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).
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).
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.
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 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.
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.
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.
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.
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.
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 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
