fix: double render call with timeline obj causes a seg fault

[devmessias]

This closes #705

Description

test_morphing https://github.com/fury-gl/fury/blob/b37669f8767d460b97e17584106ad8a6b5d26491/fury/tests/test_gltf.py#L285 causes a segmentation fault error in most of the github actions. It seems correlated with the headless mode.

• The double call Render method with a timeline and a gltf object in a headless env seems to be the cause of the seg fault.

I've created a simple repo with just the test_morphing to debug this issue. The following zip has all the logs of the failed actions: 12_DEBUG fail-test.txt : catchsegv python3 debug_fail.py 13_DEBUG catch.txt: python3 -X faulthandler debug_fail.py

logs_github_actions_traceback.zip

[codecov[bot]]

Codecov Report

Merging #706 (af9dc79) into master (e0f1613) will increase coverage by 0.05%. The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master     #706      +/-   ##
==========================================
+ Coverage   50.14%   50.19%   +0.05%     
==========================================
  Files         120      120              
  Lines       27884    28019     +135     
  Branches     2961     2987      +26     
==========================================
+ Hits        13982    14064      +82     
- Misses      13436    13489      +53     
  Partials      466      466              

Impacted Files	Coverage Δ
fury/gltf.py	0.00% <0.00%> (ø)
fury/fury/animation/timeline.py	66.08% <0.00%> (+0.21%)	:arrow_up:
fury/fury/tests/test_gltf.py	89.44% <0.00%> (+1.49%)	:arrow_up:
fury/fury/gltf.py	80.33% <0.00%> (+1.61%)	:arrow_up:

[devmessias]

Maybe there is a issue with mesa and vtk? Also, this render has no meaning there, because the buffer continue to be updated.

[m-agour]

@devmessias @filipinascimento we actually will not need any showm.render() since record function creates a new iren and a new RenderWindow. so force rendering will not help the screenshot in any way. The problem lies with in the showManager, something is not being created or cleaned properly. I think best way to debug it is to find out why the minimal example here https://github.com/fury-gl/fury/issues/603#issuecomment-1172655694 is failing.

[devmessias]

I looked at some tests being skipped because of SEG FAULT. I believe we can have multiple causes for different SEG FAULTs. Some of them happens only in the headless mode. Besides that, we have some PRs that can help to mitigate this. For example, we have an open issue with timer callbacks that are not properly being removed. https://github.com/fury-gl/fury/issues/483

[devmessias]

Hi @m-agour are you talking about this ?

The problem here is that you never removed the context, events etc from the second show_manager instance.

This code works without SEG FAULTs

import os
from fury import window, ui
from fury.data import DATA_DIR

def test_1():
    scene = window.Scene()
    show_manager = window.ShowManager(scene)
    show_manager.initialize()
    show_manager.start()
    filename = "test_ui_file_menu_2d"
    recording_filename = os.path.join(DATA_DIR, filename + ".log.gz")

    show_manager = window.ShowManager(size=(600, 600), title="FURY FileMenu")
    show_manager.play_events_from_file(recording_filename)

    show_manager.window.RemoveRenderer(show_manager.scene)
    show_manager.scene.SetRenderWindow(None)
    show_manager.window.Finalize()
    del show_manager.iren
    del show_manager.window

    filemenu = ui.FileMenu2D(size=(500, 500),
                             directory_path=os.getcwd())
    show_manager = window.ShowManager(size=(600, 600),
                                      title="FURY FileMenu")
    show_manager.scene.add(filemenu)
    show_manager.start()

test_1()

[m-agour]

@devmessias Yes this the example. It only causes a seg fault in Ubuntu (mostly in all linux distros) Your modification is still giving me a seg fault at the last line show_manager.start() in Ubuntu. Even running it as a normal code it still gives a sig fault Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)

[Garyfallidis]

This is a single line change. So will merge this and be happy to continue the segfault discussion here or another PR.

[devmessias]

m-agour

@devmessias Yes this the example. It only causes a seg fault in Ubuntu (mostly in all linux distros) Your modification is still giving me a seg fault at the last line show_manager.start() in Ubuntu. Even running it as a normal code it still gives a sig fault Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)

My mistake, I was using my custom window module to run your example @m-agour , I'll sent a PR this week with those changes.