addons cause an EXCEPTION_ACCESS_VIOLATION possibly from mesh edit optimizations. #89261

New Issue

MaybeLamia · 2021-06-18T10:06:57+02:00

MaybeLamia commented

2021-06-18 10:06:57 +02:00

System Information
Operating system: Windows-10-10.0.19041-SP0 64 Bits
Graphics card: NVIDIA GeForce GTX 780 Ti/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 466.27

Blender Version
Broken: version: 3.0.0 Alpha, branch: master, commit date: 2021-06-17 22:22, hash: 94c4a9995e
Worked: 2 weeks ago.

Short description of error
Sorry if i can't explain this correctly it's way over my head, some addons that have operations that has something to do with bmesh or BVHtrees are causing an EXCEPTION_ACCESS_VIOLATION.

Exact steps for others to reproduce the error
Just enable an addon and use an operation.

[PolyQuilt ]] or [ https:*github.com/Jrome90/Fast-Loop | Fast-Loop are free for testing just enabling the addon go in to edit mode then press the addon in toolbar should cause an EXCEPTION_ACCESS_VIOLATION for both.

I will also leave a zip file with some crash logs.
Blender-3.0-crashes.zip

**System Information** Operating system: Windows-10-10.0.19041-SP0 64 Bits Graphics card: NVIDIA GeForce GTX 780 Ti/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 466.27 **Blender Version** Broken: version: 3.0.0 Alpha, branch: master, commit date: 2021-06-17 22:22, hash: `94c4a9995e` Worked: 2 weeks ago. **Short description of error** Sorry if i can't explain this correctly it's way over my head, some addons that have operations that has something to do with bmesh or BVHtrees are causing an EXCEPTION_ACCESS_VIOLATION. **Exact steps for others to reproduce the error** Just enable an addon and use an operation. [PolyQuilt ]] or [[ https:*github.com/Jrome90/Fast-Loop | Fast-Loop ](https:*github.com/sakana3/PolyQuilt) are free for testing just enabling the addon go in to edit mode then press the addon in toolbar should cause an EXCEPTION_ACCESS_VIOLATION for both. I will also leave a zip file with some crash logs. [Blender-3.0-crashes.zip](https://archive.blender.org/developer/F10179636/Blender-3.0-crashes.zip)

MaybeLamia commented

2021-06-18 10:06:57 +02:00

Added subscriber: @MaybeLamia

blender-admin commented

2021-06-18 10:06:57 +02:00

blender/blender-addons#89311 was marked as duplicate of this issue

Fen commented

2021-06-18 15:03:48 +02:00

Added subscribers: @ideasman42, @chemicalcrux

Fen commented

2021-06-18 15:03:48 +02:00

Changed status from 'Needs Triage' to: 'Confirmed'

Fen commented

2021-06-18 15:03:48 +02:00

I ran PolyQuilt with AddressSanitizer enabled and got this crash:

==16360==ERROR: AddressSanitizer: access-violation on unknown address 0x000000000000 (pc 0x7ff6f6450f16 bp 0x12d383a9a1d0 sp 0x00aed69fbef0 T0)
==16360==The signal is caused by a READ memory access.
==16360==Hint: address points to the zero page.
    - 0 0x7ff6f6450f15 in BM_mesh_calc_tessellation_ex D:\blender-git\blender\source\blender\bmesh\intern\bmesh_mesh_tessellate.c:274
    - 1 0x7ff6f6450fcb in BM_mesh_calc_tessellation D:\blender-git\blender\source\blender\bmesh\intern\bmesh_mesh_tessellate.c:283
    #2 0x7ff70065fa4c in C_BVHTree_FromBMesh D:\blender-git\blender\source\blender\python\mathutils\mathutils_bvhtree.c:972

So, it's reading a null pointer. Line 274 is:

bm_mesh_calc_tessellation__single_threaded(bm, looptris, params->face_normals);

The only thing that reads here is params->face_normals, so params is probably null. The function it's in, where params is the third argument, is invoked on line 283:

BM_mesh_calc_tessellation_ex(bm, looptris, false);

...and there's our null pointer! I suspect it should be passing a struct with a boolean member, rather than just a boolean :)

It looks like there was a commit four days ago that added some new tessellation features. @ideasman42 could you have a look at that, since your name's on the commit?

Here's a quick replication case: creating a BVHTree from a BMesh will cause the crash. Just run the script in the text editor.

tessellation-crash.blend

I ran `PolyQuilt` with AddressSanitizer enabled and got this crash: ``` ==16360==ERROR: AddressSanitizer: access-violation on unknown address 0x000000000000 (pc 0x7ff6f6450f16 bp 0x12d383a9a1d0 sp 0x00aed69fbef0 T0) ==16360==The signal is caused by a READ memory access. ==16360==Hint: address points to the zero page. - 0 0x7ff6f6450f15 in BM_mesh_calc_tessellation_ex D:\blender-git\blender\source\blender\bmesh\intern\bmesh_mesh_tessellate.c:274 - 1 0x7ff6f6450fcb in BM_mesh_calc_tessellation D:\blender-git\blender\source\blender\bmesh\intern\bmesh_mesh_tessellate.c:283 #2 0x7ff70065fa4c in C_BVHTree_FromBMesh D:\blender-git\blender\source\blender\python\mathutils\mathutils_bvhtree.c:972 ``` So, it's reading a null pointer. Line 274 is: ``` bm_mesh_calc_tessellation__single_threaded(bm, looptris, params->face_normals); ``` The only thing that reads here is `params->face_normals`, so `params` is probably null. The function it's in, where `params` is the third argument, is invoked on line 283: ``` BM_mesh_calc_tessellation_ex(bm, looptris, false); ``` ...and there's our null pointer! I suspect it should be passing a struct with a boolean member, rather than just a boolean :) It looks like there was a commit four days ago that added some new tessellation features. @ideasman42 could you have a look at that, since your name's on the commit? Here's a quick replication case: creating a BVHTree from a BMesh will cause the crash. Just run the script in the text editor. [tessellation-crash.blend](https://archive.blender.org/developer/F10180059/tessellation-crash.blend)

blender-admin commented

2021-06-18 20:20:22 +02:00

This issue was referenced by d52b7dbe26

This issue was referenced by d52b7dbe2695c673b3bad091b55893413e7b022b

Germano Cavalcante commented

2021-06-18 20:22:28 +02:00

Changed status from 'Confirmed' to: 'Resolved'

Germano Cavalcante closed this issue

2021-06-18 20:22:28 +02:00

Germano Cavalcante self-assigned this 2021-06-18 20:22:28 +02:00

Robert Guetzkow commented

2021-06-20 21:46:37 +02:00