The cache invalidation worked on basis of timeline frame instead of
frame index, which caused, that these images were outside of range, that
was considered to be changed, after strip was moved.
In this case the issue happened only with effect strips. This was not
an issue before, mainly because raw images for these strips were not
stored. For other strip types, raw image does not usually change.
This seems to be "caused" by f4f708a54f .
Pull Request: https://projects.blender.org/blender/blender/pulls/124870
When using particle hair the optional generated color buffer was bound
as a texture (texel buffer), however in Vulkan texel buffers aren't
textures but buffers.
This change will create a texture, that will pass the needed calls to
the source vertex buffer. Only particle hair does uses
`GPU_texture_create_from_vertbuf`. I would expect that we will phase
this out eventually when particle hair is removed.
This fixes crashes with scenes using particle hair where materials use color
or uv attributes.
Pull Request: https://projects.blender.org/blender/blender/pulls/124854
It would not execute the "Frame Selected" action if the viewport already
fully zoomed out on the y-axis. However we still need to center the
viewport around the selected strips x position.
Prevent doing layer collection resync during most of
`blo_read_file_internal` process. This is dangerous (since during
blendfile read there is no way to ensure that Main is in a fully valid
sate), and a useless costly process.
There is no reason to deal with layer collections resync in non-main
case, deletion code should never trigger it anyway.
This will avoid quite a lot of the non-main-thread calls to
`BKE_layer_collection_resync_forbid` (detected and 'fixed' in previous
commit).
This commit allows re-entrant calls to
`BKE_layer_collection_resync_forbid` and
`BKE_layer_collection_resync_allow`, such that higher-level code can
block layer collections updates while calling lower-level code that does
the same thing.
E.g. `BKE_id_delete` blocks and then re-allows such updates. However,
such code can also be called from other pieces of code that also need
to prevent these updates in their own, broader scope.
In addition, realized that this call had concurrency issues since BKE
code using it is called from non-main thread too (e.g. from depsgraph
evaluation).
So made the counter atomic to ensure all calls to
`BKE_layer_collection_resync_forbid` is matched by a call to
`BKE_layer_collection_resync_allow`. Also added a cap to max amount of
're-entrant' calls (has to be quite high, due to potential concurrent
calls too).
This is required to fix#124777.
File from #124777 (from 2.79.1) cause a temp bScreen to not be properly
tagged as such, which prevents its deletion when related window is
closed, and leaves a zero-user bScreen in Main.
While this should not happen, this does not seem trivial to track down,
and not an important enough issue to spend more time on it.
Also reported as #124857.
This commit fixes a issue where the Cycles "cast shadow" setting
versioning from Blender 4.1 to 4.2 wasn't working properly due to the
use of the wrong parameter name when grabbing the Blender 4.1 setting.
Pull Request: https://projects.blender.org/blender/blender/pulls/124834
Multi-layer image passes are missing after dragging and dropping an EXR
into the compositor editor. To fix this, we tag the new nodes as needing
an update due to changed ID.
Pull Request: https://projects.blender.org/blender/blender/pulls/124855
The Cryptomatte node makes the UI very slow in Image mode. That's
because the node updates its available Cryptomatte layers on every
redraw, which is a blocking operation that is not very fast.
To fix this, we only update the layers when engine layers change for
the Scene mode, and when image settings change for Image mode.
Pull Request: https://projects.blender.org/blender/blender/pulls/124738
Cryptomatte layer initialization required the active scene to update the
image user frame, but the image user is already updated through other
mechanisms and is thus redundant to do at draw time, so we remove the
frame update call as well as the scene argument from the call chain. The
frame number of the image user is ignored at compositing time in any
case since it is set to the compositing scene frame.
This is needed for #124738.
Caused by 525293a95d
Crash due to attempt of double freeing. Fix is to make first and
last nullptr of list after freeing the memory. This avoids execution of
`folderlist_free` (`if(folderlist_clear_next())` is false)
Pull Request: https://projects.blender.org/blender/blender/pulls/124781
Part of #118145.
Change the "Pivot to Unmasked" and "Pivot to Mask Border" operations
to avoid the vertex iteration macro. The former also uses a proper
weighted average now, so mostly-masked vertices have less influence.
The whole operation is multithreaded now so I expect it's a bit faster,
but I didn't explicitly test that.
Pull Request: https://projects.blender.org/blender/blender/pulls/124743
While this is a rather fragile fix, the alternative also seems fragile.
We can instead check the operator name inside `push_begin_ex` to see if
the operator is transitioning into Sculpt Mode and initialize the
StepData.type to a custom "Mode Change" type instead of using "None"
Pull Request: https://projects.blender.org/blender/blender/pulls/124640
This commit adds SculptBoundaryPreview to reduce the scope of variables
that the SculptSession has access to when displaying the preview for
using the Boundary brush. Further changes in this area will be made once
more of the relevant flood_fill and boundary code is refactored.
In particular, the main algorithm does not need the edge position data.
Pull Request: https://projects.blender.org/blender/blender/pulls/124759
This fixes a few issues and clear up some confusion
in the code.
Note that this changes the behavior of render region;
they now reduce the internal render size. This is
matching the new design documentation.
- Data passes have correct accumulation.
- Adhere to naming conventions for extents, film and render pixels.
- Jitter over final pixels first before doing random sampling
in order to speed up convergence.
- Ensure enough sample to cover at least all the film pixels once.
- Always include the four neighbor pixels in case one is nearer.
- Fix projection matrix computation to align overscan pixels.
Pull Request: https://projects.blender.org/blender/blender/pulls/124735
EEVEE can bind layers of a texture that is also used as an attachment. When binding
the image layout of these specific layers can be different that the image layout of
the whole image.
This fixes the known synchronization issues inside EEVEE. wasp_bot, tree_creature and
wanderer scenes can be rendered without any synchronization issue reported by the
Vulkan validation layers.
Design task: #124214
When beginning to render the attachments are being evaluated. If there is an arrayed
texture (with multiple layers) the individual layers of that texture can be tracked
during until the rendering is ended.
When the same texture is bound to a shader it will be a different layer (otherwise
there is a feedback loop, which isn't allowed). The bound layers will typically need
a different layout the transition to the new layout is executed and recorded. When
the rendering ends, the layers are transitioned back to the layout the texture is
expected in.
It can happen that a layer is used multiple times during the same rendering. In
that case the rendering should be suspended to perform the transition. Image layout
transitions are not allowed during rendering.
There is one place where a layer needs to be transited multiple times that is when
EEVEE wants to extract the thickness from the shadow. The thickness is stored inside
the gbuffer_normal which is also used as an attachment. Eval then samples the thickness
from the gbuffer_normal as a sampler. To work around this issue we suspend the rendering
when a `GPU_BARRIER_SHADER_IMAGE_ACCESS` is signaled.
Pull Request: https://projects.blender.org/blender/blender/pulls/124407
Retiming draw order changes in d0cf4a4a8b accidentally broke VSE
thumbnails. The order of drawing became "for each strip: draw thumbnail,
draw retiming continuity".
But the "draw retiming" bit was changing the projection matrix, which
placed all the later thumbnails at nonsensical locations.
Fixup it so that:
- First all thumbnails are drawn, then all retiming continuities.
- Retiming continuity drawing sets up matrix just once, instead of
doing all that work for each strip.
- And it also restores the previous projection matrix after it is done.
- Also, can_draw_retiming now says "false" when no retiming keys are
present. Previously, code was going into possibly expensive drawing
setup for any strip, even when the is nothing to draw.
Pull Request: https://projects.blender.org/blender/blender/pulls/124780
When trying to delete keys from a layered action in the viewport
using ALT+I and no keyingset is set, no keys would be deleted.
That is because the code hasn't been updated to work with layered actions yet.
The fix is to check if the action is layered and iterate the
FCurves depending on how they are stored.
For this to work nicely I had to extract the function `delete_key_fcurve`.
Pull Request: https://projects.blender.org/blender/blender/pulls/124619
All the arrays in the new Action DNA are named singularly (`layer_array`,
`fcurve_array` etc). `channelbags_array` was the only plural one. For
consistency, this is now also singular.
Note that this is a backward-incompatible change, and will effectively
erase all animation from layered Actions. No effort is taken to handle this
rename as the feature is still in its experimental phase.
Action and slot assignments are not affected by this change.
Apart from the lack of backward compatibility, no functional changes are
expected.
Pull Request: https://projects.blender.org/blender/blender/pulls/124768
