The issue is a combination of following aspects:
- Missing null-pointer check in the image operation, which is probably
why the result was buggy. It is addressed by #123493.
- In certain conditions loading image was wrongfully failing.
The reason for failing to read image were items with a null-pointer
image buffer left by the cache limit enforcer, which was considered
to be an indication of failed load from disk. The reason why the cache
limiter leaves items with null-ptr as an image buffer is kind of a
legacy limitation which was never resolved. Long story short: the
system expects put() to be called on the cache to clear its empty
items.
To solve the original issue of files considered to be unreadable
only set the cache-empty if the image buffer was added empty.
Pull Request: https://projects.blender.org/blender/blender/pulls/123496
The first input of the compositor Mix node determines resolution,
leading to situation when the second input will always be attempted
to be evaluated. If the first input is a longer image sequence than
the second input it leads to a crash.
Do a null-pointer check and return transparent image in this cases,
similar to what the Movie Clip operation is doing.
Pull Request: https://projects.blender.org/blender/blender/pulls/123493
Add back the "Add-ons" preferences, removing add-on logic from
extensions.
- Add support for filtering add-ons by tags
(separate from extension tags).
- Tags now respect the "Only Enabled" option.
- Remove the ability to enable/disable add-ons from extensions.
- Remove add-on preferences from extensions.
- Remove "Legacy" & "Core" prefix from add-on names.
- Remove "Show Legacy Add-ons" filtering option.
Implements design task #122735.
Details:
- Add-on names and descriptions are no longer translated,
since it's impractical to translate text which is mostly
maintained outside of Blender.
- Extensions names have a `[disabled]` suffix when disabled so it's
possible to identify installed but disabled extensions.
- The add-on "type" is shown in the details,
so it's possible to tell the difference between an extension,
a core add-on & a legacy user add-on.
- Icons are also used to differentiate the add-on type.
- User add-on's must be uninstalled from the add-ons section
(matching 4.1 behavior).
- Simplify logic for filtering tags, move into a function.
Primarily the `evaluation_mode` enum prop was incorrectly grouped with
the previous `xform_op_mode` enum causing them to combine in the UI.
Additionally, group the `allow_unicode` option under the Blender Data
sub layout as was intended but got lost in a merge.
Pull Request: https://projects.blender.org/blender/blender/pulls/123513
Make half-size waveforms default in new files and Video Editing template.
They are more space efficient and display more detail at small sizes.
This does not change existing files.
Pull Request: https://projects.blender.org/blender/blender/pulls/123511
Some of the existing colors were hard to read with the new
strips design.
Tried following the concept from 2.83 redesign rationale:
* Same saturation for regular strips.
* Lower saturation for effect strips.
* Tried to reduce the hue shift between certain similar effects.
Other changes:
* Match saturation of all regular strips.
* Reduce value and saturation (mostly value) of color tags so
they are readable in both light and dark text.
* Image: Follow node editor Image node socket color.
* Color: Use the same hue as the color node socket.
* Text: Change it so it doesn’t use the same as Image.
* Sound: Use a greener color, less movie-like blue.
* Scene: Light gray, similar fashion to Collections.
* Other strips had minor adjustments.
Images and details in the pull request.
Pull Request: https://projects.blender.org/blender/blender/pulls/123446
This was caused by the tangent basis computation that
had a threshold that was too noticeable.
Increasing the threshold makes the artifact
unoticeable.
Fix#122949
This storage of the entire mesh for topology-changing operations doesn't
need to be stored in every single undo node. Move it to the data for the
undo step instead. This decreases the size of undo nodes from 3880 to
1816 bytes. That saves about 4MB of memory on a single stroke affecting
most of a 6 million vertex mesh.
I didn't change anything BMesh related here because it's trickier to get
right and not quite as encapsulated. Moving all BMesh undo data out
of `undo::Node` would be a good step though, because only one undo
node is used anyway.
Currently all nodes pushed in a single undo step are expected to have
the same type. Storing the undo type in the undo step itself rather than
duplicating it in every node makes it easier to move some "entire mesh"
data out of the undo nodes and into the undo step, with the goal of
making undo nodes smaller and simpler.
Pushing multiple nodes at the same time helps to reduce the amount
of time spent waiting for threads to unlock while they manipulate the
nodes map, and equalizes the amount of work per thread, since we
can iterate over just the nodes that need data stored. I observed a
2.6% speedup in the benchmark file from #118145 (0.59s to 0.57s).
Instead of counting the size of undo steps as they're being built,
add up the size of all undo nodes as a final step when the undo step
is finished. This is faster because it avoids incrementing the same
size variable from many threads (which also wasn't threadsafe).
I measured a 4% performance improvement in the brush benchmark
file from #118145 (from 0.61s to 0.59s).
Instead of locking for the whole time the undo data is being stored,
only lock while the step's per-node undo node map is being accessed.
This is fine because each PBVH node is only processed by a single thread.
Changing the node vector to not store anything until the undo step is
finalized makes this process a bit simpler because we don't have to build
both the map and the vector at the same time.
Overall this improved the performance of the sculpt brush benchmark
from #118145 by 12%, from 0.68s to 0.61s.
By itself this change doesn't look great since it's mixing up the
responsibilities of different functions. However, it's preferred
since we want to separate geometry undo state and per-node
undo state further in the future. Eventually geometry undo state
should be stored in `StepData` instead.
Sometimes undo data is stored per PBVH node, and sometimes the entire
geometry is stored. Currently these go through the same `undo::push_node`
function, which is confusing and limits improvements we can make to the
pushing of per-node undo data.
In file_browse_invoke, if the path is blank we attempt to use
BKE_appdir_folder_default_or_root to get the user's normal document
root. However we are not ensuring that the directory path has a
trailing slash, which can result in it being misinterpreted as a file
path instead, separating out the last path part. This PR only adds a
call to BLI_path_slash_ensure.
Pull Request: https://projects.blender.org/blender/blender/pulls/123504
This was caused by some pixels having no good neighbors
to get the irradiance from. The weighted sum would
have huge precision issues and the values would blow-out.
Adding an epsilon weight tailored to the report file
to fix this issue.
Fix#123488
The icon for missing media/data-block would be too close to the edges.
Decrease the size so it matches the size of the label and aligns with
the text baseline.
Update the batch specializations compilation to allow using it in an
async way.
The implementation has 2 main limitations:
- Only one batch at a time can be processed, extra batches will be
added to a queue.
- Binding a specialization variant that is still being compiled will fail.
Pull Request: https://projects.blender.org/blender/blender/pulls/123015
With the recent change for labels to be black when not active/selected,
the hardcoded colors of missing media/data-blocks make it hard to read.
- Use a lighter color just like muted strips.
- Makes active/selected muted strips draw labels in full opacity, for consistency.
- Make muted strips a little less gray (50% instead of 80%)
Pull Request: https://projects.blender.org/blender/blender/pulls/123494
