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 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.
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
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.
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.
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
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.
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!
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
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.
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
