mode.
This didn't work in 2.7x, so best disable memfile undo for now in
situations where it's going to loose data or crash.
Before that only normal component was averaged, which is not
really correct.
Unfortunately, the new code is somewhat slower due to more
involved math to deal properly with non-quad faces, but the
plan is to move averaging from runtime to edit time, This
means, that mdisps will always be continuous around the edges
and no averaging on every frame change of animated character
will be needed.
Avoids nasty code all over where such math is required, and
compilers can easily deal with such situation.
Don't prefer questionable micro-optimization which comes with
a cost of nasty actual logic code.
The idea is to run reshaping for every boundary vertex
of a grid rather than trying to copy boundary grid
elements.
While this is somewhat slower, this avoids all this
tangent flipping magic, which tempts to be rather tricky
and fragile.
The boundary copy code was not dealing correct with flipping
tangent vectors, hence causing discontinuity in the final
positions.
Now we only copy boundaries for quads, where we know how to
deal with tangent vectors and where we know that this is
needed.
More clear solution could be to change the code in a way that
handles handles displacement grids of quads in the same way
as it's done for non-quad faces.
Use normal_quad_v3 instead of normal_tri_v3 and compute the mean of all
corner distance during frustum plane extraction.
Fix T58243 Flickering of viewport when rotating and zooming
Having clipping limit selection and tools is confusing when not visible.
Disable on load until it's supported
(doing this via ifdef's isn't practical).
Fixes T59580
While changing RGBA or color wheel didn't add undo steps,
HSV and Hex values did.
Disable undo for these button types since an undo push happens when
exiting the picker.
The problem here was that when a render result is allocated, the standard render passes are added according to the
pass bitfield. Then, when the render engine sets the result, it adds the additional passes which are then merged
into the main render result.
However, when using Save Buffers, the EXR is created before the actual render starts, so it's missing all those
additional passes.
To fix that, we need to query the render engine for a list of additional passes so they can be added before the EXR
is created. Luckily, there already is a function to do that for the node editor.
The same needs to be done when the EXR is loaded back.
Due to how that is implemented though (Render API calls into engine, engine calls back for each pass), if we have
multiple places that call this function there needs to be a way to tell which one the call came from in the pass
registration callback. Therefore, the original caller now provides a callback that is called for each pass.
Some of Eevee's Bloom defaults are not very good for physically based rendering. This patches addresses this issue.
This picture shows one of the problems with current default. Bloom looks very foggy:
{F6280495}
Even worse, light emitters much dimmer than the Sun can make everything equally hazy if Clamp is set to 1.0 and intensity to 0.8 (current default). Artists often forget to adjust Clamp value and do not know what value to use for realistic intensity. Also, currently both Clamp and Intensity do not have good UI ranges. This is why often Eevee renders end up very hazy and bloom often does not look right.
Bloom effect plays important role to help to distinguish between bright and relatively dim light sources. With current defaults this is broken because Clamp set to 1.0. Also, it cannot be disabled if set to 0 like expected. This patch fixes this and sets it to 0 by default. If users need to clamp, they can do so easily with UI range up to 1000. This range is good enough for most cases and provides enough precision to control lower values, and the highest value helps to limit bloom from the Sun if necessary and will leave untouched most other light emitters. If needed, much higher values for Clamp can be entered manually up to 100000. 10000 is still affects the Sun, but up to 100000 highest limit allows to clamp anything that is much brighter than the Sun if user needs to limit bloom in such cases (for example, bright explosion in the sky or anything else very bright).
I propose new default for bloom Intensity - 0.05 and UI range to suggests realistic values. Bloom Intensity > 0.1 is not realistic for clean lens but the user can enter manually much larger values if needed.
For comparison, here is a my own photo with and without bloom caused by the Sun (on second photo the Sun was occluded with an object).
{F6280500}
{F6280492}
In real life bloom is much more subtle and does not look hazy. If Clamp is disabled, then out of 0.1, 0.05 and 0.025 values I have tried, 0.05 looks most similar to the photo. Here is test render with and without bloom with the Sun in similar position like on the photo:
{F6280496}
{F6280494}
Using color probe 27x27 I compared lightness below the horizon under the Sun. In rendered by Eevee images lightness difference was 17. In case of the photos lightness difference in similar place was 11. I then compared leftmost spot (also below the horizon) and lightness difference was approximately 2 between two photos and 1 between rendered images. In other words, with these settings bloom effect is not too strong and is not too weak. Visually it may seem like decreasing bloom intensity may increase photorealism, but then bloom effect would be too localized even for the Sun.
Besides this single test, I tested in many other scenes as well, with and without the Sun, with different HDRIs, and as far as I can tell 0.05 intensity turned out to be good default - it produces bloom strong enough to be noticeable and not too hazy.
In Cycles shutter default value is 0.50, so for consistency set to 0.5 by default in Eevee too. Besides, 0.5 is typical standard for real cameras, and values higher than 0.5 usually are needed only if very strong motion blur is desired.
Here is summary of all changes:
Bloom Intensity: 0.8 > 0.05
Bloom Intensity UI range: 0-10 > 0-0.1
Bloom Clamp: 1.0 > 0.0 (disabled by default)
Bloom Clamp manual range: 0-1000 > 0-100000
Bloom Clamp UI range: 0-10 > 0-1000
Shutter: 1.0 > 0.5
This patch is related to the discussion in this thread, there are more examples of what bloom will look like with 0.05 intensity by me and others:
https://devtalk.blender.org/t/eevee-needs-to-have-physically-based-defaults/4700
Reviewers: fclem
Reviewed By: fclem
Subscribers: pablovazquez, billreynish, rboxman
Tags: #eevee
Differential Revision: https://developer.blender.org/D4212
This is in order to make the API more multithread friendly inside the
draw manager.
GPU_shader_get_uniform will only serve to query the shader interface and
not do any GL call, making it threadsafe.
For now it only print a warning if the uniform was not queried before.
While form a strict consistency point of view it could make sense to
return identity matrix for non-instance items, it can be very handy to
get that info (common to both instances and regular objects) directly in
all cases.
This uses the same command as regular hierarchy delete, and is only
activated when debug value is set to 666 for now.
Here on file from T60419, it gives about 20% speed-up (from 5.5s to 4.4s).
This further changes the preferences organization, to avoid grouping unrelated
settings together. With more sections we can also expand more panels by default,
making it possible to quickly go through sections and see the settings of each.
Panels with less used settings are still collapsed by default, to keep all panel
headers visible without scrolling.
Differential Revision: https://developer.blender.org/D4216
This is still not fully correct, since the event loop is blocked by GHOST
and no timer events are happening for animation while the mouse is still.
But for the most part it looks ok.
When use the Alt mode to draw close strokes, if the color had the stroke diabled, the stroke was not visiblle while drawing.
Now, it's visible while drawing, but it's hidden again when the stoke is finished. To display close strokes, enable stroke mode in material or enter in edit mode.
There was no documentation at all, some very bad practices (like using
G.debug_value > 0 as some sort of global debug print switch), and even
an overlapping use of '1' value...
Also, python setter did not check for valid range (since this is a
short, not an int).
That one was used to allow specifying in system console a new path for
missing libraries, when loading a .blend file.
We now have a much more easy and user-friendly way of repairing missing
linked datablocks/libraries, so this is not needed anymore.
Main idea is to remove IDs to be deleted from Main, to avoid looping on
them to remove other deleted IDs usage (this is the most expensive
process in ID deletion, by far).
Speed improvements when deleting a large amount of IDs from a Main
containing a lot of them is quite significant, some examples for Objects:
* Removing 1k from 10k: 32% quicker (2.5s to 1.7s).
* Removing 10k from 20k: 60% quicker (59s to 23s).
* Removing 20k from 20k: 99.5% quicker (82s to 0.4s)!
Note however that this process is more risky/touchy, since we by-pass
some safety checks from regular ID removal here.
So will only give access to that code from python API for now (in
separate commit), so that it gets really tested. Also still need to
think about how to hook it up in UI (probably mostly for Outliner),
since we often do higher-level operations there...
Seems to be caused by cae3750 which changed free() function used
by bmain free to the one which does dependency graph tag. We do
no want to do any tags here.
This commit makes it so OpenSubdiv's topology refiner is kept
in memory and reused for until topology changes. There are the
following modifications which causes topology refiner to become
invalid:
- Change in a mesh topology (for example, vertices, edges, and
faces connectivity).
- Change in UV islands (adding new islands, merging them and
so on),
- Change in UV smoothing options.
- Change in creases.
- Change in Catmull-Clark / Simple subdivisions.
The following limitations are known:
- CPU evaluator is not yet cached.
- UV islands topology is not checked.
The UV limitation is currently a stopper for making this cache
enabled by default.
This fixes following errors:
- The code didn't work correctly for edges reconstructed by
the OpenSubdiv's topology refiner (due to indexing
difference).
- Sharpness of non-manifold and boundary edges was not
working correctly.
C++11 doesn't need the space between '> >' in a nested templated
declaration, so instead of `std::vector<std::pair<a, b> >` we can now
write `std::vector<std::pair<a, b> >`.
It's now possible to export curves and NURBS as mesh data to Alembic.
This allows artists to do any crazy thing on curves and export the
visual result to Alembic for interoperability with other software (or
caching for later use, etc.). It's an often-requested feature.
This works around T60503 and the fixes export part of T51311.
Note that exporting zero-width curves is currently not supported, as
exporting a faceless mesh (e.g. just edges and vertices) is not
supported by the mesh writer at all.
To test, create a curve with thickness (for example extruded), export to
Alembic and check the 'Curves to Mesh' checkbox in the export options.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4213
I moved most of the `AbcMeshWriter` code to a new class
`AbcGenericMeshWriter`. The latter is an abstract class and does not
make any assumptions about the type of Blender object being written.
This makes it possible to write metaballs, curves, nurbs surfaces, etc.
as mesh to Alembic files.
The `AbcMeshWriter` class now is the concrete implementation of
`AbcGenericMeshWriter` for writing mesh objects.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4213
If the triangulated mesh was in itself a new mesh that should be freed this
should happen before the function returns (as it only returns a single mesh,
and thus the caller can only free one).
Add 'missing' bpy code from BKE_libblock_free_ex(), now both functions
do exactly the same thing, only the later is less flexible (fewer
'exotic' behaviors supported, like handling IDs outside of bmain etc.).
Next step: nuke usages of BKE_libblock_free functions, makes no sense to
have twice the same code around!
Xorg's XIWarpPointer doesn't support multi-head display while
XWarpPointer does.
Revert since this is a known TODO in Xorg and setting a custom
xinput matrix seems not to be used often.
Resolves T50383
Xorg's XIWarpPointer doesn't support multi-head display while
XWarpPointer does.
Revert since this is a known TODO in Xorg and setting a custom
xinput matrix seems not to be used often.
Resolves T50383
The issue is that the edge fix geometry goes on top of the actual drawn
points.
This commit reduce the edge fix size to the strict minimum but does not
get rid of it.
Related to T60139
Before this Blender always needed the Wintab driver. This adds support for the
native pressure API in Windows 8+, making it possible to get pressure sensitivity
on e.g. Microsoft Surface hardware without any extra drivers.
By default Blender will automatically use Wintab if available, and if not use
Windows Ink instead. There is also a new user preference to explicitly specify
which API to use if automatic detection fails.
Fixes T57869: no pressure sensitivity with Surface pen or laptop.
Code by Christopher Peerman with some tweaks by Brecht Van Lommel.
Differential Revision: https://developer.blender.org/D4165
Base outline is 2px wide (because of how we detect them).
And since inflating this outline will only produce outlines that are 2*x
thick we map the UI scalling and the outline width setting to the closest
match.
Do note that thicker outlines have a performance cost since they need more
texture fetches and passes.
This fixes T60252 3D View Outline Width not working
The existing Add and Multiply blending modes have limited usability,
because the appropriate operation for meaningfully combining values
depends on the channel. This adds a new mode that chooses the operation
automatically based on property settings:
- Axis+Angle channels are summed, effectively averaging the
axis, but adding up the angle. Default is forced to 0.
- Quaternion channels use quaternion multiplication:
result = prev * value ^ influence
- Scale-like multiplicative channels use multiplication:
result = prev * (value / default) ^ influence
- Other channels use addition:
result = prev + (value - default) * influence
Inclusion of default in the computation ensures that combining
keyframed default values of properties keeps the default state,
even if the default isn't 0 or 1.
Strips with this mode can be keyframed normally in Tweak mode,
except that for quaternion rotation keyframing always inserts
all 4 channels, and the channel value sliders on the left side
of Graph/Action editors won't insert keys without Auto Key.
Quaternion keys are also automatically normalized.
Differential Revision: https://developer.blender.org/D4190
Supporting a strip blending type that treats quaternions as a unit
also means being able to adjust all sub-channels as a unit when
inserting keyframes. This requires refactoring keyframe insertion
code to retrieve array property values for all channels at once,
before iterating over the indices being inserted.
This disables touch gesture recognition in Blender, avoiding any initial delay
when drawing with grease pencil, texture paint, etc.
Differential Revision: https://developer.blender.org/D4203
Allows users to select a font for text strips in the video sequence editor.
Related: 3610f1fc43 Sequencer: refactor clipboard copy to no longer increase user count.
Reviewed by: Brecht
Differential Revision: https://developer.blender.org/D3621
When editing an action without a strip, or tweaking a strip without
time mapping enabled or supported, the extents of the virtual strip
can't be controlled and are purely derived from keys in the action.
Thus, cutting off evaluation of the action at these arbitrary points
gets in the way of observing the natural extrapolation of the F-Curves
and thus appears to be a mis-feature.
With this change non-mapped actions are evaluated with infinite
range, exactly like they are handled without NLA, unless extend
mode is set to Nothing.
The viewport stereoscopy support helpers are finally ported to 2.80.
We now can scale the camera and the "stereo cameras" will scale
in the viewport as well (unlike 2.7x).
At the moment I disabled the drawing of the camera frame when
stereo is selected and you are looking through the camera.
It is to be fixed later, but for now it draws the border wrong.
In 2.79 this was not a problem because the camera frame was drawn
afterwards as a hack.
Viewport > Stereoscopy:
* Cameras
* Convergence plane
* Convergence plane alpha
* Stereoscopy volume
* Stereoscopy volume alpha
shgroup_instance_alpha was getting a color[4] but would only use the
alpha defined upon creation of the shading group.
This was very limiting since it wouldn't allow for different instances
to have different alpha values.
Patch made with Clément Foucault (he made the code of it, while I fixed
all the parts of the code that were relying on shgroup_instance_alpha.
The original issue is that we were changing the camera shiftx
temporarily for the stereoscopic calculation. However we are using the
evaluated object when calculating the projection matrix.
Note: Camera framing drawing for stereo still seems to be broken.
But the viewport itself is now correct.
Not sure exactly why this happened for 'apply as shape' and not in other
cases (did not took time to fully trace what happens there). But in any
case, `BKE_key_evaluate_object_ex()` can be called from a fair amount of
places, including during depsgraph evaluation, so setting back key's
owner here is plain wrong in CoW era.
More like a band-aid than anything else really, that code is horribly
weak and need to be fully re-written at some point (putting all those
temp data-blocks fully outside of bmain...). But for now should do.
Fix T60194: Sequencer cut loses animation data for the right strip.
Fixing the first also fixes the second. First attempt was delaying
uniquename check at a later step of cut process, after everything had
been duplicated. While this fixed first issue, second one became even
more proeminent (it become active for all strips, and not only
video/audio movie strips in meta's).
So instead, passing along the list of (new) sequences, so that duplicated
seqs can be put there immediately, before checking for unique names,
henceforth ensuring even strips inside meta's get properly handled.
This partially reverts commit bb98e83b99.
It fixed 'strips having same name' issue, but broke handling of
animation then. Need to find a better way to handle this.
This commit groups a set of new tools that were tested in grease pencil object branch before moving to master. We decide to do all the development in a separated branch because it could break master during days or weeks before the new tools were ready to deploy.
The commit includes:
- New Cutter tool to trim strokes and help cleaning up drawings.
- New set of constraints and guides to draw different types of shapes. All the credits for this development goes to Charlie Jolly (@charlie), thanks for your help!
- Segment selection mode to select strokes between intersections.
- New operator to change strokes cap mode.
- New option to display only keyframed frames. This option is very important when fill strokes with color.
- Multiple small fixes and tweaks.
Thanks to @pepeland and @mendio for their ideas, tests, reviews and support.
Note: Still pending the final icons for Cutter in Toolbar and Segment Selection in Topbar. @billreynish could help us here?
This change brings back old original logic which was checking
whether worker threads do fit into an active CPU group. But
it does it a bit smarter now and is also checking affinity
within that group. This way Cycles will use all threads on a
Threadripper2 CPU if it's set to automatic number of threads,
but on another hand will not change affinity if user requested
16 threads and changed Blender affinity.
There was a problem counting the number of points for edit points and lines. Now the total size is used allocating the VBO size and not the stroke size.
This removes code duplication and put an end to the old "create at request"
batch creation.
Also it uses the same vbo as the uv layer used for shading. Reducing VRAM
usage.
Also fixes the modified uv display in uv edit mode.
This is in order to allow more spaces to have their batches created at the
same time and sharing the batches.
This is part of the effort fo making the drawing code more optimized. This
commit however should not introduce any difference.
This commit bypass the aspect ratio correction for angle stretch display
but this should be fixed in the next commit.
This is in order to use batch cache directly without using tricks like
batch presets reseting the VAOs.
Note: For now it also create a depth buffer for this area which is not
needed. We could get rid of this to lower VRAM usage.
The overlay should now use the texture interpolation setting in material
mode.
In image mode, there is now a new button to let the user choose the texture
filter. The option is located in the Texture Slots popover and only shows
in Image mode.
Debugging the edit mode selection I realized the vertices are often
occluded by edges with the same depth. Sometime it can be the center
pixel of a vertex point and that can lead to some selection issue.
So I increased the offset a bit for the vertices and it seems to fix it.
This makes it more future proof and remove baked id offset inside the vbos.
Instead we add the offset as a uniform. This makes it possible to reuse
the vbos instead of discarding them all the time.
Also using batch request may reduce batches creation time.
Tweaked scheduling so it survives this situation by scattering
"extra" threads uniformly over all the NUMA nodes.
There are still tweaks possible to make some specific hardware
configurations work better.
TH_BACK was being used when drawing the 3D view even though
there was no way to set the color in the preferences.
The color was zero'd when moving to the new 2.8x theme.
Having both gradient and background colors was confusing,
especially having to use 'TH_HIGH_GRAD' for the 3D view, 'TH_BACK' for
other views.
Move the background color back to 'TH_BACK', 'TH_BACK_GRAD' is used
when gradients are enabled.
RNA is unchanged so presets don't need updating.
This was only used by collada export metadata.
If metadata like this is needed, we can use per-filetype preferences,
to make it clear when user identifiable information is being used.
We should *never* prevent copying basic mesh CDLayers (vertices etc.),
that does not make sense.
I guess issue was not in old DM because geometry was duplicated anyway,
and in 'normal' modifier stack eval, probably because bare mesh was
awlays requested? But we should not have to be explicit here about it.
This is a part of another which, which needs to tag owner
of f-curve for an update. But since this is too many lines
changed committing separately.
Basically, for f-curve AnimElement we are now storing ID
whih owns the f-curve.
When holding down the key for a while, the pie menu will disappear when
releasing the key. This is under the assumption that in this case the user
decided to cancel the action.
Differential Revision: https://developer.blender.org/D4180
There was a problem in the caps start and in some situations, the pixel was wrong.
Also changed the way the caps are detected because use the alpha in negative was a hack that maybe will not work with all drivers.
Exiting modes shouldn't be needed since loading the new memfile
will free the old data.
Sculpt mode dynamic topology was adding undo data on exiting the mode
which isn't logical in this case and can be avoided altogether.
Too many things done wrong in original rBd12b3767f81d to list them all
here, hopefully nothing bad sneaked in again this time :|
Also cleaned up a little the 'sort by name', even though (since we only
have two options by default, sort by index and by name) we can abuse it
as a binary option for now, this is not a bitflag...
That was indeed not working properly, not at all. Except for
the basic case, but as soon as you used another object to define the
mirror plane, it would be utterly broken, in several different ways!
Exiting modes shouldn't be needed since loading the new memfile
will free the old data.
Sculpt mode dynamic topology was adding undo data on exiting the mode
which isn't logical in this case and can be avoided altogether.
- When toggling a mode that doesn't support multi editing
only do this once of the active object.
- For sculpt mode create sculpt data since this is needed
for activating other sculpt objects on reload.
This reverts commit 4d8ed937f2.
An alternative fix will come soon as a patch, since this introduced an issue.
Rolling back since the original fix (sculpt cursor on load) is less important
than the issue it introduced (crash on weight paint undo/redo).
Fix T60322.
We now perform COW -> original data flushing for all the debug values + error
status flags on Drivers/DriverVariables/DriverTargets, as these are only set
when errors are encountered when evaluating drivers.
When BPY_python_end() is not called, there can be buffered data still in
`sys.stdout` or `sys.stderr`. This generally isn't an issue when those are
connected to a terminal, but when they are read by another process (in the case
of rendering with Flamenco, for example) we could miss the actual error message
that's causing the exit in the first place.
The following script demonstrates the issue; before this commit neither the
writes to STDERR and STDOUT nor the traceback of the NameError were shown.
#!/bin/bash
cat > file-with-errors.py <<EOT
import sys
print('THIS IS STDERR', file=sys.stderr)
print('THIS IS STDOUT', file=sys.stdout)
nonexisting.monkey = 3
EOT
blender --enable-autoexec -noaudio --background \
any-existing-blendfile.blend \
--python-exit-code 42 \
--python file-with-errors.py 2>&1 | cat
Reviewers: campbellbarton, mont29
Reviewed By: campbellbarton, mont29
Subscribers: fsiddi
Differential Revision: https://developer.blender.org/D4168
This switches evaluation of vertices which are on the boundaries
of PTex faces to a single threaded one. While this introduces
some slowdown it fixes ambiguity of PTex index used to evaluate
particular vertex.
Possible alternative solutions would be:
- Do some pre-calculation of index, then do evaluation in threads.
- Try using Gregory patches and see if that makes any affect.
Fix T60235: Flickering of object instances
The units scaling was inappropriate when the bevel value was
to be interpreted as a percent, so added a separate rna property
for "Width Percent" and made UI show the width appropriate for
current offset_type.
Harden normals causes normal splitting, which will not give the
appearance expected due to autosmooth unless some edges are sharpened,
so this change fixes that. Also bevel tool will turn on autosmooth
if not already on if hardening normals.
Makes the entire Preferences UI nicely width responsive. Also, move
use_tabs_as_spaces option back to file path options, it was too lonely
in its own panel ;)
* Expand more sub-panels by default.
* Move release confirms and numeric input settings to Input.
* Move 3D cursor settings to Editing.
* Move region overlap to Interface.
- Move author to save&load
(was incorrectly under text editor).
- Rename Memory -> Memory/Limits
(some of the settings aren't obviously todo with memory).
The Collada exporter suppresses the export of flat animation curves
to optimize the animation (in fact to make the exported file smaller).
But sometimes it is important to also have the flat curves exported
because they may be needed to define an initial transformation to
a fixed location - like translating the weapon from the ground floor
to the back of the model in the report.
I added a new option "all keyed curves" which is disabled by default
but when enabled it also exports flat curves.
feedback is very welcome
In a77b63c569, the Preferences navigation region background was
made brighter. Recently stored userpref.blends (since b00963afc1,
so beta release included) would still use the slightly darker
background for the Preferences navigation region.
Now the version patch added for a77b63c569 also sets the new color
for those recent configs.
The problem was that removing entries from a vetor while iterating
the vector was implemented badly. This caused the failure when only
one element was in vector.
Partially revert rB1b8c3774a86ebc04fceb9cd, there is no good reason to
make object.dimensions read-only, it works perfectly well from python
API! Only breaking case was that weird multi-editing UI feature, due to
how it sets things. But RNA setter itself works fine, and it's a handy
shortcut/helper for scripts.
Also when breaking API, it is good practivce to at least check official
add-ons...
Implementing a new intelligent mixing mode that combines quaternions
via multiplication requires rewriting the NLA code to recombine array
properties from separate scalar channels during evaluation.
In addition, stable evaluation of NLA stack requires that any channel
that is touched by any of the actions in the stack should always be
set to a definite value by evaluation, even if no strip affects it
at this point of the timeline. The obvious choice for the fallback
is the default value of the property.
To make scanning all actions reasonably efficient, mapping paths to
channels should be done using hash tables.
Differential Revision: https://developer.blender.org/D4120
(Part 1 was 00963afc14978b)
Does the following changes visible to users:
* Use panels and sub-panels for more structured & logical grouping
* Re-organized options more logically than before (see images in D4148)
* Use flow layout (single column by default).
* New layout uses horizontal margin if there's enough space.
* Change size of Preferences window to suit new layout.
* Move keymap related options from "Input" into own section.
* Own, left-bottom aligned region for Save Preferences button.
* Adjustments of names, tooltips & icons.
* Move buttons from header into the main region (except editor switch).
* Hide Preferences header when opened in temporary window.
* Use full area width for header.
* Don't use slider but regular number widget for UI scale.
* Gray out animation player path option if player isn't "Custom"
Internal changes:
* Rearrange RNA properties to match changed UI structure.
* Introduces new "EXECUTE" region type, see reasoning in D3982.
* Changes to panel layout and AZone code for dynamic panel region.
* Bumps subversion and does versioning for new regions.
RNA changes are documented in the release notes:
https://wiki.blender.org/wiki/Reference/Release_Notes/2.80/Python_API/Preferences_API
Design & implementation mostly done by @billreynish and myself.
I recommend checking out the screenshots posted by William:
https://developer.blender.org/D4148#93787
Reviewed By: brecht
Maniphest Tasks: T54115
Differential Revision: https://developer.blender.org/D4148
When fill a stroke if the fill layer hasn't keyframe, the fill is wrong because previous fill shape hide area to fill.
Now, if the keyframe is missing in the active layer for the current frame, a new frame is added.
- clips/masks were not showing an icon [both dont have a dedicated icon,
took the ones used elsewhere]
- masks hit an assert in outliner_add_element()
- missing outliner update when adding a mask
spotted while looking into T59939
Reviewers: mont29, brecht
Differential Revision: https://developer.blender.org/D4142
It was intended to be a quarter-circle, however it was oriented wrong.
Since the triangle is no longer visible and does not overlap with the
button anymore, this just makes it a square.
Differential Revision: https://developer.blender.org/D4139
Needed to port operator to use evaluated particle system.
But also changed interface to always show Convert button when
draw type is set to Path (Hair particle system is forced to
be draws as path). This avoid rather expensive lookup on every
redraw, but will show Convert button for un-baked particle
emitter.
Probably, an acceptable compromise.
Fixed by setting the limit to the original limit I used for Cycles.
Rendering still goes extremely slow when bokeh is lower than 1.0.
But at least now it is "waitable". With lower numbers than 0.01 I don't
think we would ever get a render to finish.
@fclem feel free to address the real root of the problem, but I'm afraid
it may be a limitation of the algorithm you are using.
Even though the fragment shader was already discarding all members of
dof_bokeh_sides when blades was zero, the C code was still trying to
use this for a few divisions leading to runtime asserts.
Those are harmless yet can lead some to waste time while pursuiting
other bugs (namely a near freeze when blades aspect ratio is too low).
The original issue is that wm->paintcursors is empty until we go in and
out of the sculpt mode. To fix this we need to toggle inside the sculpt
mode.
This is usually tackled by ED_editors_init(), however the sculpt mode
toggling was never call because the object technically had "mode data".
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D4153
The intention was to flip normals when extruding in the opposite
direction, however the sign of the angle isn't meaningful unless
the geometry center and region normal are taken into account.
Disable, may add back in a way that works more predictably.
The intention was to flip normals when extruding in the opposite
direction, however the sign of the angle isn't meaningful unless
the geometry center and region normal are taken into account.
Disable, may add back in a way that works more predictably.
Word wrap and alignment layout args only used by UI_fontstyle_draw
were vars in uiFontStyle.
These were written to before drawing, so better pass as an argument.
Pass uiFontStyle & uiWidgetColors as const args.
Move the bevel hardening code all into bmesh_bevel.c.
Based on user feedback, rewrote the bevel hardening algorithm
to be more what users want.
Based on user feedback, changed the UI, removing some
not-useful options. Now hardening normals while beveling
is enabled by a simple checkbox.
Now setting face strength gives options for which faces
get their face strength set.
Am not totally convinced that generating meshes without fully valid
material info is a good thing, but this seems to be rather common in our
code base (in both mesh editing and convert-to-mesh cases).
So for now, duplicated code in mesh eval finalization to main displist
creation/eval function, synchronizing mat data at the end of modifiers
stack eval, if needed.
We are core profile now, no need to link against GLU.
This change makes it so Blender binary is not dependent on liGLU.so.
That was a weird thing that Blender was dependent on it, but was not
using any functions from it.
Caused by rB36ca072375deea4803df4681716c1d3224095e07
[one instance of `DEG_get_original_object` was neccesary, the other one
breaks getting the parent in `BLI_ghash_lookup`]
Reviewed by: brecht
Differential Revision: https://developer.blender.org/D4154
As the z-depth is calculated using the internal drawing, if we use the front mode the z-depth is wrong. The Front or Back mode must be used only for display, but not for calculation.
- Remove pathlib use
(was converting to/from string with no real advantage).
- Use user_resource(..., create=True) to ensure the path exists.
- Pass full path to BKE_studiolight_create, don't add extension after.
- Fix 'sl' filtering glob and move from ui code to operator.
- Fix string copy length.
Mixing file rename with other changes should be avoided.
Using 'module_py_api' convention here
is in keeping with imbuf, idprop, blf & bmesh.
No reason for gpu to have a different convention.
Instead of crashing, an error message is displayed if a function of the gpu module is called without a GPU context.
Reviewers: brecht, campbellbarton, JacquesLucke, mont29
Subscribers: abdelmatinboulbayam, amir.shehata
Differential Revision: https://developer.blender.org/D4143
Note this is also broken in 2.7x.
This is not a big deal since the operator is exposed in the correct
menus. But some users were accessing it via the search menu which would
lead to issues.
The original comment in the file was not acknoledging pose bones could be tacked
here as well (my fault since I should not have trusted the comments and read the code
intead).
Problem introduced on aeb8e81f27.
- Fixes missing check for unified brush in sculpt mode.
- Re-orders material first in gpencil paint mode
(matching color-first for other paint modes).
- Avoid minor differences (missing tablet pressure options from topbar).
- Don't repeat properties already displayed in the topbar
when opening the brush popover.
* 2D Animation: lots of changes from the grease pencil team. Properties
editor layouts, brush and material settings, and more.
* 3D Viewport: wireframes set to 1.0.
* World: use nodes by default.
* Node Editor: use narrow toolbar.
Add city, courtyard and interior HDRs. Replace grass field and night
HDR with different images.
Command used for compression:
oiiotool %s --resize 1024x512 --ch R,G,B -d float --compression dwab:300 -o output/%s
This operator allows to create a new stroke joining several selected points of different strokes.
The new stroke will use the current material.
To use, first select the points to be merged. Optionally can remove the old points and strokes.
The operator is available in Edit mode in the Specials menu and Stroke menu.
Steps to reproduce were:
* Open Preferences
* Choose "Input" category
* Scroll to the bottom
* Choose "Interface" category
The newly activated category should now use the scrolling set previously
in the other category, causing the contents to be out of view. You
would have to scroll to bring it back.
Now scrolling is stored per category.
The old code assumed that if the number of curves was the same, the
entire set of curves would have the same topology (in other words, it
assumed 'same number of curves => same number of vertices for each
curve').
I've added a more thorough check that also considers the number of
vertices in each curve. This still keeps certain assumptions in place
(for example that if the topology is the same, the weights won't change,
which is not necessarily true). However, when the assumption doesn't
hold, at least now Blender doesn't crash any more.
This is similar to what physics baking is doing: invoking the operator
runs a background job, whereas executing blocks. This makes Python
scripts calling the Alembic import/export operators more predictable.
For backwardward compatibility with existing Python code the
`as_background_job` parameter still exists, which overrides the
behaviour chosen by INVOKE/EXECUTE.
Reviewers: brecht
Reviewed by: brecht
Differential revision: https://developer.blender.org/D4137/new/
The clipboard is not a real user and should not be counted. Only on paste
should the user count increase.
This is part of D3621, and was implemented by Richard Antalik and me.
- own error in rB2c196de56bbb163048b08f321983234a5e72e804
- now introduce RE_PASSNAME_DEPRECATED placeholder for old passes
- also dont allocate NodeImageLayers for these
Reviewers: brecht
Maniphest Tasks: T59922
Differential Revision: https://developer.blender.org/D4132
The GP_STROKE_RECALC_CACHE identifier was changed to GP_STROKE_RECALC_GEOMETRY because the previous name was confusing and could be confused with the recalculation of the Draw Manager cache.
Please always build tests when messing with build system/libs, am tired
of fixing that kind of issues...
Also, that fix is probably not working for standalone, no idea where's
the numaapi lib then, but committing since I need a building blender
here (with the tests, yes).
Please always build tests when messing with build system/libs, am tired
of fixing that kind of issues...
Also, that fix is probably not working for standalone, no idea where's
the numaapi lib then, but committing since I need a building blender
here (with the tests, yes).
The issue was introduced by a Threadripper2 commit back in
ce927e15e0. This boils down to threads inheriting affinity
from the parent thread. It is a question how this slipped
through the review (we definitely run benchmark round).
Quick fix could have been to always set CPU group affinity
in Cycles, and it would work for Windows. On other platforms
we did not have CPU groups API finished.
Ended up making Cycles aware of NUMA topology, so now we
bound threads to a specific NUMA node. This required adding
an external dependency to Cycles, but made some code there
shorter.
Would free evaluated mesh even when it was the one cached in runtime
data by depsgraph evaluation!
Also fixes the asserts about using non-eval object in some cases.
Previously we would try to guess what the main tablet device is, but this is
error prone. Now we keep a list of X11 devices and try to match events to
them. On the Blender side there are still some limitations in regards to using
multiple devices at the same time, but this should improve things already.
Fixes T59645.
Previously we would try to guess what the main tablet device is, but this is
error prone. Now we keep a list of X11 devices and try to match events to
them. On the Blender side there are still some limitations in regards to using
multiple devices at the same time, but this should improve things already.
Fixes T59645.
Previously we would try to guess what the main tablet device is, but this is
error prone. Now we keep a list of X11 devices and try to match events to
them. On the Blender side there are still some limitations in regards to using
multiple devices at the same time, but this should improve things already.
Fixes T59645.
Curve modifier eval code was actually doing nothing to ensure we passed
mesh with valid normals when required by the modifier.
This is a bit basic, rough code, but think it should cover all cases,
time will say...
The issue was caused by NaN valid of the average spring length being
stored in the file. This caused accumulation in the springs builder
to also deliver NaNs, which then caused solver itself to not do
anything.
Not sure why these values where never initialized prior to the
accumulation. Or even, why this runime data is stored in a DNA.
Some sanitizing is possible here, but needs to be done with care
to not disrupt Spring production.
In fact, we can get valid depsgraph 99% of time from current context.
Still added extra optionnal depsgraph parameter just in case (and also
for future, when we might be handling much more temp depsgraphs).
Now, things are becoming REALLY confusing. The script does build
pugi, but is never telling OIIO to use an external one. Which makes
it to use a bundled one.
Trying to link OSL to a different version of pugi causes a lot of
linking errors.
Interestingly enough, that was me who made OSL to use external pugi
to solve configuration problem. But now i can not reproduce that
anymore.
Ideally we would either link everything against our pugi, or not
compile it at all.
ARegion.sizex/y should never have DPI factor applied. For regular panel
regions, DPI will be applied in region_rect_recursive already, causing
it to be applied twice when region size is set dynamically (= based on
content dimensions).
The root of the problem is that KnifeTool_OpData->colors was not init in
some cases. But the reason is unknown as it seems to be random and the
init function was always called.
So instead on init the color only once, we query the colors each time
we draw the knife points.
The overhead of this approach is negligeable.
Allow more flexible use of drivers on B-Bone properties by
connecting the dependencies to the actual operation node that
uses the values, instead of the whole component.
Modifiers
Since user menu entries from SPACE_BUTS/SPACE_TOPBAR are also shown in
other Editors (SPACE_VIEW3D), also allow these entries to be removed
from Quick Favorites from these Editors.
Match and deduplicate logic from screen_user_menu_draw() and
ui_popup_context_menu_for_button().
Reviewers: campbellbarton, brecht
Maniphest Tasks: T58327
Differential Revision: https://developer.blender.org/D4112
In blender 2.8, when you zoom in, the adaptive subdivisions appear earlier than previous versions.
The grid still appears a little before the snap, but since it is very small I see no advantage in snap for this case.
Object visibility is now handled by the depsgraph iterator, but this API
was incomplete as it made no distinction for visibility of the object itself,
particles and generated instances.
The depsgraph iterator API now includes information about which part of the
object is visible, and this is used by Cycles to replace the old custom logic.
Cycles and EEVEE visibility should now be consistent, which unfortunately does
means some subtle compatibility breakage for both.
Fixes T58956, T58202, T59284.
Differential Revision: https://developer.blender.org/D4109
Was only visible after going in and out (with some strokes inbetween)
hair edit mode. The edit structure was never freed during Blender
session for some reason. Now we free those when leaving particle
edit mode.
The issue was caused by shape keys datablock from evaluated mesh
being added to the main database.
This commit makes it so shape keys are not copied for the mesh
used as cage.
We were never removing the parent collection from a collection upon removal
of the parent.
Reviewers: mont29
Differential Revision: https://developer.blender.org/D4099
Has some advantages over existing options.
- Using material links color to rendering with no way to vary colors
if objects share a material.
- Random gives no control, objects may randomly have the same color,
duplicating an object often changes it's color.
The '+' widget to show a hidden region came too close to overlapping
the viewport navigation gizmo and text editor text.
Reduce size and use an arrow icon.
D4110 by @gnastacast
Now, the internal data is recalculated when add or remove a point.
The change in the API affect to stroke.points.add() that now requires a datablock parameter. This parameter is required to identify the datablock affected.
For example: stroke.points.add(gpencil, 1) instead of stroke.points.add(1)
This is the second try to fix T59600
The key indices were wrong: need to offset curve key index
by first curve key index. Also corrected calculation of the
interpolation step.
Annoyingly, can not reproduce this on a simple file, need
production rig. For the possible future look the following
file from Spring was used: 03_005_A.lighting.debug.blend
For strokes:
myframe.strokes.update(mystroke)
For datablock:
gpencil = bpy.data.grease_pencil['gpencil']
gpencil.update()
Still need a manual refresh of viewport.
Also remove special case when no items are selected,
since this only has one or two menu items, one being the add menu
which can be better accessed from the header or add shortcut.
If the no-selection case is to have it's own alternate menu - it should
be more complete before enabling.
Fix for T59219 was using low alpha-light grey for text background
so editing text would always be slightly brighter then the existing
background.
This causes outliner rename to have low alpha making text overlap
icons.
Use solid color to avoid issues with overlapping UI
elements in the future.
The perspective effect deformed the stroke. Now when you are in camera view and the lock axis is not enabled, the stroke is reprojected flat over the view to remove any deformation.
Also fixed reproject operator to use the origin set in topbar and not cursor 3D always.
Another case where editstr from search button would be used, when we
actually have desired pointer itself already available in button.
Am growing tired of doing bandaids fixes on that search menu stuff,
whole thing would require some real re-coding imho, to get rid of that
tantacular dependency over string 'identifier' only (when we should also
have access to at the very least, the active index, and also probably
active data pointer itself...).
And/or clearly separate string identifier from 'UI' string shown to user.
- Silence harmless error print about relation.
Object with particle system which doesn't use physics will
not have point cache component.
- Tag relations for update when particle system physics type
change.
This ensures correct state of point cache component.
This is all part of T59258.
There is no point having operations that iterate over the whole
bit array as macros, so convert BLI_BITMAP_SET_ALL to a function.
Also, add more utilities for copying and manipulating masks.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4101
The original issue was that different platforms will use different
hash lengths, just because defaults on Git client were different.
Now we use explicit length for the hash, and length is the same as
is used for short hashes in Linux -- apparently they started to have
collisions with length of 11.
Instead of link toggle with enum, use a single popover that contains
both settings. The code for this isn't nice - needing 3x panels for now.
See D4075
rBa520e7c85c83 defined T_OVERRIDE_CENTER(1 << 25)
which was already in use T_PROP_PROJECTED(1 << 25)
thus skipping center calculation
Fixes T58882, T59518
Reviewers: campbellbarton, brecht
Maniphest Tasks: T58882, T59518
Differential Revision: https://developer.blender.org/D4100
NLA requires a usable default value for all properties that
are to be animated via it, without any exceptions. This is
the real cause of T36496: using the default of 0 for a scale
related custom property obviously doesn't work.
Thus, to really fix this it is necessary to support configurable
default values for custom properties, which are very frequently
used in rigs for auxiliary settings. For common use it is enough
to support this for scalar float and integer properties.
The default can be set via the custom property configuration
popup, or a right click menu option. In addition, to help in
updating old rigs, an operator that saves current values as
defaults for all object and bone properties is added.
Reviewers: campbellbarton, brecht
Differential Revision: https://developer.blender.org/D4084
This commit makes it so curve path parent solving accepts an explicit
arguments for both time and curve speed flag, making it so we don't
have to mock around with scene's frame.
One unfortunate issue still is that if the instancing object is used
for something else, we might be running into a threading conflict.
Possible solution would be to create a temp copy of an object, but
then it will be an issue of preventing drivers from modifying other
datablocks.
At least the original issue is fixed now, and things behave same as
in older Blender version. Additionally, the global variable which
was defining curve speed flag behavior is gone now!
This aims to resolve a conflict where some users want to keep keyboard
axis setting global, even when the orientation is set to something else.
Move/rotate/scale can optionally each have a separate orientation.
Some UI changes will be made next.
- Use the user orientation when pressing XYZ keys,
second press switches to global.
- Pressing again switches to global, or local
if you're have global orientation set.
The option for gizmos to have their own orientations will be added,
see: D4075
Show backface culling option even with rendered shading since it doesn't
yet support meshes two-sided option (noted as TODO).
Also correct bad string comparison.
This has been unbelievably painful to understand... And solution is only
partially good actually, we may even want a single axis for all the
islands in that case? But for now this is giving much better results
already, compared to the random crazyness it used to produce.
this was working for object/collection display/render but lattice was
not taken into account for non object/collection display/render types
(halo, axis, cross, circle, ...)
Reviewers: sergey, brecht
Maniphest Tasks: T59484
Differential Revision: https://developer.blender.org/D4096
Display statistics from CCG structure.
This makes values to be different from what is shown in object
mode, since CCG is operating on individual grids, and object
mode will stitch those grids. But on another, those values from
CCG is what sculpt mode is actually "sees" or "uses".
The number of faces should be the same in both sculpt and object
modes.
While the operator needs a depth to work as intended,
it feels buggy if the initial drag does nothing until a depth is found.
If the cursor isn't over any geometry calculate an initial depth.
This resolves this issue where users would enable a snapping mode
besides incremental (vertex for eg), then notice strange behavior w/
rotate and scale.
While this ability can be useful, it's quite an obscure use case.
Now changing snap-modes keeps rotate and scale using incremental snap,
with the option for these modes to be affected by other snapping modes.
D4022 by @kioku w/ own minor edits.
- This extends context menus, checking the selection in some cases
to conditionally show operators.
- When nothing is selected, add, paste .. etc are added to the menu.
- Use columns when mixed mesh modes are used (vert/edge/face).
- Move armature naming operators into sub-menu.
See D4043
This now only upload data per loops to the GPU, making use of index buffer
to draw polygon. This make use of the vertex cache, speed up renders
and saves a lot of vram.
Update performance is also slightly faster and can even be improved further
by updating only uvs or vcol independently.
This commits breaks texture paint batches. It will be added back in another
commit.
This is a second attempt to get the crash fixed. The original fix
worked, but it was reverted by d3e0d7f082.
Now the logic goes as:
- All pointers which we can not have shared (the ones which are
owned by the runtime) are cleared.
- The rest of runtime stays untouched.
This seems to be enough to keep particles happy.
This changes the bent normal effect to be a bit more subtle.
I also tuned down the bent normal blending factor so mesh faceted look may
appear more in occluded regions. this is to increase the fidelity of the
indirect lighting. This blending might be a parameter in the future.
Based the calculation on "Bent Normals and Cones in Screen-space"
by O. Klehm, T. Ritschel, E. Eisemann, H.-P. Seidel
This commit adds support for new curve tool and adds more functionalities to the existing primitives, including new handles, editing, stroke thickness curve, noise, preview of the real stroke, etc.
Thanks to @charlie for his great contribution to this improvement.
This replaces the test of consistency and capacity made with `GL_PROXY_TEXTURE_..` on AMD GPUs with one that checks only if the texture fits the limits of size and layer.
Differential Revision: https://developer.blender.org/D4081
Pinning doesn't really change the geometry, so we don't really have a
tag for this. Using '0' for now (same we use for Mark Seam).
I will check with Sergey Sharybin which flag to use instead.
We could use ID_RECALC_GEOMETRY to match the notifier, but even the
simpler ID_RECALC_SELECT would do it.
NLA strips support using the keyframe values in a variety of ways:
adding, subtracting, multiplying, linearly mixing with the result
of strips located below in the stack. This is intended for layering
tweaks on top of a base animation.
However, when inserting keyframes into such strips, it simply inserts
the final value of the property, irrespective of these settings. This
in fact makes the feature nearly useless.
To fix this it is necessary to evaluate the NLA stack below the
edited strip and correctly compute the raw key that would produce
the intended final value, according to the mode and influence.
Differential Revision: https://developer.blender.org/D3927
Basically uiRNACollectionSearch->but_changed was always NULL for the
search templates. So skip_filter would always be true, and we would
never filter.
An alternative fix would be to add the following to the begin of
`template_search_add_button_searchmenu`:
```
static bool always_true = true;
template_search->search_data.but_changed = &always_true;
```
This changes a bit the batches data structure. Instead of using one
vbo per material we use one for all material and use index buffers for
selecting the correct triangles.
This is less optimized than before but has potential to become more
optimized by merging the wireframe data vbo into the shading one.
Also the index buffers are not strictly necessary and could be just
ranges inside the buffer. But this needs more adding things inside
GPUIndexBuf.
Shaded triangles are not yet implemented (request from gpumaterials).
This also changes the mechanism to draw curve normals to make it not
dependant on normal size display. This way different viewport can
reuse the same batch.
The issue was caused by a special code in node tree freeing function
which will free extra fields in the case when tree is not in bmain.
This is how old code was dealing with "nested" trees, but is now
making behavior different from other datablocks. This is exactly
what was confusing copy-on-write logic.
Ideally, ntreeFreeTree() need to behave same as all other datablocks,
ad freeing of data of nested trees should be up to the owner of the
tree (this way it's all explicit and does not depend on check of
some special flag.
Prefer legacy OpenGL library, for the compatibility and portability
reasons.
Also use proper OpenGL libraries to be linked against, so we can
change preference to GLVND.
this can be different though (e.g. vertex parenting) and correct location
is already stored in ob->orig
spotted while looking into T59332
Reviewers: fclem, brecht
Differential Revision: https://developer.blender.org/D4076
This restores the object->data to a non-modifier evaluated state.
So this allow us to change evaluated object modifier stack directly and
get BKE_mesh_new_from_object() for the evalauted object.
By default the text button background color was a similar brightness
to the cursor, making it hard to see at times.
Button types number/slider/text background brightness when editing
varied quite a lot too.
- Change the background while editing to match the number button.
- Darken the selection for greater contrast.
Resolves T59219
Hex color values are now always in sRGB space, as would be expected by
most other applications. Previously they were in display space and using
the view transform.
In 2d655d3 the color picker was changed to use display space HSV values.
This works ok for a simple sRGB EOTF, but fails with view transforms like
Filmic where display space V 1.0 maps to RGB 16.292.
Instead we now use the color_picking role from the OCIO config when
converting from RGB to HSV in the color picker. This role is set to sRGB
in the default OCIO config.
This color space fits the following requirements:
* It is approximately perceptually linear, so that the HSV numbers and
the HSV cube/circle have an intuitive distribution.
* It has the same gamut as the scene linear color space.
* Color picking values 0..1 map to scene linear values in the 0..1 range,
so that picked albedo values are energy conserving.
In production files that use a lot of linking I measured loading speedups between 5% and 18%. In files that use less linking the speedup might not be noticeable at all, but it should not be slower.
Reviewer: brecht
Differential Revision: https://developer.blender.org/D4038
The new data structure uses open addressing instead of chaining to resolve collisions in the hash table.
This new structure was never slower than the old implementation in my tests. Code that first inserts all edges and then iterates through all edges (e.g. to remove duplicates) benefits the most, because the `EdgeHashIterator` becomes a simple for loop over a continuous array.
Reviewer: campbellbarton
Differential Revision: D4050
Mostly rewrite logic which now matches (de)select picking,
share between both operators.
- Support all selection operations (eSelectOp), fixes T59255.
- Add function that selects using 'BONESEL_*' flags & eSelectOp.
This avoids lasso & box select having to handle selection flushing.
- Fix strange behavior with lasso where selecting a bone in a chain
would only select the tip (from 2.7x).
There are some minor differences in center calculation but I don't think
they're important, if so we can update the code.
- Edit-bone uses head (instead of the middle when both are selected).
- Edit-bone flag for restricting components uses the selection instead
of the active bone.
It should be possible to use this when the active element is unselected, but
there still needs to be something else selected. Otherwise it is not possible
to deselect all as a way to get the gizmo out of the way.
- Key-map items properties now override tool-options
so modifier keys can have different behavior to the default action.
- Box & circle select now have `wait_for_input` properties
instead of detecting this based on selection options being set or not.
This relied on the key-map setting properties which may need to be
initialize from the tool settings.
Basically what we address here is to make sure the active object and the cage
are not interferring with the baking result (e.g., when baking Combined).
To do so, we take advantage of the fact that we create our own depsgraph
for baking. So now we can change the cowed objects, instead of the
original ones.
Note: There is still a way to get a crash. If you try to bake from
selected to active when is_cage, but with no cage object, we get an
assert:
```
BLI_assert failed: //source/blender/blenkernel/intern/DerivedMesh.c
mesh_calc_modifiers(), at
(((Mesh *)ob->data)->id.tag & LIB_TAG_COPIED_ON_WRITE_EVAL_RESULT) == 0
```
We can bypass this by passing ob_low instead of ob_low_eval to
bake_mesh_new_from_object on object_bake_api.c:847 . But then the edge
split modifier change will take no effect.
The alignment makes it so the button edges overlap, now one pixel is removed
to account for this.
Differential Revision: https://developer.blender.org/D4063
This enables ffmpeg to encode each frame in its own thread. However in most
cases Blender does not pass frames to ffmpeg fast enough to actually use the
more than two threads. In some tests the speed was measured to be about 20%.
If other parts of the video sequencer get optimized, this should improve.
Differential Revision: https://developer.blender.org/D4031
Applying the effect of bone parent is much more complicated than
simple matrix multiplication because of the various flags like
Inherit Scale. Thus it is reasonable to provide access to this
math from Python for complicated rest pose related manipulations.
The simple case of this is handled by Object.convert_space, so
the new method is only needed for complex tasks.
Differential Revision: https://developer.blender.org/D4053
There are some changes in API of OpenImageIO, but those are quite
simple to keep working with older and newer library versions.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4064
It used to be used for some sort of ignoring automatically
generated bump nodes. But nowadays it causes one of the shaders
in Classroom demo file to be compiled wrong.
as opposed to the 'real' Dopesheet e.g. keyframes were not merged when
placed on the same frame
Reviewers: brecht, aligorith, angavrilov
Maniphest Tasks: T59005
Differential Revision: https://developer.blender.org/D4061
Also this display is optimized. It does not use blending and pixel discard.
Working with scanned data should be more pleasant with this.
A better option would be to use gl_FragDepth to have a better sense of
volume but this discards early depth test.
This makes it possible for engines to ask for batches and only fill their
data after all engine populate functions have run.
This means that, when creating the batches data we already know all the
batches that are needed for this redraw and the needed data.
This allows for less redundant data preparation and better attrib masking.
Ideally, we should run all viewports populate function before executing
the batch construction but this is not the scope of this patch.
Conversion from the old request method will be progressive and both can
coexist (see uses of mesh_create_pos_and_nor()).
This is a small change. We delay all gl calls at the first use of the
GPUIndexBuf / GPUVertBuf in order to be able to create multiple buffers
from different threads without having many gl contexts.
Basically, armature update is not supposed to be run in edit mode.
This worked in master and new dependency graph because nobody was
tagging armature for an update.
But with all those copy-on-write and other things we can't ensure
tag doesn't happen (and we shouldn't). So now we ensure unwanted
code is not run from the code itself.
P.S. Deeper reason of this goes to the optimization of not updating
pose channels when in edit mode. Since pose doesn't define anything
there we don't want to be bothered with a pose update after every
operation which changes it.
Textures are now hooked up to the RESET operation of particle
settings, which ensures particles being re-distributed when
texture is changed.
This is limited to a direct user modifications, which matches
old behavior in 2.79.
Seriously... There is no point in having those subversions if one does
not take advantage of them to reduce doversion work on file load! Now we
have to raise subversion again just for that. :(
We cannot let those data be generated on-the-fly in RBW evaluation
anymore, since those would be added to CoW eval object and never ported
back to orig objects.
We *could* get orig objects in eval code, of course, but as in
constratints, this is not really threadsafe and future proof, depsgraph
evaluation should really write back to orig data as little as possible.
So instead, add code to ensure required data is generated to objects
when their collection is added to rigidbody world.
Note that we *may* want to clean that up once collection is no more used
by RB? On the other hand, people might want to keep those data around to
be able to switch between different setups easily... So think it's OK to
keep them at least for now.
There is no guarantee that object in rigidbody collection already have a
valid rgigidbody data when rebuilding deg relations, that is often
generated on-the-fly by actual rigid body simulation.
Note that this can be an issue when generating deg relations I guess...
But at least it won't crash anymore.
The issue here is that in the new dependency graph drivers are
individual nodes which depends on what they are driving. This
means that changes to RNA path or property index should ensure
those nodes are updated. Easiest way to do so is to tag relations
for update.
One issue that especially newer users often run into is that they accidentally reset changes to the scene by switching frame without creating a keyframe first.
Therefore, this commit adds a new color that is used to draw properties if their current value differs from the one that would be set when switching to this frame.
This works both for existing keyframes as well as for currently interpolated frames.
Unfortunately the flags in but->flag are full, so I had to move the new flag to but->drawflag and pass that to all relevant functions.
I went with orange for the color since afaics it fits with the green and yellow that are currently used for keyframe states and since it's somewhat reddish to signify that there might be something to look out for here.
Reviewers: campbellbarton, #user_interface, brecht
Reviewed By: campbellbarton
Subscribers: brecht, predoe
Differential Revision: https://developer.blender.org/D3949
This is for 2.80 (though bug I mistakenly merged into was for 2.78.
Duplicate bugs T58127, T58411, T58440, and T58789 all fixed.
Bevel weights and crease are not real Mesh layers so get lost
on coversion of mesh to bmesh unelss the mesh's cd_flag member
tells the converter to create layers for them.
Most code the copies or partially copies meshes uses
mesh_new_nomain_from_template_ex, so copied the flag there.
Hit normal originates from tesselated triangles and isn't the
actual normal used for shading of flat faces. Thus, it is better
to use the actual polygon normals when available.
This makes the `#include <Windows.h>` use more localised and out of the
`image.c` file.
Reviewers: LazyDodo
Reviewed by: LazyDodo
Differential revision: https://developer.blender.org/D4049
Fix T58237: Exporters: Curve Modifier not applied when "apply modifiers" are selected.
Fix T58856: Python: "to_mesh" broken in 2.8.
...And many other cases... ;)
Thing is, we need target IDs to always be evaluated ones (at least I
cannot see any case where having orig ones is desired effect here).
Depsgraph/Cow system ensures us that when modifiers are evaluated by it,
but they can also be called outside of this context, e.g. when doing
binding, or object conversion...
So we need to ensure in modifiers code that we actually are always
working with eval data for those targets.
Note that I did not touch to physics modifiers, those are a bit touchy
and rather not 'fix' something there until proven broken!
Since we started using looptris we no longer need a triangulation
modifier in the highpoly object. In fact having was causing a bug
where baking would be utterly broken.
This fix normal baking. Combined pass still needs a fix to hide the
objects during baking.
Having the hostname allows us to identify which machine rendered which
frame in our render farm.
This simply uses the host's name, and doesn't do any DNS lookup of any
IP address of the machine. As such, it's only usable for identification
purposes, and not for reachability over a network.
Reviewers: sergey, brecht
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D4047
better have this vertex color layer cover the whole 0-1 range
thx @sergey for checking
Maniphest Tasks: T57994
Differential Revision: https://developer.blender.org/D3976
This is unfortunate, but the number of bugs in this configuration
keeps growing, and almost all of them are caused by bug in OpenCL
compiler.
The compiler is not likely to be fixed, since Apple declared OpenCL
deprecated.
This evil commit is aimed to keep officially supported features
of Blender in a good working and stable state.
The fix itself simply is to store the cage object as a pointer instead
of a string/name.
That said baking with or without cage is yielding very different results
than in 2.7.
The issue was caused mpoly array urequired by the cache filling,
but the pointer was never set when preparing render data.
Seems this change is safe enough, in terms it shouldn't be
causing slowdown, since the assignment of mpoly is cheap, but
hard to tell if there is anything else affected by thing underneath.
This fix aims to fix crash/assert failure related on wrong
evaluation order which happens when there is a cyclic
dependency involved.
The rationality of this change is that we can allow use
of uninitialized scalar value, but memory is better be
allocated.
This might not be ideal still, but worth a try.
There were at least three copies of those:
- OB_RECALC* family of flags, which are rudiment of an old
dependency graph system.
- PSYS_RECALC* which were used by old dependency graph system
as a separate set since the graph itself did not handle
particle systems.
- DEG_TAG_* which was used to tag IDs.
Now there is a single set, which defines what can be tagged
and queried for an update. It also has some aggregate flags
to make queries simpler.
Lets once and for all solve the madness of those flags, stick
to a single set, which will not overlap with anything or require
any extra conversion.
Technically, shouldn't be measurable user difference, but some
of the agregate flags for few dependency graph components did
change.
Fixes T58632: Particle don't update rotation settings
We had two different ways of doing it, SurfaceDeform and LaplacianDeform
would do it through a special modifier stack evaluation triggered from
binding operator, while MeshDeform would do it through a regular
depsgraph update/eval (also triggered from its binding op).
This enforces the later to search back for orig modifier data inside
modifier code (to apply binding on that one, and not on useless CoW
one).
Besides the question of safety about modifying orig data from threaded
despgraph (that was *probably* OK, but think it's bad idea in general),
it's much better to have a common way of doing that kind of things.
For now it remains rather dodgy, but at least it's reasonably consistent
and safe now.
This commit also fixes a potential memleak from binding process of
MeshDeform, and does some general cleanup a bit.
The shader is way simpler and run way faster on lower end hardware
(2x faster on intel HD5000) but did not notice any improvement on AMD Vega.
This also adds a few changes to the way the wireframes are drawn:
- the slider is more linearly progressive.
- optimize display shows all wires and progressively decrease "inner" wires
intensity. This is subject to change in the future.
- text/surface/metaballs support is pretty rough. More work needs to be done.
This remove the optimization introduced in f1975a4639.
This also removes the GPU side "sharpness" calculation which means that
animated meshes with wireframe display will update slower.
The CPU sharpness calculation has still room for optimization. Also
it is not excluded that GPU calculation can be added back as a
separate preprocessing pass (saving the computation result [compute or
feedback]).
The goal here was to have more speed for static objects and remove
the dependency of having buffer textures with triangle count. This is
preparation work for multithreading the whole DRW manager.
Right now does not add padding at the end of the buffer.
This seems not necessary but may cause problem on some platform. If needed
we will add this padding (only 2 more vertices).
Using listener here, although I suspect we should be using message
subscriber only. That said, this mimics the behaviour of the buttons
main region.
As for the original bug report what was happening was that when
switching between viewlayers (or when creating one) we may not get the
same active object. So the context breadcrumbs are different.
And the bug itself was that we were missing a redraw on view layer
change.
Aka all the thousand of reports duplicated here.
I should have seen this coming, since I had to add a hack in the first
place because things were "not working".
I should have figured out earlier that COW handles base in a really
special way, with its own special object_runtime_backup hack.
Other software uses this to define UV islands, so we can't just merge
any UVs with the same coordinate. They have to share a vertex too.
Contributed by Maxime Robinot, with changes by me.
Differential Revision: https://developer.blender.org/D4006
This fixes our workaround for until proper solution is accepted
in upstream.
Now, when default view behaves same as it was supposed to (and
as it behaves in OCIO-1.0.9) it is obvious that our configuration
violates own design -- default view is used for cases when
images don't want to be displays using "render" settings.
LaplacianDeform binding handling is a catastrophee in CoW context,
because half of the binding (the laplacian solver cache thingy) is not
saved, and can be re-generated on the fly from stored vcos.
This means that binding is not only done when hitting 'bind' button, but
also at file load, and when some things change.
And this utterly breaks with CoW design, not sure how to fix, will add a
task about that.
But this also means that NULL laplacian solver cache pointer is not a
good check to know whether it is binded or not, only stored vcos are
relevant for that (and the binding flag, of course).
This fixes/clarifies Surface Deform evaluation code that does the
binding, since that part should only be called outside of depsgraph
evaluation, with orig data-blocks and not CoW ones.
Now we have a decent amount of asserts and checks to ensure eveything
works as expected.
Also had to add a special case to get target's mesh in binding case,
since often target's evaluated mesh is not available, in that case (and
in that case only), we can actually compute that mesh (because we are
out of depsgraph evaluation).
Binding and unbinding *has* to happen outside of 'normal' depsgraph
evaluation of modifiers now that we have CoW, otherwise persistent data
stored in modifier data are always lost!
Note that this is only first step of the fix, modifiers code needs also
some work. Surfacedeform one is in next commit, Laplacian case is much,
much more complicated to handle, given how it uses its cached data. :(
This would overlap with buttons in the header. It's smaller to hit, but
still wider than the outlines for resizing, so hopefully it's fine.
Differential Revision: https://developer.blender.org/D4033
Was initially reported when painting on a mesh with armature,
which was failing due to missing bbone cache. The issue was
deeper, and was related on the way which object was used to
calculate crazyspace.
The layout changed when the radius property was added to shape
keys in 2.8, but the RNA code wasn't updated.
Also, even before that, the code didn't do anything to correctly
handle mixing sub-curves of different type (nurbs vs bezier) in
the same Curve object. Now that case is handled correctly but not
very efficiently by allocating a mapping table when necessary. To
recover some performance, a custom index lookup function is added.
When enable annotations with a grease pencil object, the GP Object must be set to Object mode because the annotation Draw and the GP draw are incompatible.
This is something what is caused by OCIO library. The patch
has been submitted there:
https://github.com/imageworks/OpenColorIO/pull/638
For until it is refined and checked we do workaround from
our side.
Solves weird situation when default display name is queried
from OCIO, but Default view being assumed to be set for it.
Now view is initialized to a default view of that display.
That one was flagged as useless since 2.77, and only kept 'to be sure'
everything was OK. This was years ago now, and never got any report on
this, so 2.8 sounds like a good time to nuke it.
Nice side-effect of using new __annotations__ thingy to store
dynamically-generated fields in a class: __annotations__ dict is not
ensured to exist for a given class, so we may end up modifying on of the
parents' one!
Passing depsgraph instead of scene, since a scene does not fully define the
state of object you want to use for the BVH.
Also, mesh_create_eval_final_view and mesh_create_eval_final_render are pretty
much the same, so mesh_create_eval_no_deform and
mesh_create_eval_no_deform_render are as well.
Issue reported on: T58734
Reviewers: sergey
https://developer.blender.org/D4032
No functional changes, but it makes all the coordinates more consistent
(going from small to larger values). It helps debugging in the future
to be able to rule out vertex order as a culprit.
The top-left and bottom-right corners were creating the new area in the
wrong place.
Blender 2.7x only had action zone corners in the top-right, and
bottom-left corners. So it had some hardcoded assumptions based on that.
This commit feels a bit like a hack, but I think it may be fine.
Bug reported via IRC, how to reproduce:
* Change shading to Rendered.
* Split viewport from the top-left corner.
Since it seems that CD_ORIGINDEX is not available for loops,
the only choice is to simply use the loop normals already
computed by depsgraph after evaluating modifiers.
This revealed a bug where the Auto Smooth settings would be lost
from the mesh after complex modifiers, or after edit mesh to mesh
conversion, so restoring them is needed to get correct results.
This only happens after a certain wireframe threshold.
We sort triangles into 2 bins (start and end of the buffer) based on a
threshold and just draw the first bin if the wireframe slider is low enough.
This optimization is disabled for deformed meshes when playback is active.
This optimization is only implemented for meshes object for now.
This should help resolve (to some extent) T58188.
This only happens after a certain threshold.
We sort triangles into 2 bins (start and end of the buffer) based on a
threshold and just draw the start bin if the wireframe slider is low enough.
This optimization is disabled for deformed meshes.
This should help resolve (to some extent) T58188.
The issue was caused by transflag set in geometry evaluation
never copied back top original object.
Now we have a dedicated operation which does all sort copy
back to original object, so we don't have to worry about
atomic assignments or what gets set where.
Still need to move boundbox to the same function, but it
needs some careful doublechecking first.
This seems to be a bug in OpenSubdiv. For now simply use Catmark
subdivision scheme with infinitely sharp edges.
Later on it's either gets fixed in OpenSubdiv or we do bilinear
subdivision on our side.
All localized datablocks are not supposed to have animation
data associated with them.
There was an easy way to reproduce assert failure: toggle
animation decorator for Viewport Display -> Color.
COW nodes in the graph are mostly connected via a relation type
that doesn't propagate the update flags. Unfortunately, due to
the scheduling implementation that means the relations don't
actually guarantee execution order for indirect dependencies.
Relations also don't guarantee order in case of cycles.
As mentioned in IRC, the simplest way to fix possible problems
is to execute all COW nodes as a separate execution stage. This
seems to fix crashes with Data Transfer modifier in a cycle.
Staging works by simply delaying actual scheduling of tasks for
non-COW nodes until the second run of schedule_graph.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4027
To make the pool more usable for running multiple stages of tasks,
fix local queue handling in BLI_task_pool_work_and_wait.
Specifically, after the wait loop the local queue should be empty,
or the wait part of the function contract isn't fulfilled. Instead,
check and run any tasks in queue before the wait loop.
Also, add a new function that resets the suspended state of the pool.
Add special handling for both edge cases (:p):
* 180° is same as no splitting by angle;
* 0° is same as split on all edges unconditionnaly.
In both cases we can also avoid computing poly normals.
This adds an elliptical arc primitive.
Press CKEY for toggling closed/open arc.
Press FKEY key for flipping arc.
Additional changes to gpencil primitives.
Increases default edges of circle to 64.
Keymap changes to allow primitives to be drawn with Shift or Alt key.
Allow Plus/Minus key to adjust number of edges.
Missing: Toolbar icon
Differential Revision: https://developer.blender.org/D4024
After update of the mesh some of that data is so broken that using
it would crash. To reduce the risk of crashes in case of dependency
cycles, clean it up immediately.
Presets were not updated when parameter were changed in rBe3d31b8dfbdc.
Note that will also check on generating more resistent py code for that
kind of presets, since that will also affect any custom preset made by
users...
The hang was due to the nodes being "evaluated" for every incomming link.
Solution: only evaluate once per nodetree.
Also merge the tagging of SSS and SSR into one traversal only.
We separate the background and foreground shading passes to be able to make
the object id pass optionnal if we don't need it.
This saves a bit more memory. Also not clearing all rendertargets saves
some GPU time too.
We exploit the fact that we are using the metallic workflow for material
and pass the metallic parameter instead of the specular color.
Pack the front facing bit in the color buffer only for matcap display.
Change buffer formats to use less bytes as possible.
Also don't request buffers that we won't use.
Saved 40MB on 2K screen on StudioLight + Shadows + Specular Lighting.
Includes several cleanups.
Move all mask-related fields from Object and OperationDepsNode
to Object_Runtime and IDDepsNode. Auto-apply DEG_TAG_GEOMETRY
if the mask changes after DEG rebuild. Update DEG API and all
code that uses it.
This fixes "source mesh data is not ready" errors from Data
Transfer modifier when parameters are changed in the UI after
the recent mesh_get_eval_final fix.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D4025
Our mesh validation was only checking cd layout so far, not their actual
data. While this might only be needed for a few types, this is a
required addition for things like imported UVs, else we have no way to
avoid nasty things like NANs & co.
Note that more layer types may need that callback, time will say. For
now added it to some obvious missing cases...
Previously the shift key for line primitives only allowed diagonals.
This change allows the line to constrain to vertical and horizontal lines.
Differential Revision: https://developer.blender.org/D4012
Ensure we use lists for keymap items and item properties.
This means scripts can access keymap definitions from other layouts,
manipulating them without sometimes encountering a tuple that needs
to be converted into a list.
Instead of doing a lot of alpha blended drawing with jittering, use the
fragment shader to do the masking using a circle mask.
This is much simpler and requires much less resources.
Hopefully this may solve the issue we have with the Intels UHD Graphics 620
on linux.
The cause is that FOREACH_OBJECT_IN_MODE_BEGIN assumed that the active
object is in the correct mode, which is wrong in this case. It also
only considered objects of the same type as active, which had to be
replaced with an explicit type parameter.
Fix the old code that propagates selection changes to the
evaluated mesh directly without rebuilding, and avoid tagging
DEG_TAG_COPY_ON_WRITE if it succeeds.
It's a very bad idea to call this on non-COW instances - see T58150.
Also, when rebuilding mesh it's better to accumulate mask flags to
avoid possible repeated rebuilds from different users.
There was a bug due to non-aligned struct in the DNA that prevented us
to increase the size of the userdef light array.
Since the studio lights are now presets and stored in external files,
there is no need to keep backward compatibility with theses lights.
Remove the old array and create a new one.
Add blue tint light for specular.
Shadow focus let the user choose how hard are is the shadows transition.
Harder shadow transition can be used for stylistic effects or more uniform
shading.
Make shadow orientation respect the same orientation as the studio light
(view from +Y direction aka. front view). Make the default shadow direction
more similar to the default light position (the default light object, not
the default studio lighting).
Texture paint code was retrieving the evaluated mesh from the
original object, which isn't supposed to happen, so the cached
mesh isn't properly cleaned up by Edit Mode toggle.
Note that am not sure that is actually needed, since switching to that
mode does not actually use any eval data, it's only needed during init
of first stroke... But in doubt, that won't hurt to have it here anyway.
This reverts commit 3f31c28a02.
Gives issues zooming, could be resolved but it mostly worked OK before,
and it's not a priority to spend time on, so leave as is for now.
Now used the original dist instead, since using the distance between
the camera and the views offset may seem random from the users POV.
This addresses strange behavior noticed in T56934.
* Move the curvature computation to the cavity pass: One can argue it's not
the best performance wise (it gets a tiny perf pernalty if it is done
alone without the ssao), but it make the code cleaner and reduce
considerably the number of shader variation possible.
* Lower shader variation to 2^8 instead of 2^12
Now when try to add annotation, if a grease pencil object is selected, first the object is unselected. This solution is not perfect but it's better than cancel the annotation.
Thanks Dalai for his help.
This option is per viewport.
Having view space shading make sense when working on isolated objects like
if you were holding them in your hands. But for entire scene work, it is
better to have the lighting fixed to have a better spatial representation.
This changes a bit how the userprefs solid lights works. They are not
visible until enabling the "Edit Solid Light" checkbox. Once enabled the
current studiolight used for solid mode will be overwritten.
Once the lighting settings are tweaked, the user can click the
"Save as Studio light" button to save the current settings.
This makes it easy to create new lighting without messing the other
presets.
The studio lights are stored as ASCII files on the disk using a dead
simple custom format.
The UI/UX is not perfect and will be improved in other commits.
Also includes:
* Separate LookDev HDRI selection from Solid Lights
* Hide LookDev HDRIs from the Solid Lights selection list
For most brushes, texture painting uses a special mask accumulation
table in order to ensure that the amount of added color only increases
when the same pixel is touched multiple times by the stroke.
Unfortunately, only the mask texture was added to the mask before
this check, while normal, stencil, texture alpha masks were applied
after this check. This means that the check can pass if e.g. the
pressure is increased, but the final mask value is actually lower.
One might think that the mask values are fixed per pixel, but with
symmetry that isn't true. The result is a nasty stripe artifact due
to the discrete cutoff nature of the accumulation test.
In order to fix this, apply all masks before accumulation.
This allows primatives to be drawn from the center using the ALT key.
Also fixes SHIFT constraint not working correctly in all directions.
Both options can be used together.
Differential Revision: https://developer.blender.org/D4009
Not sure what those #ifdef's were supposed to do exactly... But one
thing is for sure, clearing that flag in particlesettings after first
encounter would prevent transferring it properly to other objects that
would use same particlesettings.
The original offset was wrong because it applied a constant to
homogenous coordinates (the actual depth is z/w), which broke
totally if near clip distance was reduced.
A correct depth offset has to take slope into account like
glPolygonOffset in order to avoid dotted lines caused by
interpolation precision variations. When drawing wire lines
however only the slope of the line itself is accessible, so
also generally increase the offset when the object is close.
Sculpt (and paint) modes rely on valid evaluated data at their initialization.
Added code to ensure that in `ED_object_mode_toggle()`, when relevant
toggle operator requires it (looks like sculpt/paint should be the only
ones affected, although particle edit may be too...).
While atomics library was trying to use "user-space" defined
LIKELY() and UNLIKELY(), this is not always true that user
code was checking for those macro coming from an unrelated
area.
Some space types are exposed as multiple space types,
previously the key binding to set the space type would use the last
used space-type.
Now pressing the key again cycles to the next space sub-type.
Without this, shortcut display is confusing since some space types share
a key. Keymap display will need to be updated to support this.
This commit adds a sample-based profiler that runs during CPU rendering and collects statistics on time spent in different parts of the kernel (ray intersection, shader evaluation etc.) as well as time spent per material and object.
The results are currently not exposed in the user interface or per Python yet, to see the stats on the console pass the "--cycles-print-stats" argument to Cycles (e.g. "./blender -- --cycles-print-stats").
Unfortunately, there is no clear way to extend this functionality to CUDA or OpenCL, so it is CPU-only for now.
Reviewers: brecht, sergey, swerner
Reviewed By: brecht, swerner
Differential Revision: https://developer.blender.org/D3892
To make consistent with Left click select, now if click outside any point, all points are deselected.
Reduced the circle of selection to get more precission. The radius used before was too wide.
Note: There is a minimum distance to consider outside selection area.
The old onion skinning used in 2.7x has been ported and converted to 2.8. Only basic features have been included. For more advanced onion skin features, use grease pencil objects.
Onion Skin is supported in View 3D and Sequencer.
Not exactly sure why we did not have cached displist for bevel object
here... But anyway, that conversion operation should really happen
outside of depsgraph evaluation area, so makes sense to do it as when
generating geometry for rendering, imho. Also solves issues like loosing
hidden parts of the curve/surface, etc. Still using viewport resolution
for curves, though.
Meshes from evaluated objects may already have modifiers applied, but
that's not the case for curves, we need to do that when converting them
to meshes.
This is in order to have more flexible ligthing presets in the future.
The diffuse lighting from hdris was nice but lacked the corresponding
specular information. This is an attempt to make it possible to customize
the lighting and have a cheap/easy/nice-looking pseudo-PBR workflow.
* Add cheap PBR to Workbench with fresnel and better roughness support.
This improves the look of the metallic surfaces and is easier to control.
* Add ambient light to studio lights settings: just a constant color added
to the shading.
* Add Smooth option to studio lights settings: This option fakes the
effect of making the light bigger making the lighting smoother for this
light. Smoother lights gets reflected like a background hdri.
* Change default light settings to include the smooth params.
* Remove specular highlights from flat shading. (could be added back but
how do we make it good looking?)
* If specular lighting is disabled, use base color without using metallic.
* Include a lot of code simplification/cleanup/confusion fix.
The idea is to make main thread and job threads to be scheduled
on CPU dies which has direct access to memory (those are NUMA
nodes 0 and 2).
We also do this for new EPYC CPUs since their NUMA nodes 1 and 3
do have access but only to a higher range DDR slots. By preferring
nodes 0 and 2 on EPYC we make it so users with partially filled
DDR slots has fast memory access.
One thing which is not really solved yet is localization of
memory allocation: we do not guarantee that memory is allocated
on the closest to the NUMA node DDR slot and hope that memory
manager of OS is acting in favor of us.
See T57857 for discussion. This reverts:
"Outliner: Do not gray out empty collections"
4521d3e707.
"Remove eye column from the outliner"
fd16b35997.
Fix/workaround issues in pose and edit mode"
6d2e2e30d5.
"Per view-layer collection visibility"
4de6a210c6.
Tools can define a function that generates the tooltip using a function,
this takes the tools keymap as an argument which can be used
to extract keys to include in the tip.
A few reasons motivating this change:
* It works well for all devices: mouse, trackpad, and tablet pens.
* For beginners or users coming from other software, it's easier to get
started and avoids an initial stumbling block.
* Many users in 2.7 (about half?) were already using left click select, so
combined with the above advantages it makes for a practical default.
Note that we continue to support right click select, as many experienced
Blender users (and developers) see efficiency advantages in this approach.
The option to switch is in the first time setup splash screen, and in the
user preferences.
Curve/surface/text final data may be an evaluated mesh instead of a
displist, when some modifiers (constructive ones in particular) are
applied to it.
Note that this is just getting feet wet, whole draw code suffers from
the same issue! :P
We still control this in the viewport collections visibility menu. But
now we are actually changing the visibility of the collections, not of
the objects.
If a collection is indirectly invisible (because one of its parents are
invisible) we gray it out.
Also if you click directly in the collection names, it "isolates" the
collection by hiding all collections, and showing the direct parents and
all the children of the selected collection.
Development Note:
Right now I'm excluding the hidden collections from the depsgraph.
Thus the need for tagging relations to update.
If this proves to be too slow, we can change.
To control per-viewlayer, per-object visiblity users should either use
the menu or the H shortcuts (H, Alt+H, Shift+H).
Following next will be to have proper per-viewlayer collection
visibility. This will replace the functionality currently at the
viewport collections visibility menu.
Previously we tried this but reverted (see 64d40c82c3)
because there wasn't a predictable set of keys to use global-space.
Now the keys are swapped:
- 'GX' always transforms in the user defined orientation.
- 'GXX' always transforms in global space.
As before 'GXXX' cycles back to disabling constraints.
This does have a down side that GXX won't be used for local-space
when the user has global space set.
Also, when global is the user-orientation, pressing GX and GXX
does the same thing.
Note: examples here use GX but could be any transform-mode/axis.
This has been a contentious topic: Artists at the Blender-Studio prefer
this behavior, while the user community overwhelmingly prefers 2.7x
operator search. Previously this defaulted to accessing tools
(eg: Space-T activates transform.. Space-R rotate etc) which I still
believe is a better long term default - otherwise we don't have
efficient tool switching for a system we intend to make more use of,
nevertheless as far as I can tell users haven't been keen on adopting
this so far. Show the preference on the setup screen since many users
don't animate and will may want to quickly search or switch tools.
When a modifier depends on some other object's position, then it also
depends in its own position, this has to be also told to depsgraph.
Fixes several modifiers where moving target would update the modifier,
while moving modified object itself would not.
Also fixes a few issues (like meshdeform's EM variant not using editmesh
data), and adds a few optimizations (like only generating that source
mesh when we do have a vgroup defined in parameters, for modifiers only
using it to access vgroup)...
These changes are necessary. Need to mark vertices of edges passed
in geom; also the normals.out slot has a custom element type, not
ELEM, so need to prevent attempt by python code to convert it to
an elem. But this leaves a memory leak. I will rework code to not
use normals.out slot at all, but that's a bigger fix.
Now there is a crash in a different place (GPU code). Think that if
using Op on its own (instead of from edbm_bevel_calc, there needs to
be a dependency graph update and maybe more?
Fixes T57931 Particle weight edit mode is not supported.
There is a bug that prevent refresh of the toolsettings on which is based
the weight / non-weight display selection (see T58086).
* The exporter now gets the view layer from the context
instead of the depsgraph.
* The depsgrap is now fetched only on demand since the graph
is not always needed for exporting (currently only for armature exports).
was caused by NC_SCENE notifier being skipped with a non-scene reference
showed e.g. in timeline not updating keyframes/cachelines
Maniphest Tasks: T57929
Differential Revision: https://developer.blender.org/D4000
Accessing ob->bb directly is not a good idea anyway.
Still, would like to know why/where this bbox is freed, since it is
allocated at least once by depsgraph eval, as part of
`BKE_object_handle_data_update()` function...
Not sharing caused duplication in the keymap and
required a factory class generator.
Simplify tool & keymap definitions by sharing them.
It's highly unlikely we will ever want these to use different keys
once they're set as the active tool.
B-Bone shape is a non-trivial computation, so access to
the results would be useful for Python scripts working with
B-Bones, e.g. rig generation.
This exposes both final segment matrices, and the tangent
vectors computed from the custom handle bones.
Since the handle tangents use the axis+roll orientation math
of edit bones, add matrix conversion static methods to Bone.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D3983
Sometimes the text doesn't fit. What to do in this case?
* Overflow: The default behaviour still is to overflow the text.
* Truncated: If any text box is defined we can also not draw the text
that goes outside the text boxes.
* Scale to Fit: For single-text box texts we can scale down the text until
it fits.
To support textboxes we are bisecting the scale until we find a good
match. Right now the hardcoded iteration limit is 20, and the threshold 0.0001f.
An alternative in the future would be to tackle this by integrating existing
layout engines such as HarfBuzz.
Note: Scale to fit won't work for multiple text-boxes if any of them has
either width or height as zero.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D3874
Feature development sponsored by Viddyoze.
Implement strand selection visualisation but without any shading.
I think this is not the overlay job to draw the strands shaded.
We can already view the children strands shaded for now but we might add
an option to draw the shaded strand instead of (or in addition to) the
guide strand.
This modifier only uses mesh to get vgroup, which is only needed in case
modified object is indeed a mesh! Building a mesh from curve here is not
only useless and time-consuming, it will also easily fail the assert
about same number of vertices!
Note that surface_project and subsurf option also need more work at some
point, but this is probably not that urgent for now.
Also, use MOD_get_vgroup() helper in modifier code itself and pass
resulting MDeformVert & index to BKE_shrinkwrap's `shrinkwrapModifier_deform()`,
this is simpler and avoids duplicating vgroup handling code.
Related to T57972.
This modifier only uses mesh to get vgroup, which is only needed in case
modified object is indeed a mesh! Building a mesh from curve here is not
only useless and time-consuming, it will also easily fail the assert
about same number of vertices!
Also, use MOD_get_vgroup() helper in modifier code itself and pass
reluting MDeformVert & index to BKE_curve's curve_deform_verts(),
this is simpler and avoids duplicating vgroup handling code.
Also fixes crash when used on lattice.
Related to T57972.
This modifier only uses mesh to get vgroup, which is only needed in case
modified object is indeed a mesh! Building a mesh from curve here is not
only useless and time-consuming, it will also easily fail the assert
about same number of vertices!
Also fixes crash when used on lattice.
Related to T57972.
There is a new `bpy.app.timers` api.
For more details, look in the Python API documentation.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D3994
In 2.79 hiding works in paint modes with selection enabled,
so it is a missing feature. This implements it in texture
paint overlays and in workbench base shading.
Reviewers: fclem
Differential Revision: https://developer.blender.org/D3989
Now it's possible define the blend mode between layers including the option to clamp the layer using underlying layers.
Also a new Simplify option has been added to disable blend layers.
The approach is fairly simple, just apply an edge detection filter to the view normal and scale the brightness based on that.
The overlay is disabled at object boundaries to avoid dark lines around objects.
Generally, this implementation follows the proposal of @monio at https://blender.community/c/rightclickselect/J9bbbc.
The changes are:
- Dynamic filter radius (on high-DPI displays, a radius of two is used)
- Options to reduce the strength of both ridges and valleys
- Tweaked function for the strength reduction (the original method actually had a local maximum, resulting in a brighter line inside valleys)
- Multiplication for blending instead of overlay, which doesn't work reliably with scene-referred intensities
- Renamed to point out the distinction between it and the SSAO-based cavity overlay
Reviewers: jbakker
Reviewed By: jbakker
Subscribers: billreynish, manitwo, linko, monio
Differential Revision: https://developer.blender.org/D3617
This fixes conflicts where the tool and editor keymap use different event
types. Tools need to be able to mouse buttons on PRESS without triggering
CLICK events in the editor keymap.
This commit makes it so that subsurf/multires modifiers will respect
the WITH_OPENSUBDIV option. The WITH_OPENSUBDIV_MODIFIER option is
now gone.
For artists it mean that subsurf modifier will behave same as it is
planned for 2.80. Multires will now support sculpting, but it has some
known limitations. Those will be worked on before the final release.
If OpenSubdiv is disabled, no subsurf/multires functionality will
present.
For the details see:
https://wiki.blender.org/wiki/Reference/Release_Notes/2.80/Modeling#Subsurf.2FMultires
The old behavior with two clicks evolved out of a gesture system, and it can
have some advantages if you want to press more keys to constrain for example. But
this seems a better default.
This is a local fix.
The problem with duplicate looptris still remains.
That is, it can still be released in one place but not upgraded in the other.
(note: setting the looptris to NULL in the evaluated mesh and assert whether it is still NULL when the mesh is freed could indicate where those cases are).
Go for the simple solution for now (disable auto-texspace in evaluated mesh).
Proper fix would be part of known TODO redesign of bbox handling.
Solution suggested by @sergey, thanks!
For users that want the 2.7 LMB keymap behavior, this provides a way of
working without tools interfering. For RMB select this operator is quite
redundant with the Cursor tool, we may have to find a solution for that.
Note that we also might later add transform tweak to the transform tools,
when nothing is selected. But this is important for existing users who
preferred the existing workflow.
mesh_finalize_eval() may set ob->data to evaluated mesh, needs to be
done *after* call to BKE_mesh_texspace_copy_from_object(), else that one
is meaningless.
Related to investigations on T57985, but does not solve it at all. :(
Thinks whole bbox code needs a complete rewrite, one can see a lot of
old history in it, it has way too many functions doing
nearly-the-same-thing(c), it spreads in very inconsistent ways across a
lot of files, ... But have no time for this right now, and would not be
a good idea with Beta comming up close anyway.
So for now going the simple and (hopefully) sane & safe way: forbid
object-level functions to affect data-level bbox. Mesh and curve ones
would generate bbox in obdata instead of object, for some reason (all
other obdata types only use object's bbox ever). That may have been
working in old ages, but with CoW and threaded depsgraph this is just
calling for piles of issues.
Implements the first changes for T54115:
* Rename "User Preferences" window to "Settings" in the UI.
We'll likely put workspace settings in there, separate from the global
user settings. System settings should become separate from user
settings in future to allow settings for specific hardware.
* Add sidebar region for navigation (scrolls independently).
Addresses space problems, so we can add more categories as needed now.
* Increase size of Settings window to compensate new navigation bar.
* Group sections into User Preferences and System.
Icons for section groups by Andrzej Ambroz. Thanks!
* Bumps subversion for file compatibility.
Screenshot: https://developer.blender.org/F5715337
I also added categories for future work, but commented them out.
We may also want to redesign contents of each section now.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D3088
Design Task: https://developer.blender.org/T54115
If backface culling is off, the user obviously wants to paint on
back faces, so the normal angle cutoff designed to prevent painting
at glancing angles shouldn't do the culling as a side effect.
Bring back per-viewport localview. This is based on Blender 2.79.
We have a limit of 16 different local view viewports.
We are using both the numpad /, as well as the regular /.
Missing features:
* Hack to make sure lights are always visible.
* Make rendered mode with external engines to support this as well
(probably just need to support this in the RNA iterators).
* Support over 16 viewports by taking existing viewports out of local view.
The code can use a cleanup pass in the future to unify the test to see
if an object is visible (or we can use TESTBASE in more places).
Logic to update child windows on workspace changes should simply ignore
temporary child windows. Users opened those for a specific purpose (i.e. edit
user preferences or show render result). Blender should not come in and
repurpose it.
- Class constructors without body (only attribute initialisations)
can safely be kept in the class header files
- Constructor variables should be initialized in the order of their
definition in the header files
This change is also aimed to remove a couple of
build warnings from the linux builds.
Computing the shape of a B-Bone is a quite expensive operation, and
there are multiple constraints that can access this information in
a variety of useful ways. This means computing the shape once per
bone and saving it is good for performance.
Since the shape may depend on the position of up to two other bones,
often in a "cyclic" manner, this computation has to be a separate
node with its own dependencies.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D3975
Most important changes are in the Animation exporter and Animation Importer.
There is still some cleaning up to be done. But the Exporter/Importer basically
work within Blender 2.8
Some details:
User Interface:
The interface has been reorganized to look more like the FBX interface.
New options in user interface:
* keep_keyframes:
When sampling the distance between 2 keyframes is defined by
the sampling rate. Furthermore the keyframes defined in the
FCurves are not exported. However when this option is enabled
then also the defined keyframes will be added to the exported fcurves
* keep_smooth_curves:
When sampling we do not use FCurves. So we also have no Curve handles
for smooth exporting. However when this option is enabled, Blender
does its best to recreate the handles for export. This is a very
experimental feature and it is know to break when:
- the exported animated objects have parent inverse matrices
different from the unit matrix
- The exported objects have negative scaling
There may be many other situations when this feature breaks.
This needs to be further tested. It may be removed later or replaced
by something less wonky.
BlenderContext:
is a new class that contains the bridge to Blender. It contains
pointers to the current export/import context plus derived values
of Depsgraph, Scene, Main
Reporting:
I reorganized the output on the Blender Console to become more
informative and more readable
Preservation of Item names:
name attributes are now encoded with XML entities. This makes
sure that i can export/import names exactly defined in the tool.
This affects material names, bone names and object names.
Hierarchy export:
* Object and Bone Hierarchies are now exported correctly
by taking the Blender parent/child hierarchy into account
* Export also not selected intermediate objects
Problem:
When we export an Object Hierarchy, then we must export
all elements of the hierarchy to maintain the transforms. This
is especially important when exporting animated objects, because the
animation curves are exported as relative curves based on the
parent-child hierarchy. If an intermediate animated object is missing
then the exported animation breaks.
Solution:
If the "Selected" Optioon is enabled, then take care
to also export all objects which are not selected and hidden,
but which are parents of selected objects.
Node Based Material Importer (wip):
Added basic support for Materials with diffuse color and
diffuse textures. More properties (opacity, emission) need
changes in the used shader.
Note: Materials are all constructed by using the principled BSDF shader.
Animation Exporter:
* Massive optimization of the Animation Bake tool (Animation Sampler).
Instead of sampling each fcurve separately, i now sample all
exported fcurves simultaneously. So i avoid many (many!)
scene updates during animation export.
* Add support for Continuous Acceleration (Fcurve handles)
This allows us to create smoother FCurves during importing Collada
Animation curves. Possibly this should become an option ionstead of
a fixed import feature.
* Add support for sampling curves (to bake animations)
* The animation sampler now can be used for any animation curve.
Before the sampler only looked at curves which are supported by
Standard Collada 1.4. However the Collada exporter currently
ignores all animation curves which are not covered by the 1.4.1
Collada Standards. There is still some room for improvements
here (work in progres)
Known issues:
* Some exports do currently not work reliably, among those
are the camera animations, material animations and light animations
those animations will be added back next (work in progres)
* Exporting animation curves with keyframes (and tangents)
sometimes results in odd curves (when parent inverse matrix is involved)
This needs to be checked in more depth (probably it can not be solved).
* Export of "all animations in scene" is disabled because the
Collada Importer can not handle this reliably at the
moment (work in progres).
* Support for Animation Clip export
Added one extra level to the exported animations
such that now all scene animations are enclosed:
<Animation name="id_name(ob)_Action">
<Animation>...</Animation>
...
</Animation>
Animation Importer:
* Import of animations for objects with multiple materials
When importing multiple materials for one object,
the imported material animation curves have all been
assigned to the first material in the object.
Error handling (wip):
The Importer was a bit confused as it sometimes ignored fatal
parsing errors and continued to import. I did my best to
unconfuse it, but i believe that this needs to be tested more.
Refactoring:
update : move generation of effect id names into own function
update : adjust importer/exporter for no longer supported HEMI lights
cleanup: Removed no lopnger existing attribute from the exporter presets
cleanup: Removed not needed Context attribute from DocumentExporter
fix : Avoid duplicate deletion of temporary items
cleanup: fixed indentation and white space issues
update : Make BCAnimation class more self contained
cleanup: Renamed classes, updated comments for better reading
cleanup: Moved static class functions to collada_utils
cleanup: Moved typedefs to more intuitive locations
cleanup: indentation and class method declarations
cleanup: Removed no longer needed methods
update : Moved Classes into separate files
cleanup: Added comments
cleanup: take care of name conventions
... : many more small changes, not helpful to list them all
In some instances, the number of control vertices of a hair could change mid-frame.
Cycles would then be unable to calculate proper motion blur for those hairs. This adds
interpolated CVs to fill in for the missing data. While this will not necessarily result in
a fully accurate reconstruction of the guide hair, it preserves motion blur instead of disabling it.
Reviewers: #cycles, sergey
Reviewed By: #cycles, sergey
Subscribers: sergey, brecht, #cycles
Tags: #cycles
Differential Revision: https://developer.blender.org/D3695
Explicitly tag copy-on-write form library remap. Previously, this
tag was used implicitly via geometry/transform tagging, which worked
ok for objects. For non-objects we do need to ensure all copies has
correct pointer and the only way to do so is to pass tag explicitly.
There is probably more places in the library remap where this is
needed, but not being familiar with the code makes it difficult to
spot where possible tags are missing.
When dragging in a vertical or horizontal region,
there is no need to detect the drag axis.
Gives minor usability improvement for dragging over vertical tabs.
- Z shows pie menu (removed wire/xray toggles).
- Alt-Z toggles x-ray.
- Shift-Z toggles wireframe.
- Shift-Alt-Z toggles overlays.
Note that toggle overlays had no binding for 2.7x,
this is likely not a heavily used option and could even be left out.
COMMAND${CMAKE_COMMAND}-Ecopy"${PYTHON_OUTPUTDIR}/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib"${BUILD_DIR}/python/src/external_python/run/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}.lib#missing postfix on purpose, distutils is not expecting it
COMMAND${CMAKE_COMMAND}-Ecopy"${PYTHON_OUTPUTDIR}/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib"${BUILD_DIR}/python/src/external_python/run/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib#other things like numpy still want it though.
COMMAND${CMAKE_COMMAND}-Ecopy"${PYTHON_OUTPUTDIR}/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib"${BUILD_DIR}/python/src/external_python/run/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}.lib#missing postfix on purpose, distutils is not expecting it
COMMAND${CMAKE_COMMAND}-Ecopy"${PYTHON_OUTPUTDIR}/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib"${BUILD_DIR}/python/src/external_python/run/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib#other things like numpy still want it though.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
