This PR fixes live viewport stability issues on Mac when MetalRT is enabled.
There were two sources of instability:
1) `MTLAccelerationStructure` instances were not being correctly retained meaning that use-after-free crashes could occur following a geometry sync.
2) `MTLIntersectionFunctionTable` objects could be unsafely shared between multiple `MetalDeviceQueue` instances (in this case, `setBuffer` being the unsafe mutation)
The solution to 2 involves creating a new `MetalDispatchPipeline` type which is strictly used by only 1 `MetalDeviceQueue` instance.
Pull Request: https://projects.blender.org/blender/blender/pulls/124055
the object in volume stack should be used instead of `isect.object`.
NOTE: this solution does not work for overlapping volumes. But since
light linking of overlapping volumes did not work before, it should be
fine to implement this partial solution. We read the bottom of the stack
instead of the top to avoid looping through the entire stack.
Pull Request: https://projects.blender.org/blender/blender/pulls/124341
This PR fixes a buffer overrun crash in the MetalRT backend. When non-traceable objects are in the scene, 'num_motion_transforms' is undercounting and the downstream buffer writes (i.e.`motion_transforms[motion_transform_index++]`) are overrunning.
Pull Request: https://projects.blender.org/blender/blender/pulls/124351
Similar to !124267 and !124261
This normal data is eventually passed into `BKE_mesh_set_custom_normals`
and through to `mesh_normals_corner_custom_set` which expects normals to
actually be normalized per its documentation.
Not doing so would yield meshes with incorrect "sharp" data for affected
edges.
Thx @OmarEmaraDev for the initial patch
Pull Request: https://projects.blender.org/blender/blender/pulls/124336
This PR adds a tag to prevent `kernel_data.integrator.seed` being baked into Metal pipelines as a specialisation constant when full kernel specialisation is enabled. This stops new pipelines from being continually compiled when animation is playing in live viewport mode.
Pull Request: https://projects.blender.org/blender/blender/pulls/124349
The cause of the crash was that asset libraries are freed, but the filtered
trees used for things like node tool menus weren't marked dirty, so we
continued to use the freed asset data. On a design level this is quite
concerning because it means the lifetimes of the asset tree caches and
the asset system's asset libraries are connected tenuously.
Eventually there should be a better solution, but I found the simplest
fix for now is just to add a notifier when opening a new file that causes
the asset tree caches to be tagged dirty.
Pull Request: https://projects.blender.org/blender/blender/pulls/124254
The `enum_definition` property was changed by #121234 and now returns
the node itself. This creates an infinite recursion in the override
system that must be avoided by adding this flag.
Pull Request: https://projects.blender.org/blender/blender/pulls/124347
The issue here was that the left handle of the first point was calculated
incorrectly (set to the position of the point itself), which caused a
NaN when attempting to normalize a zero vector when computing extrapolation
vectors.
Pull Request: https://projects.blender.org/blender/blender/pulls/124239
Update the description to mention the node is now based on the OpenPBR
model rather than the Disney Principled Model. This has been the
case since Blender 4.0.
Pull Request: https://projects.blender.org/blender/blender/pulls/124209
When clearing depth attachment the write depth state should be enabled.
When this isn't the case a clear texture is being performed. Before
the render graph this used to be a performance issue as the rendering
would be suspended.
The render graph currently reorders these texture clears before of
the rendering context and therefore isn't seen as a performance issue.
Pull Request: https://projects.blender.org/blender/blender/pulls/124340
The draw tool was creating catmull-rom curves because the
curve type attribute was being reset to the default (0) which
happens to be the catmull-rom type.
The fix makes sure to skip over the `curve_type` attribute when
initializing the other attributes to their default.
This change allows to pass optional freeing and copy callbacks to
functions setting the `func_argN` member of the `uiBut` and `uiBlock`
structs. These callbacks are used to free or duplicate the data stored
in these void pointers.
Defaults are set respectively to `MEM_freeN` and `MEM_dupallocN`, to
match previous hard-coded behavior.
Utils template functions are added to easily generate callbacks to
types that are managed with the `MEM_new`/`MEM_delete` C++-style
allocation, as long as they provide a copy constructor.
Note that this change should be considered as a temporary, transitional
solution. A more robust and future-proof solution is to move towards
more usage of `std::function` for these callbacks, which can then own
and manage their custom data themselves.
Allow translation of a few missing messages in the Rigify add-on. In
particular, the most important "Generate Rig" button!
Some issues reported by Gabriel Gazzàn.
The main change from this commit is the usage of ASAN poisoning (if
available) to trigger an ASAN report on the erroring memory block.
The main benefit is the report of the allocation backtrace of that
faulty memory block.
Pull Request: https://projects.blender.org/blender/blender/pulls/124231
- Suppress pylint warnings via comments or be minor changes.
- Any errors disabling add-ons before upgrading or uninstalling are now
reported to the operators.
- Disable cyclic import warning is it's impractical to resolve.
- Enable useless-suppression warning.
The `[build.generated]` section now includes a filtered list of wheels
to simplify checking existing wheels on the server.
Also disallow quotes & control characters in wheel paths because they
already shouldn't be used and doing so would cause escaping issues in
the generated TOML.
Implements #124242.
This normal data is eventually passed into `BKE_mesh_set_custom_normals`
and through to `mesh_normals_corner_custom_set` which expects normals to
actually be normalized per its documentation.
Not doing so would yield meshes with incorrect "sharp" data for affected
edges.
Pull Request: https://projects.blender.org/blender/blender/pulls/124267
Calling `MEM_freeN` on data allocated with `MEM_new` is bad, since it will not
call a destructor matching the one invoked as part of `MEM_new`.
Would crash with the guarded allocator, which detects such mismatches now.
Exceptions need be caught and forwarded so they show it in Blender's
interface. While most common errors where accounted for, various IO
errors & malformed JSON/TOML could cause internal operations to fail
with unhandled exceptions.
