Blender will crash if a file with an Alpha Over node was saved in 4.4,
then in 4.5 alpha, then back to 4.4. This is because the node was
missing forward compatibility code that allocates storage that was
necessary for the file to load correctly on older versions.
This was essentially missed in b62ef2cdd6 and 976efdcac9.
Pull Request: https://projects.blender.org/blender/blender/pulls/141873
Don't attempt to fill curve caps when the direction vector is invalid.
This prevents the crash in #141612 however the root cause of that
report isn't directly related to curve filling.
The check to prevent overly complex tessellation checked the objects
scale directly instead of the final evaluated scale.
Also corrects the scale check which wasn't accounting for negative axes.
When the layer tree in the evaluated state of the Grease Pencil object changed,
the code would fail to get the crazyspace deformation.
Currently we rely on a 1 to 1 index mapping of the original and evaluated
layers. For obvious reasons, this is very weak and can easily break.
The new implementation works as follows:
* Caller that wants to get the crazyspace deformation passes the evaluated and original
object + the original drawing to get the deformation of.
* Fallback deformation are the original positions.
* If there are drawing edit hints in the evaluated geoemtry set, then
* find the edit hint that corresponds to the original drawing
* use the positions in the edit hint.
To create the drawing edit hints, we need to know what evaluated layer corresponds
to which original layer. Currently, this simply stores the original layer index on the
evaluated layer runtime data.
The solution is not ideal and there are some possible improvements like:
* Find a way to solve the more general case, e.g. when there are multiple original
IDs involved.
* Propagate the "mapping" to original layers even when the type of geometry is
changed, like going to curve instances and back.
Pull Request: https://projects.blender.org/blender/blender/pulls/139285
The memfile undo data-block change detection didn't work for meshes
because we ended up writing a new pointer every time. In practice the
array the pointer references is always empty anyway, so we can just add
a check and write null instead.
Unfortunately this fix only applies to 4.5, since the attribute DNA
data (which is actually used at runtime in 5.0) is created temporarily
specifically for writing, so it gets a new address every time.
We'll probably need to solve #127706 in 5.0 to fix this.
Pull Request: https://projects.blender.org/blender/blender/pulls/141457
Tune the grain size used for the parallel_for to alleviate excessive
mutex contention inside `handle_fan_result_and_custom_normals`.
I happened to notice that the 4004 Moore Lane USD scene[1] experienced a
load time regression compared to the prior release. It looks due to the
grain size used and here are some 3-run averages for the import:
```
Grain | Time in seconds
256 (main) | (14.6+14.6+14.8)/3 = 14.6667
1024 | (13+12.8+12.9)/3 = 12.9
4096 | (13.3+13.1+13.1)/3 = 13.1667
16384 | (12.2+12+ 12.5)/3 = 12.2333
65536 | (9.4+9.2+9.6)/3 = 9.4
131072 | (7.9+7.7+8)/3 = 7.8667
262144 | (7.3+7.1+7.2)/3 = 7.2
max(16384, #verts/2) (PR) | (7.1+6.9+6.8)/3 = 6.9333
```
This PR gets the scenario loading in just under 7 seconds now compared
to over 14 originally.
[1] https://dpel.aswf.io/4004-moore-lane/
Pull Request: https://projects.blender.org/blender/blender/pulls/141249
When moving a handle so its length is 0, the other handle was automatically set to 0 as well.
Cancelling the action didn't move the other handle back to 0, but set its length to 1.
The reason for that was that the code tries to maintain the relation between the two handles,
but that would lead to a divide by 0 when either of the handles has a length of 0.
So it would set the handle length to 1 in that case.
The fix is to ensure that the length of a handle is never 0. This is done by clamping it at the key
position with a threshold of 0.001 and an extra floating point step to support large floating
point values.
This should not affect animation in any way.
This was discussed in the animation & rigging module meeting
https://devtalk.blender.org/t/2025-06-26-animation-rigging-module-meeting/41272#p-153605-patches-review-decision-time-5
Co-authored by Nathan Vegdahl
Pull Request: https://projects.blender.org/blender/blender/pulls/141029
The `CurvesGeometry::resize` function would not work properly when
passing a new size of zero.
1) The `CustomData_realloc` would allocate an array with zero size.
2) A new size of 1 is passed to `implicit_sharing::resize_trivial_array`
for the curve offsets when it should pass 0.
3) A value is written out-of-bounds to `curve_offsets`
This fixes the issues in the following ways:
1) `CustomData_realloc` now doesn't allocate any data when the new
size is 0 and sets the `layer->data` to `nullptr`.
2) When the new number of curves is 0, resize the `curve_offsets` to 0
not to 1. This also ensures that the data pointer is `nullptr`.
3) Make sure to only write the first and last curve offset when the
number of curves is greater than 0.
Pull Request: https://projects.blender.org/blender/blender/pulls/141333
We do not do any audio time stretching when doing time remapping.
As we also use the audio to keep track of time passed, we should always
pass the actual timestamp we are starting playback from.
The issue was that the old logic would try to undo time remapping on cfra.
However cfra is already correct and is not remapped. I'm guessing that
this changed at some point and we forgot to fix it up in the sound code.
Pull Request: https://projects.blender.org/blender/blender/pulls/141288
Detection of whether an object is in Edit mode can rely on checking its
obdata ID status (for meshes e.g.), which will falsly make remapping
code think it is remapping the obdata of an edited object.
Somewhat work around the issue by forcing such remapping for the time
being. On the long run, check should be updated to only rely on the
Object's status.
On a Suzanne mesh with 125k faces subdivided at level 3, this patch
results in a 40MB savings of memory, from 1.24 GB measured in Blender
itself to 1.20 GB with this patch applied.
Pull Request: https://projects.blender.org/blender/blender/pulls/141167
The issue was that the depsgraph was not rebuilt, thus
the driver node still stuck around. This then crashed when the
depsgraph evaluated.
The reason why this wasn't caught in the unit tests, was because the
depsgraph was not updated between creating and removing the data.
Pull Request: https://projects.blender.org/blender/blender/pulls/141272
#138435 was an attempt to fix the issue in #138168 where the Smooth
Operator modifies locked vertex groups. Unfortunately, the fix actually
changed some already-correct code to be incorrect to compensate for the
buggy code in the Smooth Operator.
This reverts that fix and applies a correct fix, which is to exclude
locked vertex groups in the Smooth Operator's code itself.
Pull Request: https://projects.blender.org/blender/blender/pulls/141093
The `StringRef` in `MemoryBlobReader` was referencing data
that was freed under some circumstances. While the actual packed
data is shared between the original object and copy-on-eval object,
the name of the blobs are not.
This check was meant to catch unknown failure cases but instead it
meant any attribute with an empty array (e.g. from a mesh with no
faces) would not be loaded. Instead the failure case should only be
when there is no data when the array is not meant to be empty.
Pull Request: https://projects.blender.org/blender/blender/pulls/141096
When armature is selected for custom shape of a bone, we run into
stack overflow due to recursive calls. To prevent this situation,
exclude armature object type from the custom shape list.
This also clears the `pchan->custom` pointer after loading/linking a
blend file. Otherwise a linked Armature object can still be assigned
indirectly, by renaming objects in the library file.
Fixes: #140747Fixes: #140959
Co-authored-by: Sybren A. Stüvel <sybren@blender.org>
Pull Request: https://projects.blender.org/blender/blender/pulls/140851
Caused by change in `AVCodecID` enum. Our `IMB_Ffmpeg_Codec_ID` enum,
which was stored in .blend file did match `AVCodecID`. But after change
in external library headers it fails to initialize correct codec.
Function `MOV_av_codec_id_get()` was added to map these 2 enums and
the values are no longer hard coded.
Functions, that accepted `int codec_id` argument were modified to accept
`IMB_Ffmpeg_Codec_ID` or `AVCodecID` types.
Because `codec_id` was still stored as integer in DNA, get/set struct functions
were added. This way, compiler always knows what type is used and emit
error/warning when incompatible types are used.
Unfortunately, there is no way to ensure, that DNA `codec_id` field is not
accessed directly in future code other than comment.
Pull Request: https://projects.blender.org/blender/blender/pulls/140745
The metaball selection radius was inside the cube and could only
be selected with wire-frame shading.
Resolve by expanding the radius by the dimensions of the cube.
After a recent commit, multiple drawings were always written using the
same address because we used the same stack memory in a loop. This
causes the blend file reading to not be able to distinguish between the
structs, meaning the file is corrupt. However, we already had the same
problem in some cases because of the inline buffer in `BlendWriteData`.
To resolve this, make all "temporary" data for writing live as long as
the writing is going on for the ID. This is somewhat inefficient since
it makes memory reuse impossible for this temporary data. In the future
we should use a technique like #127706 to address this.
For testing, I saved and loaded multiple production files with Grease
Pencil objects, with and without the "write with attribute storage"
option enabled. The fix still goes to 4.5 though, because the first
mentioned issue is present there, and the `CustomDataLayer` vector
address reuse is potentially a problem too.
Pull Request: https://projects.blender.org/blender/blender/pulls/140667
GreasePencil ID type did not list `ID_OB` as its potential ID dependency
type, leading to lack of remapping of its layers' parents objects
pointers (e.g. on Object deletion), leading to crash from accessing
freed data.
`blend_write_prepare` expects to work on a shallow copy of the
`CurvesGeometry` struct. This is necessary because it changes
the struct to contain only the layers being written. This is
particularly important with writing with the new format enabled,
where attributes are written as part of `AttributeStorage`.
Writing with a shallow copy of the struct also gives us the
flexibility to zero the runtime pointer in the written file which
is generally a good practice.
Pull Request: https://projects.blender.org/blender/blender/pulls/140560
Create a good default name for saving individual frames of a movie file loaded
as an image datablock, instead of the movie file name.
Changes ImBuf to store the frame separate from the filepath, to implement this.
Seems more clear for ImBuf.filepath to be an actual filepath anyway.
Thanks to Jesse and Aras for investigating this bug.
Pull Request: https://projects.blender.org/blender/blender/pulls/140471
Previously `blend_name` was using the same path to get its file name
as `//` used. However, the choice of path for `//` seems to be
bespokely chosen at each call site. There may be a logic to it, but
if so it's not immediately clear what it is.
This PR changes `blend_name` to instead always use the currently open
("global") file path, making it well-defined and predictable for
users.
This also prepares better for PR #139438 for 5.0, which adds
`blend_name_lib`, which always uses the blend file that the ID owning
the path is linked from. Over-all this should be much more predictable
and controllable for users.
Pull Request: https://projects.blender.org/blender/blender/pulls/140474
Auto normalization used to not work on assign/remove vertex group
operator, this was due to `BKE_object_defgroup_validmap_get` and
`vgroup_parray_alloc` did not handle grease pencil type objects. Now
added grease pencil cases in them and auto normalization works as
expected.
Pull Request: https://projects.blender.org/blender/blender/pulls/139912
This patch removes node init functions that currently exist only for
forward compatibility and moves the logic to the forward compatibility
section of node writing. This is to avoid allocating unused data
throughout the 5.x series.
Pull Request: https://projects.blender.org/blender/blender/pulls/140273
This assert added by 321ec72c74 notes that changing the mesh
wrapper type from subdivision to normal mesh data just discards the
potentially-subdivided geometry and doesn't make sense, and possibly
breaks other assumptions.
This wrapper type change was done years ago by 0f89bcdbeb to
fix issues with shared object data during evaluation. It noted that
the mesh drawing data extraction didn't handle BMesh wrappers correctly
when the object isn't in edit mode, but that doesn't seem to be the
case after two trivial changes. The other bugs mentioned by that commit
are still solved after this change.
Pull Request: https://projects.blender.org/blender/blender/pulls/140290
The transition to `AttributeStorage` requires the active and default
UV map status to be stored separately, since currently they're
stored as flags on `CustomDataLayer`. This commit adds forward
compatible reading for when the status is written as attribute names
on `Mesh` instead, which is what we already do for color attributes.
Pull Request: https://projects.blender.org/blender/blender/pulls/140134
Previously, the subdivision level was clamped to 11. While such high subdivision
levels are rare, there are still use-cases for going above those if hardware
resources allow for it.
This patch sets the hard upper limit to 15. When going above that, even
subdividing a single triangle would result in data that's too large to store in
`Mesh` because it has too many face corners for an `int`. Furthermore, instead
of clamping the subdiv level, there is an error when going above that. This
might help when accidentally connecting a value >= 16 to the subdiv level input.
Pull Request: https://projects.blender.org/blender/blender/pulls/140057
In 884ef238c0 line art added a safe guard in depsgraph iterator to
properly handle dupli-objects, but it should check original objects id
for inclusion instead of evaluated objects. Now dupli-objects will show
up correctly.
Pull Request: https://projects.blender.org/blender/blender/pulls/140095
This commit effectively reverts the changes introduced in
23951e1b12 that changed how the brush
strength curve calculation works. Since that commit, a number of high
severity regressions have been reported, and the decision to remove
filtering from this method now seems dubious and likely to cause further
problems.
Other interim fixes that were made have not been reverted, as they are
generally correct to do, even if they cause an extra loop over all of
a node's factors & distances. In general, all existing unnecessary calls
to `filter_distances_with_radius` are planned to be removed in 5.0.
Finally, to address the original issue that the mentioned commit was
intended to fix, we now check distance for the constant falloff and
zero out the factor if necessary.
Pull Request: https://projects.blender.org/blender/blender/pulls/140181
Fixed by not doing async loading and always stage correct
texture reference.
Unfortunately the code is getting a bit messy since the
texture loading is not done at the GPUMaterial level.
So we need one async and one synchronous path inside
`PassBase<T>::material_set`.
`ImageGPUTextures` now contains references to the location
of the future `GPUTexture *`.
Also fix#140001
Pull Request: https://projects.blender.org/blender/blender/pulls/140203
Code selecting objects in collections that have been instanced was only
handling objects directly owned by the collection, not the objects in
the hierarchy of children collections.
The objects from the whole hierarchy need to be hanlded here, since
children collections will not be processed (as they are already
instanced by their parent one).
Pull Request: https://projects.blender.org/blender/blender/pulls/140068
Depending on things like order or processing etc., appending
liboverrides data could generate error message in the console regarding
invalid override hierarchy data.
While this could probably be solved, it feels fairly brittle and risky
to directly make linked liboverride hierarchies locale anyway, so just
alsways make local copies of these on append.
Currently, there is the general rule that outputs must be above inputs in any
given panel. Since internally a panel toggle input is part of the panel that it
toggles and has to be the first item in a panel, it was not possible to have a
panel toggle for a panel with outputs.
This patch relaxes the rule for ordering of inputs and outputs for the special
case of panel toggles.
Pull Request: https://projects.blender.org/blender/blender/pulls/140054
This allows to reduce the waiting time caused by
shader compilation on some GPU-driver combo.
A new settings in the User Preferences make it
possible to override the default amount of worker
threads and optionally use subprocesses.
We still use only one worker thread in cases where
there is no benefit with adding more workers
(like AMD pro driver and Intel windows).
It doesn't scale as much as subprocesses for material
shader compilation but that is for other reasons
explained in #139818.
Add some heuristic to avoid too much memory usage
and / or too many stalls.
Also add some heuristic to the default number of subprocess for
the platform that shows scalling.
Historically, multithreaded compilation was prevented by the
need of context per thread inside `DRWShader` module.
Also there was no good scaling at that time. But
nowadays numbers shows different results with
good scaling with reasonable amount of threads on many
platforms.
Even if we are going for vulkan in the next release
most of the legacy hardware will still use OpenGL for
a few other releases. So it is relevant to make this
easy improvement.
See pull request for measurements.
Pull Request: https://projects.blender.org/blender/blender/pulls/139821
