This was caused by a use after free. The issue was that the motion steps
were successfully gathered but failling the last vertex count check,
discarding the copied VBOs but not removing the reference of the first
VBO which was passing the test.
When the displacement space is set to SHD_SPACE_WORLD, the GLSL method
"node_displacement_world" is used instead of the "node_displacement_object" method. The two GLSL methods:
```
void node_displacement_object(
float height, float midlevel, float scale, vec3 N, mat4 obmat, out vec3 result)
{
N = (vec4(N, 0.0) * obmat).xyz;
result = (height - midlevel) * scale * normalize(N);
result = (obmat * vec4(result, 0.0)).xyz;
}
void node_displacement_world(float height, float midlevel, float scale, vec3 N, out vec3 result)
{
result = (height - midlevel) * scale * normalize(N);
}
```
In contrast to the "node_displacement_object" method, the "node_displacement_world"
does not require an "obmat" parameter. Attempting to still pass "GPU_builtin(GPU_OBJECT_MATRIX)"
as additional parameter will result in a memory leak. The "GPUNodeLink" allocated in
the "GPU_builtin" method will never get released.
Fixes T83941 Memory leak when using the Displacement shader node in Eevee with the displacement
space set to "World Space"
This change removes copy-on-write operations from ID nodes which do not
need copy-on-write.
Should be no functional changes, as before the copy-on-write operation
would do nothing for those nodes anyway.
Building IDs which are not covered by copy-on-write process was not
implemented, which was causing parameters block not present, and, hence
causing crashes in areas which expected parameters to present.
First part of this change is related on making it so Copy-on-Write is
optional for ID nodes in the dependency graph.
Second part is related on using a generic builder for all ID types
which were not covered by Copy-on-Write before.
The final part is related on making it so build_id() is properly
handling ParticleSettings and Grease Pencil Data. Before they were not
covered there at all, and they need special handling because they do
have own build functions.
Not sure it worth trying to split those parts, as they are related to
each other and are not really possible to be tested standalone. Open
for a second opinion though.
Possible nut-tightening is to re-organize build_id() function so
that every branch does return and have an assert at the end, so that
missing ID type in the switch statement is easier to spot even when
using compilers which do not report missing switch cases.
As for question "why not use default" the answer is: to make it more
explicit and clear what is a decision when adding new ID types. We do
not want to quietly fall-back to a non-copy-on-write case for a newly
added ID types.
Differential Revision: https://developer.blender.org/D10075
Previously this relied on the dependency graph to detect changes in the screen
datablock, which would then notify the renderers. This was rather indirect an
not even really by design. Instead use notifiers to tag specific 3D viewports
to be updated.
Includes changes to BKE_scene_get_depsgraph to accept a const Scene pointer.
Testing if this works correctly requires adding back commits 81d444c and 088904d,
since those have been temporarily reverted.
Differential Revision: https://developer.blender.org/D10235
The grease pencil merge depth shader is designed to only work correctly
in octographic mode. The uv coordinates used `noperspective` attribute.
Somehow this doesn't lead to render artifacts on most platforms and was
only detected on OSX + AMD cards.
This fix would calculate the uv coordinate inside the fragment shader
and isn't passed along from the vertex shader.
Thanks to Sebastián Barschkis for providing the hardware and time and
Clément Foucault for helping out with the final fix.
This commit fixes T84588's second issue. The `particle` parameter was
declared optional in the Python API of `bpy.types.ParticleSystem.uv_on_emitter`
due to a typo in the RNA definition. This commit marks it as required.
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D10127
Introduced with swapped axis in rB0d67eb277f9b.
Similar was fixed for the translate gizmo in rB567212c3434a.
Now do the same for scaling as well.
Maniphest Tasks: T85169
Differential Revision: https://developer.blender.org/D10245
When this behavior was added it made sense,
since then show_edges has changed to make edge-display more subtle
(see 1a4b60c30d) instead of removing
edge-selection display entirely.
For jpeg, an image.resolution was always based on the default 72dpi, now
read the pixel density from the jpeg_decompress_struct, convert
according to unit and store in IMBuf's ppm.
Not 100% sure of all implications tbh., files I have checked seem to work
as expected now in the context of the report.
Maniphest Tasks: T84661
Differential Revision: https://developer.blender.org/D10166
If a force field was of type "Texture", any changes of that texture (e.g.
its type - as reported in T85139 - or also its properties) were not
properly updating rigid bodies and particle systems.
Now ensure that texture is actually in the depsgraph and set up relation
accordingly.
Also fixes T75198.
Maniphest Tasks: T85139
Differential Revision: https://developer.blender.org/D10234
- Grey out in wire/xray display.
- Expand the description for when this is used.
While this is working, the intended behavior wasn't clear,
address T85177.
The outline for the active modifier was abusing the property search
match theme color, as noted in a comment. This commit adds a new
theme color in RNA specifically for the active modifier outline.
This enum is only used by the node. So it does not need to be declared outside
the scope of its function.
Originally I thought this may be relevant to the collection info node as well,
but the patch for it is defining its own enums.
The issue was that the `offset` in `dst_span[offset]` was out of bounds
when the domain size is 0. The fix is to simply skip copying attributes
in that case.
This reverts commit 76fd41e9db. This should have
been reverted along with 0f95f51361, since this
change by itself is causing crashes when the depsgraph accesses a non-existent
copy-on-write component.
Ref T84717
For buttons that edit array properties, the soft min/max and slider ranges are
based on the range of all values in the array. However for alpha this does not
make much sense, the only reasonable range is 0..1 even when there are RGB
values larger than 1. So treat alpha as an individual property.
Now the icon's rect are drawn wider than it should be, and with
overlapping, probably in order to compensate for the icon's offsets
inside the rect. The solution is to draw icon's rect of the correct
size and center the icon itself.
And make the hotspot exactly match the icon's rect. The last/right
button's hotspot also covers the extra padding on the right.
Differential Revision: https://developer.blender.org/D9936
Reviewed by: Julian Eisel
Since 216d78687d the depth function
(glDepthFunc) was left in an undefined state for drawing callbacks that
use the `bgl` module.
This meant enabling depth-test from Python's bgl module also needed
to set the depth function (which previously wasn't necessary).
Set the depth function as part of GPU_bgl_start
Gaussian filter with a too large kernel doesn't make much sense.
This commit caps the Gaussian function radius to MAX_GAUSSTAB_RADIUS.
Reviewed By: sergey
Maniphest Tasks: T84512
Differential Revision: https://developer.blender.org/D10122
Technically, the crash was caused by revert which happened in
rBcd24712c2c5: it reverted some code which is essential for
rB76fd41e9db1.
Bring back the essential code for the removal of un-needed
Copy-on-Write operations, so that the crash doesn't happen.
What was causing the crash is the ID tag assuming Copy-on-Write
operation always exists.
Activating buttons was running too early, before the popup was
positioned on the screen,
causing data-selectors to be positioned incorrectly.
As the early button activation isn't needed, remove this call.
Ensure this doesn't happen again with an assert.
Disable key-accelerators for key-repeat events.
When a key was held it could open the menu and activate the menu
item associated with that key.
With the RMB select option: edit-meshes & edge-selection caused
holding W to open & activate "Edge Bevel Weight".
When using the viewport render animation and saving to an 8 bit
image/movie file, the color management is applied twice. Once on the GPU
and once when saving to disk. Removing this option currently leads to
incorrect Scene strip rendering so needs more research.
When uv editing objects that share the same mesh only the selection
state can get confused. The cause is that the UV editor uses a
particular order of objects and store its state in the first object of a
mesh it hasn't handled. During drawing this state is updated into the
GPU buffers. In the case of linked meshes it can happen that the GPU
buffers are updated based on the object that does not have the correct
selection state making th selection VBOs incorrect.
This patch adds a work around that uses the order that the UV editor is
also using so the GPU buffers are built with the right data.
Revert "Fix T83411: Crash when using a workspace/layout data path in a driver"
The fix for the crash exposed design violation in the viewport shading updates,
which is for some reason relying on dependency graph tag of interface data.
The viewport module did not respond to the issue in 2 weeks, and the architect
considered missing update for multiple users a more serious issue than a crash
in a very specific case.
This reverts commit 0f95f51361.
BLI_strncpy_wchar_from_utf8 made the assumption that
wchar_t is UTF-32 bit regardless of environment, while
this holds true on both mac and linux, on windows
wchar_t is actually actually UTF-16.
This resulted in the upper 16 bits being dropped from
from some string conversions and prevented blender
from starting when installed in a path with unicode
code-points over 0xffff.
There was also a fair bit of code duplication between
BLI_strncpy_wchar_from_utf8 and BLI_str_utf8_as_unicode_and_size
this change essentially removes all logic from
BLI_strncpy_wchar_from_utf8 and calls the right function
for the right environment.
Reviewed By: brecht , Robert Guetzkow
Differential Revision: https://developer.blender.org/D9822
An extra `p->` was added for some reason in a recent commit.
Additionally, there's no reason for the flag to be kept as a pointer,
so just dereference it at the start of both functions.
When the drawing plane was set to view and the user would pan their
camera sideways, the drawing would be more and more distorted and
projected further back.
The fix projects the strokes to view when the user is looking through
the camera or has the drawing plane set to view.
Reviewed By: antoniov
Maniphest Tasks: T85082
Differential Revision: https://developer.blender.org/D10213
Use transform matrices instead of loc, rot, scale variables to store instance transforms.
Reviewed By: JacquesLucke
Differential Revision: http://developer.blender.org/D10211
`seq->tmp` was left in invalid state after
`SEQ_sequence_base_dupli_recursive()` which is run by
`scene_copy_data()`
Ensure all strips have `seq->tmp` set to `NULL` before splitting strips.
Meta strip cache is not invalidated if strip inside of meta strip is
invalidated.
Find all metastrips that contain invalidated strip and invalidate them
recursively.
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D10192
Workbench engine used active camera to setup DOF effect even when
camera was overridden.
Store camera override in `WORKBENCH_PrivateData` and use it in
`workbench_dof_engine_init()`
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D9952
This improves the fix introduced in rBdc8b31af578bfcf9f77dbce6d7b7da1006a3c8c1.
While it did not result in other bugs afaik, I thought it would be
good to make the fix more specific, in case some other code did
expect `layer->data` to be null. This updated fix checks for the
very specific case the fix was intended for.
Change `"active_id"` to `"id"` so that tab completion in the asset browser
context returns the correct properties.
055ef5df61 renamed the `active_id` property
to `id`, but `dir(thecontext)` still returned `"active_id"`.
