Fixes#108730
The call to `CustomData_interp` also interpolated the face corner
topology attributes, which broke the mesh completely. To fix this,
avoid relying on the CustomData API and move the attribute
interpolation to use the attribute API instead. This change also
makes it trivial to support all attribute types in the future.
Also, only retrieve the closest vertex once, instead of doing it
again for every attribute. Combined with building the topology
maps in parallel this approximately doubled the performance
of color attribute propagation in a test with 2 million vertices
and 8 color attributes on both domains (63 to 34 ms).
Pull Request: https://projects.blender.org/blender/blender/pulls/108773
Ensure buffer read operations wait for pending GPU work to
complete. This is only required for unit tests, but required to
ensure correct results in all cases. Also adds support for staging
buffer if GPU memory type is private.
Buffer clearing routine added to support values above 1 byte.
The previous buffer clearing functionality using fillBuffer writes
values to each byte, rather than the full 32 bit uint used by the
GPU API.
Authored by Apple: Michael Parkin-White
Pull Request: https://projects.blender.org/blender/blender/pulls/108757
The graph editor has the function to hide the display of FCurves and to only draw keyframes on selected curves.
Both options were not respected when executing the frame jump operator,
meaning it would jump to potentially hidden keyframes.
This patch adds a new operator that is specific to the Graph Editor and that respects those features.
The only possible issue that could arise from this is a slight confusion for the user
why they behave differently between editors.
Especially with only the Graph Editor and the 3D view open.
In general I think it's an improvement and follows the "you can only modify what you see" mantra in blender
Also
Resolves#97701
Pull Request: https://projects.blender.org/blender/blender/pulls/108549
Issue here was that Quick Favorites use a property's **identifier** when
adding [which is the bare name without brackets etc. from id properties]
but when spawning the actual menu, `RNA_struct_find_property` expects
the identifier to already include the brackets to know these are id
properties (later on in `screen_user_menu_draw`).
So to solve this, now include the needed syntax when storing the
`bUserMenuItem_Prop` identfier.
Seems the quickest way to append the needed characters is using
`RNA_path_property_py` (not sure if there are better ways to do this).
Also note that we (need to) ignore the actual array index constructing
the string [always pass -1 here] since the index is handled separately [I
tested boolean arrays and these work].
Pull Request: https://projects.blender.org/blender/blender/pulls/108713
During uv unwrapping and uv packing, certain floating point algorithms
have extreme sensitivity to round-off errors. These can produce very
different layouts even when given inputs which are only slightly different.
The root cause is that the two main types of CPUs used to run Blender,
namely x86 and Apple Silicon, produce slightly different results on some
math functions, including `sinf()`, `cosf()` and `atan2f()`.
* sinf(0.8960554599761962890625) = 0.780868828296661376953125 (Intel i7)
* sinf(0.8960554599761962890625) = 0.78086888790130615234375 (Apple M1)
This fix, and others that came before it [0], improve accuracy by using
double-precision to hide the differences between the CPUs.
[0] e.g. 0eba9e41bf
[1] See also #107829
Caused by 98bfa8d458
Occlusion must consider several factors:
- The `Viewport Display` > `Display As` setting of the object.
- The X-Ray mode of the scene.
- Whether only Snap to Face is being used.
- Whether Snap is set to Cage (always transparent) in edit mode.
- Whether the object is hidden.
In the case of Cage, the final geometry should be utilized for occlusion.
However, to prevent excessive and risky modifications, it was followed
the same approach as before and verified the `Display As` setting of
the object.
In particular, the root was not sticking to mesh with hair dynamics.
Reason for this is missing HairKey weight upon conversion (everything
had zero weight). This then can not be corrected (since weight tool does
not allow to paint on roots).
So now, give the hair a default 1.0 to 0.0 weight gradient upon
conversion [which is in line with the default weight distribution for a
hair particle system].
Pull Request: https://projects.blender.org/blender/blender/pulls/108651
Part of the checks to see if translating a UI string is allowed or not
is that current thread is the main one. This was a requirement years ago
of the Boost backend for translations.
Not sure whether this can be removed or not now, but it needs to be
carefully checked, done it as its own commit, and not in a beta release
branch. ;)
Sorry for the noise, totally missed this during the patch review
yesterday.
Pasting numerical array buttons happens with `Ctrl + Alt + V`.
Holding `Alt` also triggers uiSelectContext, so having other nodes/
objects etc. selected while doing this would try to copy the pasted
values back to other objects (possibly to the ones you pasted from) and
that happens relative to the original value, so the value actually
changes.
NOTE: the `Ctrl + Alt + V` shortcut can also be used on non-array buttons, so was an issue for them as well.
To prevent the "copy-to-selected" behavior, refine the `IS_ALLSELECT_EVENT` macro to be more specific.
Pull Request: https://projects.blender.org/blender/blender/pulls/108270
References to data-blocks in a material were stored in-memory and could
crash if the data-blocks referenced by the material no longer existed
when pasting.
Resolve by using a blend-file for material copy/paste, matching how the
clipboard works in the 3D view-port.
Currently there is no support for including indirectly linked
data-blocks when pasting the material. Instead, data-blocks are restored
by name, by inspecting the current file.
This also fixes a crash where the `SpaceNode::nodetree` could point to
freed memory when pasting a material.
Ref !108496.
Includes contributions by @mont29.
---
Fix back-ported to main [0], including fix [1].
[0]: 5177e2f20b
[1]: 64aa96d421
`hair_out_mesh` and `hair_in_mesh` implicitly share edges.
In `hair_create_input_mesh()`, edge data of `hair_in_mesh` needs to be
updated and therefore are copied to a new location. In the subsequent
frames, `psys->clmd->clothObject->edges` won't be updated and point to
freed memory block. Therefore, Blender crashes.
By freeing `hair_out_mesh` first,
1. in `hair_create_input_mesh(),` at least edge data copying is avoided
2. `psys->clmd->clothObject->edges` always points to correct memory
However, since it's possible that similar situation will happen again
by adding another strong user to the same `CustomData` in the future,
it is safer to update `psys->clmd->clothObject->edges` for every frame.
Pull Request: https://projects.blender.org/blender/blender/pulls/108480
The "Fill" message can be either a noun or a verb. This commit
disambiguates the verb usages for translation through various
translation contexts.
The more involved change is in the generation of keymaps from paint
modes. By default, the enums defining brush names are in the default
context, but this commit changes the ones including a "Fill" item to
"Brush". In order to get the same contexts in the keymap, we introduce
a specific function in `paint.cc` to return the appropriate context
depending on the tool.
Issue reported by Gabriel Gazzán (@GabrielGazzan) in #43295.
Pull Request: https://projects.blender.org/blender/blender/pulls/108561
This came up in #108096
Reason this fails is the `ui_but_has_array_value` check [which depends
on a property subtype that supports arrays]. The default `VectorBuilder`
has a `PropertySubType` of `PROP_NONE` though, so one possibility
would be to change this to `PROP_XYZ`.
However, RNA should know much better which RNA property buttons
have arrays than the UI code, so use RNA_property_array_check now
(instead of checking particular UI property subtypes).
Pull Request: https://projects.blender.org/blender/blender/pulls/108349
When adding a texture paint slot to an object, the object could have no
material, this patch handles that by checking the material first in
`default_paint_slot_color_get`, if material is null, then it will return a
fallback default color so the operator can proceed normally.
Pull Request: https://projects.blender.org/blender/blender/pulls/108592
- Paste flipped created invalid paths when the size of the name changed
left -> right: lost the 't' in right.
right -> left: lost the entire end of the RNA path.
- Correct the fixed buffer size as it may include escaped characters.
Fix a mistake in commit 2ce5fc4a3e that caused a crash when detaching
node links from input sockets.
When a link is detached from an input socket, `nodeRemLink` nulls the
`link` pointer of the socket.
So before the next update inputs are linked but don't have a valid `link`
pointer causing the crash, when trying to access the link in
`std_node_socket_draw`.
The introduced check avoids the crash and is more correct since it
doesn't just check one link for multi-input sockets.
Pull Request: https://projects.blender.org/blender/blender/pulls/108623
- "Invalid" in transformation messages.
- For three messages, translation occured after a string
- concatenation, so the full message was not found.
Instead, translate a format pattern and format it afterwards.
- Alembic errors when there is an import type mismatch.
Pull Request: https://projects.blender.org/blender/blender/pulls/108212
Code was plainfully buggy, early-out check in
`BKE_undosys_stack_limit_steps_and_memory` was plainfully wrong.
Also added some more logging for memory limiting code.
Produce optimal layouts for `n` squares, where n == 11, 18, 19 and 26.
With thanks:
* Walter Trump
* Pertti Hamalainen
* Robert Wainwright
* Erich Friedman
Prevent make links operator from creating links to sockets that are
already linked to a muted link.
The `SOCK_IS_LINKED` flag is used to check if there already is a link
connecting to the socket but when the link is muted, the flag wasn't set
leading to issues in parts of the code that used the flag to check
for any type of connected link.
This commit now also sets `SOCK_IS_LINKED` when links are muted and
adds an additional check in places where different behavior is expected
for muted links.
Pull Request: https://projects.blender.org/blender/blender/pulls/108375
Fix node link insertion during transform not working properly for
rotation and scale.
Inserting nodes by rotating or scaling...
* ...didn't offset the attached nodes.
* ...could lead to unfreed memory.
This commit fixes that by always calling the `NODE_OT_insert_offset`
operator at the end of the node transform operator rather than having
to explicitly append it into a macro operator for each transform
operator.
Pull Request: https://projects.blender.org/blender/blender/pulls/107460
The solution using `reverse_index_array` didn't work because it lost the
order that the corners were processed in (the order around the vertex).
This is important when setting custom normals because the process
sets sharp edges when the normal of the current and previous corner
is too different.
Because of how the Clip Editor managed region state for different view
modes before region polls were introduced (fa0f295b53), regions could be
written to files in a state that's not valid for displaying them. State
for clip editors written to files is corrected with versioning now.
The user count decrement and unassigning the action were split
between two different conditionals, which weren't guaranteed
to both execute/not-execute together.
Objects were excluded from the ID operation, as originally they had
their own entry in their own menu. Joys of nightmare tools handling in
the Outliner...
Painting weights in Grease Pencil with the auto-normalize option enabled,
gave unpredictable results when vertex groups with a weight of zero
were involved.
This patch resolves the issue. Vertex weights of zero are excluded from
auto-normalization. Which is a better fit for the Grease Pencil workflow,
where weights are mostly painted from scratch since the 'parent to
armature with automatic weights' operator doesn't give good results.
Pull Request: https://projects.blender.org/blender/blender/pulls/108524
Steps to reproduce were:
- Open File Browser, set it to Thumbnails display mode, disable filters
- Navigate to a directory with documents that have no preview
- Change Preferences > Themes > User Interface > Icon Colors > File
Folders to light and dark colors
- See how the type icons on top of the white document icon image get
changed unexpectedly. Only type icons drawn on top of folder icon
images should change.
When drawing the document type icons on top of the folder or document
image in the center of thumbnails, they would always use the folder icon
color (Preferences > Themes > User Interface > Icon Colors > File
Folders) to determine if they should be drawn with a light or a dark
color. It should only use that when drawing on top of a folder icon
image, and the color for document icon images otherwise.
When doing the second undo, the image wasn't properly marked for full
update.
There reason was that in some cases the partial_updater was
totally reconstructed: `first_changeset_id` and `last_changeset_id` were both 0.
While the `user_imp->last_changeset_id` was still its last value (e.g., 3).
The fix is to have the partial updater validator to check for the
last change set as well (it was only checking for the first change set).
This way we cover both scenarios when user_imp->last_changeset_id is out
of the range of the partial update history.
Pull Request: https://projects.blender.org/blender/blender/pulls/108533
Pasting MTex referenced Texture & Object data-blocks without any
check for their validity.
This isn't ideal as it doesn't handle re-allocated ID's,
it just prevents a crash referencing freed data-blocks.
Excessive memory pool bloating could occur for certain workloads in
Metal. Particularly those which continuously allocate increasingly
large buffers with minimal re-use of existing buffers.
New logic added to the memory pool flushes old buffers if they
have not been used for a set period of time. Timing is calibrated
against system resources and overall memory pressure.
Metal memory pressure will run higher than OpenGL, however,
this is an active decision to provide significant performance
improvements for scenarios which allocate lots of memory and
for keeping frames queued in flight without stalling on pending
GPU work.
Authored by Apple: Michael Parkin-White
Pull Request: https://projects.blender.org/blender/blender/pulls/108083
Certain textures require explicit HOST READ support for reading
data back to CPU. Given the compute based texture read path may
be used for this, if format conversion is required, this flag should
also imply general texture read support.
This resolves a bug with GreasePencil fill which relies on this.
Authored by Apple: Michael Parkin-White
Pull Request: https://projects.blender.org/blender/blender/pulls/108463
