When enabled prefetching(preview panel>view settings), a pernament running job
is created, that will render frames in the background until the cache is full.
If the cache is not filled fast enough, prefetch job suspends itself
at the last moment and will wait until it has chance to "catch up".
Effectively this will decouple rendering to separate thread, so rendering
itself is a bit faster.
Cache recycling behavior will be changed to "free furthest frame to the left
of playhead if possible, otherwise rightmost frame".
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5386
This change allows the user to select a renderpass in the 3d viewport.
Added support for external renderers to extend the `View3DShading` struct.
This way Blender doesn't need to know the features an external render engine wants to support.
Note that the View3DShading is also available in the scene->display.shading; although this is
supported, it does not make sense for render engines to put something here as it is really
scene/workbench related.
Currently cycles assumes that it always needs to calculate the combined pass; it ignores the
`pass_flag` in KernelFilm. We could optimize this but that was not in scope of this change
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5689
Before there were two options: Paste to original layer called "Paste" and Paste to active layer called "Paste & Merge"
Now, by default the paste is in active layer and the "Paste & Merge" has been renamed "Paste".
For old "Paste", now is called "Paste by Layer" and it's not the default value anymore.
Note: Minor edits to add icons not present in Differential revision.
Differential Revision: https://developer.blender.org/D5591
The specifc bug here came fro; some IDProperties ID pointer storing
references to workspaces.
But that was actually a main loophole in that 'unndoing data while
keeping same UI' process, as we never know who might store a pointer to
one of those datablocks that we want to keep the 'old' version off.
It might actually be ever more needed when we start undoing (changing)
only the IDs actually modified in an undo step...
Notes:
* While not ideal, I think we can afford an extra looping over the whole
Main DB here... Remapping process in itself is fairly cheap, thanks to
the hashes.
* This commit is considered rather risky (especially thanks to 'private'
IDs), think it should work fine for now, unless some IDPointers start
storing references to private IDs...
Once D5559 is in, we shall do another pass here, probably also forbids
assigning private IDs to IDProperties, etc.
This reverts commits 54fd8176d7, 4c5becb6b1 and 8f578150e.
Those kind of commits must be reviewed and approved by project owners.
That one:
* Broke Collada building by not properly updating all calls to modified
function.
* Broke *whole* ID management by not properly updating library_query.c.
And in general, I am strongly against backward ID pointers, those are
*always* a serious PITA for ID management. Sometimes they cannot be
avoided, but in general other ways to get that kind of info should be
investigated first.
NodeTree structures of materials and some other data blocks are
effectively node group data block objects that are contained inside
the parent block. Thus, direct references to them are only valid
while blender is running, and are lost on save.
Fix Copy As New Driver to create a reference that goes through
the owner data block, by adding a new runtime field to bNodeTree.
This commit also provide a compatibility code that will convert old
materials using Additive or Multiply Blend mode to their node equivalent.
This conversion is only done on outputs that are enabled for eevee.
Not even in append case. If those objects are already part of a
collection, we can now leave them fully un-instantiated in any scene,
since user can easily do it themselves as they need to.
This fixes inconsistencies in materials between objects and obdata
due to placeholders generation for missing libdata.
Note that we cannot do that when generating the obdata placeholder,
as not all objects using it might be already loaded...
So this has to be done near the end of the reading/linking process.
Reported here by Blender Studio.
Reviewers: brecht
Subscribers: jbakker, zeddb
Tags: #datablocks_and_libraries
Differential Revision: https://developer.blender.org/D5428
There was a fixed limit to the number of points available in a buffer stroke.
Now, the array is expanded as needed using a predefined number of points for each expansion, instead to add one by one. This is done to reduce the number of times the memory allocation is required.
As part of the fix, some variables have been renamed to reflect better their use.
Some ugly very low-level collection code was using the generic
LIB_TAG_DOIT tag... should never happen, that one is for rather
high-level code to use, core process shall use own tags.
The issue was caused by dependency graph always ignoring animation
update when it is first time constructed. This was a way to make it
preserve unkeyed changes on undo/redo. This, however, made it so
changes of animation data itself (such as deleting/moving keyframes)
did not trigger animation update by the dependency graph.
This worked prior to copy-on-write because animation recalc flags
were stored in the DNA and never re-set on file/undo load. This was
giving dependency graph a clue that animation is to be re-evaluated
when operator explicitly asked to (more precisely, when such operator
was undone/redone).
This change makes it so original ID's recalc flags are storing
recalc flags when ID is tagged for update as an response to user
input. This way re-building dependency graph can force animation
to be updated on redo.
Tricky part here is that ID's recalc flag is no longer to be zeroed
when loading undo step (which is the same as reading .blend file).
This is something what works differently comparing to legacy
dependency graph, which was zeroing object's recalc flags there but
not animation data's recalc flags.
Shouldn't be causing issues, since unkeyed changes are not preserved
upon opening a file anyway, at least to my knowledge.
Related reports which are to be taken into account and verified
they are not re-introduced when making changes in the area:
- T63111: Auto-Bake stuck at constant re-rendering
- T54296: Cycles viewport render stuck on constant re-render
Reviewers: campbellbarton, brecht
Reviewed By: campbellbarton, brecht
Maniphest Tasks: T66325
Differential Revision: https://developer.blender.org/D5316
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
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
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.
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
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.
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.
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.
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. :(
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 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
