| Age | Commit message (Collapse) | Author |
|---|
|
|
|
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.
|
|
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.
|
|
|
|
|
|
\Something went wrong in the rename script..
|
|
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.
|
|
It is clear enough already that it goes From to To.
|
|
|
|
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
|
|
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.
|
|
|
|
|
|
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.
|
|
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
|
|
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.
|
|
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
|
|
|
|
Using those functions during multithreaded evaluation is a sure
way to have a race condition and crash.
|
|
|
|
|
|
|
|
|
|
|
|
See T57857 for discussion. This reverts:
"Outliner: Do not gray out empty collections"
4521d3e7074d2e08ca813e1f4a2297f5000f335b.
"Remove eye column from the outliner"
fd16b359977c8932ada8db4ff0e43d0402fdc280.
Fix/workaround issues in pose and edit mode"
6d2e2e30d50c40f302b62d3601b5742d7c7056c6.
"Per view-layer collection visibility"
4de6a210c69fe254518ca8d6c860782c54f03749.
|
|
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.
|
|
|
|
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).
|
|
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
|
|
Ensure that pose cleanup is never run before all IK solvers are finished.
|
|
|
|
|
|
|
|
This way we guarantee that action tagged for CoW will ensure
order of updates with objects which are using it.
Fixes T57870: Scaling keyframes in Graph Editor causes crash
Fixes T57971: Moving keys in the Dopesheet crashes with Segfault 11
Fixes T57530: Moving keys on a rig, Crash
Fixee T57491: Crash when trying to move keyframe in dope sheet
|
|
Five years later since the original commit, is probably not so bad
timing on calling things by their clear name.
|
|
Previously each of the objects which has armature modifier will
request deformation matricies from bbones. Thing is, all those
deformations are the same and do not depend on object which is
being modified. What's even worse is that this calculation is
not cheap.
This change makes it so bbones deformation is calculated once
and stored in the armature object. After this armature modifiers
simply use it.
With a rigs we've got here dependency graph evaluation time
goes down from 0.02 sec to 0.012 sec.
Possible further optimization is to make bbone deformation
calculated at the time when bone is calculated. This will avoid
an extra threaded loop over all bones.
|
|
Ideally would need to make check based on evaluated mesh,
but is now a good time to go away from derived mesh.
|
|
|
|
|
|
The issue was caused by dependency cycle solver killing relation
which was guaranteed various things: i.e. copy-on-write component
orders and pose evaluation order (which must first run pose init
function).
Now it is possible to prevent such relations from being ignored.
This is not a complete fix, but is enough to make this specific
rig to work. Ideally, we also need to run copy-on-write operation
prior to anything else.
|
|
|
|
They are self-contained now, and should not cause any sync or
free happening when freeing them.
|
|
It was missing handling of collections there, which caused collection
used for smoke colliders to not be in the dependency graph.
|
|
Do early output when trying to add target relation with wrong rna_path.
We can't do anything reliably in that case anyway, so hopefully it is
a no-functional-change for artists, just avoids noisy error prints in
the terminal.
|
|
This finished old standing TODO which was attempting to
ignore objects of all invisible collections.
The difference here is that we remove invisible bases from
view layers. This guarantees that the evaluated state is
consistent and does not reference original objects.
|
|
For some reason relations can not always be found. This is to be
investigated, but doesn't hurt to be safe here. Also allows to
unlock production.
|
|
|
|
|
|
Rig generators (e.g. rigify) use dummy drivers to attach extra
ID references to rigs for Append/Link. These references use
Single Property variables with empty path, which is intentionally
invalid and shouldn't produce low-level error logs from depsgraph.
The removed code would always print an error because RNAPathKey
requires a valid path to produce a non-NULL reference.
|
|
With the current implementation it only confuses logic around
checks like vewport/render subdivision levels.
If this mode is really needed for any decision making, implement
this properly.
|
