- User Since
- Nov 23 2018, 1:25 AM (8 w, 4 d)
agreed optimising the global undo sounds like a better way. at least i learnt how the undo system works. kind of
- Even if you detect all that, you still can't reverse the operation exactly due to floating point precision errors.
- reverted err_out parameter related changes.
- removed blen_font refactored code
- added an exciting new ERROR_OUT macro for gpu_*
Sun, Jan 20
i started to write some code that could speed up undo by over 9000 times
Sat, Jan 19
ok, thanks for those tips, i'll keep them in mind for next time
Fri, Jan 18
removed parts of diff that weren't supposed to be in this set of changes
Wed, Jan 16
thanks for committing the timestamp functionality
Tue, Jan 15
I couldn't reproduce this on my computer using blender 2.80 (2019-01-12) with the attached file
i've made a fix for this here: https://developer.blender.org/D4201
I think i've found the problem that causes this crash. MOD_surface.c deformVerts, calls BKE_id_free on surmd->mesh. This frees the cached tree used by surmd->bvhtree. surmd->bvhtree is not rebuilt until later, so other threads can access the invalid bvhtree. setting surmd->bvhtree to NULL stops this happening.