BMesh.free causes crash under curtain conditions #96791

New Issue

Jan Ondrej · 2022-03-26T09:28:38+01:00

Jan Ondrej commented

2022-03-26 09:28:38 +01:00

System Information
Operating system: Linux-5.16.16-200.fc35.x86_64-x86_64-with-glibc2.34 64 Bits
Graphics card: NVIDIA GeForce GTX 1660 SUPER/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 510.47.03

Blender Version
Broken: version: 3.2.0 Alpha, branch: master, commit date: 2022-03-25 21:17, hash: 1909fd2781

all 3.x versions are broken, 3.0.0 sometimes works

Worked: 2.93.8 LTS
On Fedora also some 2.91 versions have been affected, according to rebuild agains newer pythons.
Reported on redhat bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1872980

Short description of error
blender.crash.txt
Using my python addon blender crashes on _PyEval_EvalFrameDefault or _Py_Dealloc.

Exact steps for others to reproduce the error
Unable to describe simple steps to reproduce. This happens using my python addon, but if a bug is in addon, still shouldn't crash blender but throw an python traceback.
When I delete line describen in crash log's backtrace, then another row is marked as a problem, so unable to identify exactly what's happening.
Please, let me know, if you need more information.

Edit, run this script with the following command line arguments: boxmaker_crash_single_script.py
blender -b --factory-startup --python boxmaker_crash_single_script.py

**System Information** Operating system: Linux-5.16.16-200.fc35.x86_64-x86_64-with-glibc2.34 64 Bits Graphics card: NVIDIA GeForce GTX 1660 SUPER/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 510.47.03 **Blender Version** Broken: version: 3.2.0 Alpha, branch: master, commit date: 2022-03-25 21:17, hash: `1909fd2781` ``` all 3.x versions are broken, 3.0.0 sometimes works ``` Worked: 2.93.8 LTS On Fedora also some 2.91 versions have been affected, according to rebuild agains newer pythons. Reported on redhat bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1872980 **Short description of error** [blender.crash.txt](https://archive.blender.org/developer/F12947980/blender.crash.txt) Using my python addon blender crashes on _PyEval_EvalFrameDefault or _Py_Dealloc. **Exact steps for others to reproduce the error** Unable to describe simple steps to reproduce. This happens using my python addon, but if a bug is in addon, still shouldn't crash blender but throw an python traceback. When I delete line describen in crash log's backtrace, then another row is marked as a problem, so unable to identify exactly what's happening. Please, let me know, if you need more information. ---- Edit, run this script with the following command line arguments: [boxmaker_crash_single_script.py](https://archive.blender.org/developer/F13052503/boxmaker_crash_single_script.py) `blender -b --factory-startup --python boxmaker_crash_single_script.py`

Jan Ondrej commented

2022-03-26 09:28:38 +01:00

Added subscriber: @ondrejj

Pratik Borhade commented

2022-03-26 10:15:04 +01:00

Added subscriber: @PratikPB2123

Pratik Borhade commented

2022-03-26 10:15:04 +01:00

Changed status from 'Needs Triage' to: 'Needs User Info'

Pratik Borhade commented

2022-03-26 10:15:04 +01:00

Hi, thanks for the report.
Not necessary that the crash caused by Python script is a bug in Blender code.
We use this canned response if crash or error is introduced by their own python script.
Crashes or errors involving your own C/C++ code (Python C/API for example) will only be accepted as bugs if you are able to identify the error in Blender's code. Finding that your code isn't working as expected in Blender isn't sufficient to consider it a Blender bug as there may be other reasons that aren't necessarily caused by errors in Blender's code.

Hi, thanks for the report. Not necessary that the crash caused by Python script is a bug in Blender code. We use this canned response if crash or error is introduced by their own python script. ```Crashes or errors involving your own C/C++ code (Python C/API for example) will only be accepted as bugs if you are able to identify the error in Blender's code. Finding that your code isn't working as expected in Blender isn't sufficient to consider it a Blender bug as there may be other reasons that aren't necessarily caused by errors in Blender's code.```

Stephen Swaney commented

2022-03-26 16:30:00 +01:00

Added subscriber: @StephenSwaney

Stephen Swaney commented

2022-03-26 16:30:00 +01:00

/me bets on a reference counting error.

Jan Ondrej commented

2022-03-26 18:23:47 +01:00

Hello. Thanks for reply. My script is only plain python, using only bpy, mesh and mathutils (only Vector) modules. No extenal C code. This is the reason why I reported this here.

If you wish, I can share my python script, even if it's not ready for public. My script also needs some interaction, which I can't fully describe, so I don't know if it helps.

Hello. Thanks for reply. My script is only plain python, using only bpy, mesh and mathutils (only Vector) modules. No extenal C code. This is the reason why I reported this here. If you wish, I can share my python script, even if it's not ready for public. My script also needs some interaction, which I can't fully describe, so I don't know if it helps.

Jesse Yurkovich commented

2022-03-26 21:29:40 +01:00

Added subscriber: @deadpin

Jesse Yurkovich commented

2022-03-26 21:29:40 +01:00

Either way, we need a way to reproduce the crash. We can't, and won't, debug through an entire addon to do so. A simplified repro is necessary. Strip out as much as possible until you arrive at only the pieces that are necessary.

Jan Ondrej commented

2022-03-27 11:10:43 +02:00

Sorry, I can't reproduce this in a small script, however I found a problem. Using this function fails:

def bevel(cover, offset, segments=10):
    bm = bmesh.from_edit_mesh(cover.data)
    bmesh.ops.bevel(bm,
      geom=[x for x in bm.verts[:]+bm.edges[:] if x.select],
      offset=offset,
      offset_type='OFFSET',
      segments=segments,
      profile=0.5,
      affect="EDGES", #vertex_only=False,
      loop_slide=True,
      clamp_overlap=True
    )
    bmesh.update_edit_mesh(cover.data)
    bm.free()

After removal of bm.free(), script works again.
@StephenSwaney, thank you for pointing me to an referencing error.

Why this bm.free() exists and should or shouldn't be called? In blender's documentation I found, that I should call it.

Sorry, I can't reproduce this in a small script, however I found a problem. Using this function fails: ``` def bevel(cover, offset, segments=10): bm = bmesh.from_edit_mesh(cover.data) bmesh.ops.bevel(bm, geom=[x for x in bm.verts[:]+bm.edges[:] if x.select], offset=offset, offset_type='OFFSET', segments=segments, profile=0.5, affect="EDGES", #vertex_only=False, loop_slide=True, clamp_overlap=True ) bmesh.update_edit_mesh(cover.data) bm.free() ``` After removal of bm.free(), script works again. @StephenSwaney, thank you for pointing me to an referencing error. Why this bm.free() exists and should or shouldn't be called? In blender's documentation I found, that I should call it.

Pratik Borhade commented

2022-03-28 05:20:32 +02:00

I think it's not necessary to call bm.free()
Data is expected to get freed at the end of function execution.

We may handle this as a valid report if that call is responsible for the crash.

I think it's not necessary to call bm.free() Data is expected to get freed at the end of function execution. We may handle this as a valid report if that call is responsible for the crash.

Jan Ondrej commented

2022-03-28 07:52:55 +02:00

I can confirm, that reenable of bm.free() will crash again and after removal it's back.

This script doesn't crash on bm.free() or at end of function, but crashes aprox. 10 rows after return from this function. May be crash log from this report can help to identify, what's happening.

My initial problem has been solved by removing of bm.free(), but I am ready to do more tests if needed. Just let me know, how I can help.

I can confirm, that reenable of bm.free() will crash again and after removal it's back. This script doesn't crash on bm.free() or at end of function, but crashes aprox. 10 rows after return from this function. May be crash log from this report can help to identify, what's happening. My initial problem has been solved by removing of bm.free(), but I am ready to do more tests if needed. Just let me know, how I can help.

Pratik Borhade commented

2022-04-05 09:27:45 +02:00

Added subscriber: @lichtwerk

Pratik Borhade commented

2022-04-05 09:27:45 +02:00

This script doesn't crash on bm.free() or at end of function, but crashes aprox. 10 rows after return from this function. May be crash log from this report can help to identify, > what's happening.

There might be an issue in your script itself (Still I'm not sure unless we have steps to reproduce it).
Please try one of the community websites: https://www.blender.org/community/
If you think you found a bug, please submit a new report and carefully follow the instructions. Be sure to provide system information, Blender version, and a .blend file with exact steps to reproduce the problem.

Just in case I'm missing something, @lichtwerk, can you check?

> This script doesn't crash on bm.free() or at end of function, but crashes aprox. 10 rows after return from this function. May be crash log from this report can help to identify, > what's happening. There might be an issue in your script itself (Still I'm not sure unless we have steps to reproduce it). Please try one of the community websites: https://www.blender.org/community/ If you think you found a bug, please submit a new report and carefully follow the instructions. Be sure to provide system information, Blender version, and a .blend file with exact steps to reproduce the problem. --- Just in case I'm missing something, @lichtwerk, can you check?

Philipp Oeser commented

2022-04-05 10:12:52 +02:00

To speed the process up (there are multiple people looking at this already), please always attach a .blend file where this script is already set up to just be executed from the Text Editor.
(This also ensures we are all on the same page)

To speed the process up (there are multiple people looking at this already), please always attach a .blend file where this script is already set up to just be executed from the Text Editor. (This also ensures we are all on the same page)

Jan Ondrej commented

2022-04-05 10:25:48 +02:00

There is no .blend file. My script is running from blender addons and is using multiple scripts.

Philipp Oeser commented

2022-04-05 10:37:52 +02:00

Added subscriber: @ondrew

Philipp Oeser commented

2022-04-05 10:37:52 +02:00

@ondrew : than please share that here or try to isolate relevant parts into a script (embedded in a .blend file) with clear steps to reproduce

Jan Ondrej commented

2022-04-05 10:46:11 +02:00

@lichtwerk, you mentioned wrong people. :-)

So ok, I am adding my script. First is: boxmaker.py:
boxmaker.py

Copy this to addons and enable it.

Download pir.py:
pir.py

and run:

blender -P pir.py

This crashes my blender. Not tested today.
Removing of "bm.free()" can fix this script.

Please, do not ask me to make this script simpler. I can't figure, where is the exact problem.

@lichtwerk, you mentioned wrong people. :-) So ok, I am adding my script. First is: boxmaker.py: [boxmaker.py](https://archive.blender.org/developer/F12974711/boxmaker.py) Copy this to addons and enable it. Download pir.py: [pir.py](https://archive.blender.org/developer/F12974677/pir.py) and run: blender -P pir.py This crashes my blender. Not tested today. Removing of "bm.free()" can fix this script. Please, do not ask me to make this script simpler. I can't figure, where is the exact problem.

Philipp Oeser commented

2022-04-05 11:11:15 +02:00

Removed subscriber: @ondrew

Philipp Oeser commented

2022-04-05 11:11:15 +02:00

Changed status from 'Needs User Info' to: 'Needs Triage'

Philipp Oeser commented

2022-04-05 11:11:15 +02:00

In #96791#1335640, @ondrejj wrote:
@lichtwerk, you mentioned wrong people. :-)

Oops, thx noting

> In #96791#1335640, @ondrejj wrote: > @lichtwerk, you mentioned wrong people. :-) Oops, thx noting

Campbell Barton commented

2022-05-04 07:43:54 +02:00

Added subscriber: @ideasman42

Campbell Barton commented

2022-05-04 07:43:54 +02:00

@ondrejj please make a simple test script that causes the error without relying on a separate add-on.

Any logic that isn't related to reproducing the crash should be removed.

@ondrejj please make a simple test script that causes the error without relying on a separate add-on. Any logic that isn't related to reproducing the crash should be removed.

Jan Ondrej commented

2022-05-04 07:46:40 +02:00

Sorry, but I can't. My blender script programming is not good enough to do this. I did some experiment but without any success.

Campbell Barton commented

2022-05-04 07:52:07 +02:00

@ondrejj it should not be so difficult to merge the add-on into a script, then remove function calls - checking the crash still occurs.

Jan Ondrej commented

2022-05-04 07:54:34 +02:00

But don't know which calls to remove. After removal of one line which makes a problem this problem moves to another line. Sorry, but as I said, I did some experiments, but without success.

Campbell Barton commented

2022-05-05 07:26:09 +02:00

@ondrejj uploaded a slightly simplified script with tracing enabled that runs in background mode and crashes reliably, this needs some more investigation though.

Edit: this seems to be memory corruption however it's not clear where exactly this occurs, tested with ASAN and Valgrind don't point to anything obvious.
It looks like the Python state is corrupted some time before the crash occurs.

@ondrejj uploaded a slightly simplified script with tracing enabled that runs in background mode and crashes reliably, this needs some more investigation though. Edit: this seems to be memory corruption however it's not clear where exactly this occurs, tested with ASAN and Valgrind don't point to anything obvious. It looks like the Python state is corrupted some time before the crash occurs.

Campbell Barton commented

2022-05-05 07:26:30 +02:00

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

Campbell Barton commented

2022-05-05 07:26:58 +02:00

Changed status from 'Confirmed' to: 'Needs Developer To Reproduce'

Campbell Barton changed title from ~~Crash using python addon on _PyEval_EvalFrameDefault~~ to BMesh.free causes crash under curtain conditions

2022-05-05 07:28:32 +02:00

Pratik Borhade commented

2022-08-17 02:04:42 +02:00

Report is already classified as a Bug so can we also mark it as "Confirmed"?

Report is already classified as a `Bug` so can we also mark it as "Confirmed"?

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

BMesh.free causes crash under curtain conditions #96791