|Operating system:||Windows-10 64 Bits|
|Graphics card:||Radeon (TM) RX 480 Graphics ATI Technologies|
- User Since
- Oct 7 2012, 2:37 AM (454 w, 4 d)
- Rebase on master
The crash happens in this function: https://developer.blender.org/diffusion/B/browse/master/source/blender/editors/sculpt_paint/sculpt_smooth.c$213
But we still don't know what the source of the problem is.
You still don't have enough information for us to investigate the problem.
If the bug is caused by the addon "extra objects", I suggest creating another report.
Be sure to carefully follow the instructions and provide system information, Blender version, and a .blend file with exact steps to reproduce the problem.
All crash-logs show that the crash happens in do_enhance_details_brush_task_cb_ex.
But without reproducible steps, unfortunately we can't triage this as a complete bug report.
Thanks for the report, I can confirm.
The problem is not "preview set to zero" or "sculpt & render set to two", the problem with the file is that the modifier is disabled for viewport:
It's the same problem described in T89383: Dont' working Blender for apple m1 at frist time.
The crash apparently occurs in BLT_lang_set which is involved in the translation of Blender for different languages.
Has this worked before?
Can you test the latest blender ARM builds? https://builder.blender.org/download/daily/
Thanks for the report, but unfortunately this can't be triaged as a complete bug report.
Thanks for the report, but as informed, this is by design. So closing.
For user feedback and feature requests you can see: https://wiki.blender.org/wiki/Communication/Contact#User_Feedback_and_Requests
The undo system has more to do with IDs than with modifiers, so the Core module is, in my opinion, the most suitable.
Tue, Jun 22
This is not really a bug, but a design issue.
We could solve this by supporting units in the proportional editing value:
diff --git a/source/blender/editors/transform/transform_mode_translate.c b/source/blender/editors/transform/transform_mode_translate.c index 3088f6a7776..8f7ef4ae4c2 100644 --- a/source/blender/editors/transform/transform_mode_translate.c +++ b/source/blender/editors/transform/transform_mode_translate.c @@ -201,8 +201,19 @@ static void headerTranslation(TransInfo *t, const float vec, char str[UI_MAX_ }
NaN coordinates indicate corrupted mesh that can trigger serious problems.
So I'm setting this report as having high priority.
- Resolve TODO in BKE_MESH_BATCH_DIRTY_DEFORM
- Cleanup: Remove redundant assignments of open_node
- Rename GEOMETRY_NO_COW --> GEOMETRY_SHARED
- Fix typos
- Rename _TAG_DIRTY_ --> _UPDATE_
- Rebase master
Mon, Jun 21
- Rebase on master
- Free buffer at the end of the operation
A similar solution is already in master (rBb11a463e4fcd)
I can confirm the memory usage spikes, but I couldn't find the source of the problem (I'm having issues with the latest Visual Studio update).
Maybe it's a limitation, but I'm confirming it as a bug for now.
Overall looks good. Just a few opinions on the style:
D11599 was postponed in favor of this one.
They are not mutually exclusive.
That one will be edited according to the decision made here.
I will be satisfied whichever path is chosen.
Committed a different solution.
Fri, Jun 18
I can confirm.
This is because bpy_class_call is being called like recursively and so rna_disallow_writes is reset ahead of time.
How to reproduce the problem in the attached file. Enabling and disabling the modifier doesn't change memory that much.
I had an idea that might be worth considering.
Based on D11599, (which "solved" the problem by creating a GEOMETRY_NO_COW component), my idea is to create a component called GEOMETRY_SHARED and an ID_RECALC_GEOMETRY_SHARED flag.
The name "SHARED" indicates that only parameters shared on different evaluated ids have been modified.
And since they're shared, there's no need to COW.
I think it's best to wait for a decision on the D11337: Depsgraph: remove redundant mesh data duplication in edit-mode first.
With that patch, it will no longer be necessary to create the new component (GEOMETRY_NO_COW).
The idea is good and it really makes a difference.
The flush flags and assert make things safe.
Thanks for the report, but build problems are off-topic on the bug tracker (as explained in the submission form).
Good to know :)
If issues creep up again and you find clear instructions on how to reproduce it from scratch, feel free to open a fresh report, thx!
What does not work?
Cylinder is not seen by the camera and the cube's shadow does not appear as expected either.
From the debug output, the error is not in the gpu but in VCRUNTIME140.dll
VCRUNTIME140.dll can indicate a lot of things, so we don't have enough information to know what's going on.
Please try the latest daily build: https://builder.blender.org/download/
Go to File → Defaults → Load Factory Settings and see if you still can reproduce this issue.
Thu, Jun 17
In fact the use of PROP_UNIT_LENGTH as subtype doesn't seem right, (even more so with those warnings).
The other subtypes listed must also be wrong.
Thanks for the report. Unfortunately the scenario described is too time consuming for us to track down, we require the bug reporter to narrow down the problem.
Is it necessary to create an "image strip" with 240 images?
If there are many steps to reproduce the problem, please attach a .blend file saved right before the bug occurs, so only a few steps are needed to trigger the bug.
You mention F-curves but the images are from the timeline editor.
Timeline and Dopsheet share the same bl_space_type, 'DOPESHEET_EDITOR'.
In the Dopsheet, the sidebar is used in the context of GPENCIL when an object of type GPENCIL is selected.
Wed, Jun 16
In fact, in pose mode, the position of each bone (in td->loc) is relative to the rest position, and the actual real position is td->center. Although it works in some modes, this is not the convention of the transform operator.
Deserves an investigation.
For BMesh we have BM_iter_parallel which works even with elem_table_dirty dirty.
I wonder if it could be used here (and what would be the impact on performance compared to BLI_task_parallel_range).
It's always nice to see the transform code being simplified and it makes sense to move some updates to the internals of the sequencer.
The structure of the transform code hasn't changed and I confess that I don't touch the Sequencer's transform much (Is it just the "Translate" and "Sequence Slide" that affects this area?)
Tue, Jun 15
I saw now that the patch did not match the description!
Maybe the name mat_offsets.tri could be more descriptive like mat_offsets.tri_len, but it's not really a problem.
It is no longer necessary.
- Fix wrong use of GPU_BATCH_CLEAR_SAFE
- Cleanup: use DRWBatchFlag type
Mon, Jun 14
- Isolate the operation responsible for ID_RECALC_GEOMETRY_DEFORM in its own component