Quite obscure issue, seemingly caused by the fact that HIP-RT is
passing a large (about 20 MB) global array to a different library
(HIP driver, via hipModuleLoadData). Having global variables of
such size seems to be always problematic as they can not be stored
on stack and, possibly, extra mapping is involved here. It is not
clear whether it is a quirk of the HIP driver, or Linux, or, maybe,
something completely different.
It is possible to work-around the problem by making a temporary
copy of data on heap memory and pass it to the hipModuleLoadData().
This is how other areas are dealing with modules in Blender.
This change contains patch against HIP-RT and the new HIP-RT
library compiled with the patch. It seems to fix the problem
reported in the report.
This change does not resolve OIDN on HIP GPU which seems to have
the same issue. However, it is not a recent regression and the
bug with OIDN GPU denoising can be reproduced using Blender 4.3.
Pull Request: https://projects.blender.org/blender/blender/pulls/135403
Caused by 6d832ee5b3
The bone names were flipped correctly, however, we were missing the
actual FCurve BezTriple flipping.
This was caused by passing the wrong FCurve into
`do_curve_mirror_flippping` -- this function is checking for existence
of the FCurve in the `KeyframeCopyBuffer`, so we gotta pass
`fcurve_in_copy_buffer` (instead of the new fcu).
Pull Request: https://projects.blender.org/blender/blender/pulls/135381
Caused by 40ac21e5a5 [does not remember/resore the previous ScrArea &
ARegion anymore].
Without this, the operator reporting might get confused by using the
wrong area/region, so added back
NOTE: I tried to just notify `NC_SPACE | ND_SPACE_INFO_REPORT` in
`screen_opengl_render_end` (same as in `wm_operator_finished` >
`wm_operator_register`), but to no avail...
Think this is quite good pratice to leave us with the original area/
region anyways though.
Thx @brecht for improvements (restoring in `screen_opengl_render_init`
already, also taking care of restoring in some early out cases)
Co-authored-by: Brecht Van Lommel <brecht@blender.org>
Pull Request: https://projects.blender.org/blender/blender/pulls/135394
For this to work, we have to take into account handles in TransVert as
well.
Behavior is the same as in legacy curves (meaning that if the control
point itself is selected, the handles are ignores).
Part of #133448 (same thing for grease pencil, which I plan to also make
part of TransVert, but better solve for Curves first, so we can share
code here).
Pull Request: https://projects.blender.org/blender/blender/pulls/134945
The compositor backdrop is still visible even after a new completely
different file is loaded. This is because compositor contexts cache
things like the backdrop, and contexts are stored on interactive
compositor renders, which are not freed until Blender is closed. To fix
this, we cleanup the interactive compositor renders that were created
upon file load.
Pull Request: https://projects.blender.org/blender/blender/pulls/135390
Enable VK loader debug messages when running blender_debug_gpu_*
scripts.
Allows easier debugging for users not familiar with using terminal.
It won't log anything when using OpenGL.
Vulkan loading logging can be big (1000 lines) but gives a lot of insight
how vulkan and related applications are configured on the user system
Sadly it does contains actual paths but think it is OK as the user is still
responsible for uploading it in the tracker.
An example of a log has been attached.
See #135327.
Pull Request: https://projects.blender.org/blender/blender/pulls/135399
This commits fixes a type error issue when reporting that a addon was
designed for a version of Blender newer than the current one.
Thanks to @Rakete for the fix.
Ref: !135392
This was caused by the normals not being available (default to
`vec3(0)`) which produced `NaN`s down the line after the
unsafe normalize. Changing to `safe_normalize` fixes the issue.
Add a new function `bke::animdata::prop_is_animated()` that returns
whether an RNA path + array index is animated by anything. This covers
the assigned Action, NLA Action strips, and drivers.
This function is now used in the determination whether the visibility
of a GreasePencil layer is animated.
The easiest way to implement this with the existing F-Curve-visiting
logic was to call `adt_apply_all_fcurves_cb()`. However, that function
did not allow the callback function to signal "stop looping, I found
what was I was looking for", so I extended it to do just that. I don't
expect the extra conditions to significantly slow down the other uses,
as the branch predictor will very likely optimise for the "returning
true" case for those calls that simply visit everything.
Pull Request: https://projects.blender.org/blender/blender/pulls/135317
The issue was that during playback, the evaluated Grease Pencil ID
is not being copied again for re-evaluation. This meant that the
adjustments get re-applied over and over again.
Ideally, we would restore the drawing pointers before modifiying them
again, but this is a bit trickier to implement.
To fix this issue for now, we do the layer adjustments during object
data evaluation. Using the geometry set for modifier evaluation
means that we copy the data and apply the adjustments for every
object.
In the future this can be optimized to only happen once per
Grease Pencil ID (in `BKE_grease_pencil_eval_geometry`).
Pull Request: https://projects.blender.org/blender/blender/pulls/135387
If a user wants to expand the "bug fixes per release" script with
new classifications, it requires them searching for and expanding a few
lists spread through out the code.
This annoying for anyone not familiar with the code, so refactor these
lists out to the "Constants" section to make it easier to see and
expands.
The transform gizmo in the sequence editor does not show highlights properly for rotation, when hovering over it or while dragging it.
This patch has no further side effects on operators that use the 2d cage gizmo without rotation such as the crop gizmo and the viewer (transform) gizmo in the compositor.
Pull Request: https://projects.blender.org/blender/blender/pulls/135065
ED_area_newspace correctly changes an area's space when the type and
subtype are set beforehand. But SCREEN_OT_space_type_set_or_cycle, when
changing to a new editor rather than cycling, needs newspace to ignore
the set subtype and use the last-used saved in the space. This
situation cannot be tested for in newspace. Instead have the operator
pass -1 as subtype to signal this intended behavior.
Pull Request: https://projects.blender.org/blender/blender/pulls/135367
The line-width is used to calculate the `U.pixelsize` which should only
be used to control the size of lines & points.
Update doc-strings to mention the intended use, remove unused defines.
This is due to accessing member of nullptr Asset. After deleting the
action id from asset library, `CTX_wm_asset` won't return value because
force reset for files was tagged (`FL_FORCE_RESET_MAIN_FILES`).
Instead of updating library from active asset, use library reference to
clear then read/draw the updated asset library
Pull Request: https://projects.blender.org/blender/blender/pulls/135219
Patch #132957 added helpful UI feedback when attempting to rotate or
scale an object whose transforms are set to "only affect locations."
But it neglected to check for invalid trackpad rotations that are
affected the same way. This patch renames the original `HLP_ERROR` to
`HLP_ERROR_DASH` to distinguish it from a `HLP_ERROR` transform cursor
with no dashed line present, which matches trackball's normal state.
Pull Request: https://projects.blender.org/blender/blender/pulls/134653
While joining and docking areas there is hint shown near the mouse that
describes the potential operation. But it is cut off if your mouse is
at the extreme right or bottom edge of the window. This PR just clamps
the position so it is always visible.
Co-authored-by: Jonas Holzman <jonas@holzman.fr>
Pull Request: https://projects.blender.org/blender/blender/pulls/135211
Various fixes in the HIP-RT BVH building related on making sure
curves motion blur is supported and is working correctly, as well
as properly handle motion pass configuration when path tracing is
to ignore motion blur (and instead write vector pass).
This PR contains #134797 with fixes needed to fully finish it:
moving commits from that PR here made it easier to ensure all
moving parts are tested without mental overhead.
Fixes#134510
Co-authored-by: Sahar A. Kashi <sahar.alipourkashi@amd.com>
Co-authored-by: Sergey Sharybin <sergey@blender.org>
Co-authored-by: Brecht Van Lommel <brecht@blender.org>
Pull Request: https://projects.blender.org/blender/blender/pulls/135125
The operator refused to pack libraries with absolute paths (wasnt the
case in its original implementation 16411da41e, but was added in
129fb516f4 -- for the reason of preventing "bad things happen on
unpacking" without an explanation of what these exactly are). It did so
by cancelling as soon as **one** library with ab absolute path was
found.
Now with the introduction of essential assets, we have those absolute
path linkages more or less "by default" as soon as e.g. a brush is
used, so the operator is more or less unusable now. NOTE: these absolute
essential asset library paths seem to be converted to relative on save?
Upon reload, these are then gone... (might be another hint for an
alternative fix, see below)
By "bad things happen on unpacking" I would assume the scenario of
folders being created in unwanted locations (e.g. when moving from one
OS to another), but the same thing is also true for packing **files**
instead of libraries (there, absolute paths are allowed, and unpacking
in original locations can equally fail or create folder structures that
are "unexpected"). NOTE: for files though we have the choice of
unpacking to a relative folder (which wouldnt really be possible since
libraries can be nested and we would have to correct paths all over the
place). NOTE: the chance of creating "unwanted" folder structures with
relative paths might be slimmer, but if you have a lot of "upwards"
parent folders, relative can easily "break" as well.
Possible ways to resolve this:
### [1] skip libraries identified as essentials assets (still cancel on all other absolute paths)
Can check a library path to be contained in `EssentialsAssetLibrary`
`essentials_directory_path`. This would be the safest imo since it is a no-behavior change.
### [2] lift the limitation of absolute paths alltogether
Like mentioned above, things could break with "relative" almost as
easily as with "absolute", there might even be scenarios where
"absolute" is wanted. It is a more behavior-changing fix that we might
explore more after 4.4 is out.
### [3] skip absolute libraries (but continue with non-absolute libraries)
This does change behavior as well (it does not cancel as soon as **one**
library with ab absolute path was found anymore, but the worst case
scenario is that you end up with an "incomplete" file if you really
mixed absolute and realtive lib linking).
This PR implements [1] for now, [2] or [3] can follow for 4.5.
Fixes#134665
Pull Request: https://projects.blender.org/blender/blender/pulls/134839
`ColorManagedDisplaySettings`, `ColorManagedViewSettings`,
`ColorManagedInputColorspaceSettings` were affected since these are used
from multiple places (Scene, File output nodes, ..).
Similar to how we are getting the path for `ImageFormatSettings` (which
pretty much has the distinction logic already), we can do so for the
ColorManage Settings as well (piggbacking on the already correct path to
the `ImageFormatSettings`.
With this, we can also remove these "known failures" from the test
introduced in 4032b853c3.
Pull Request: https://projects.blender.org/blender/blender/pulls/135238
Volumes' "Sequence Mode" recently got the "Sequence" context in
2b3f0d0110. The intention was to disambiguate the "Extend" enum item,
but this caused another item, "Clip" to also get this context. That
one already existed and referred to a sequencer clip.
Use the more specific "Volume" context instead.
Reported by Gabriel Gazzán.
Commit 8ff8fc4859 changed the context for the modes in Merge Nodes
and Change Nodes operators to Nodetree. The enum items were thus
extracted using this context, but the UI Menus were not updated to
reflect this change and a few items were no longer translated.
This commit fixes the issue by using the same translation context for
the menus' drawing.
These have bugs in with the latest HIP-RT and HIP SDK, so just disable them
as we do not expect a fix in time, and rolling back would re-introduce other
bugs. As RDNA1 does not have hardware raytracing, it is also less important
to use HIP-RT.
Note that only RDNA2+ is officially supported by HIP, so these GPUs working
at all is somewhat lucky.
Fix#134979Fix#134978Fix#134975
Pull Request: https://projects.blender.org/blender/blender/pulls/135179
OpenEXR DWA compression in Blender is derived from a more user-friendly
quality slider which has an intuitive range 0 .. 100.
Initially the mapping was done so that the visually lossless JPEG
quality of 97 was mapped to the default DWA compression 45. A point was
made that we should make it so default quality is mapped to the default
compression, following the intent of DWA for rendering and compositing
the main target.
This change adjusts the mapping so that quality of 90 is mapped to DWA
compression 45.
This change relies on the library update to fully utilize the DWA
compression #135037.
This change leads to the difference in the way proxies of EXR images
are generated:
```
DWA compression Size (bytes)
Before the change 750 175,208,243
After the change 225 77,838,827
```
It is worth noting that the DWA compression seemed to be ignored in
the 4.4 branch before this change (this is what the original report is
about, a bit indirectly).
This is measured on the Fabrik Eingang footage converted to EXR. The
absolute value is ptobably not that important, it just shows the
reduction in size. This also leads to a lower quality of the proxy
image, but it is not worse than an actual JPEG proxy: the quality is
set to rather low 50 for the strip proxies.
Ref #134802
Pull Request: https://projects.blender.org/blender/blender/pulls/135103
`IDProperty` of keyitem and operator in UI were not matching in
`wm_keymap_item_find_in_keymap()`. Now fixed by removing explicit
property value from UI menu
Pull Request: https://projects.blender.org/blender/blender/pulls/134724
The `with_smoke` case is handled correctly and checks for reallocation,
but the liquid case was ignored for unknown reasons.
The `with_smoke` distinction was added years ago (c7596cd820)
but for some reason the `with_liquid` case was ignored for reallocation.
Pull Request: https://projects.blender.org/blender/blender/pulls/135247
The code-path for coercing a dictionary to operator/gizmo properties
was being used RNA functions where it's not supported.
Raise a type exception instead of crashing.
