Make manim crash when ffmpeg crashes

[leotrs]

Motivation

Sometimes, if FFMPEG crashes, manim happily keeps chugging along without reporting anything. It even prints a message as if everything was ok. In the picture below, FFMPEG complains that it didn't find a necessary file, but manim reports that the file is ready to view. In fact, manim didn't render anything at all!

Importantly, this allows some tests to pass even if FFMPEG has NOT rendered anything!

Explanation for Changes

Instead of calling FFMPEG and just waiting for it to finish, use subprocess.check_call, which will raise an exception unless the subprocess terminates successfully.

Testing Status

Tests pass locally.

Acknowledgement

• I have read the Contributing Guidelines

[eulertour]

Building the docs fails for me locally too 😕
Which sort of makes sense, because although the docs do build on master they output errors like this a few times:
manim/docs/source/media/videos/480p30/partial_movie_files/Example3DNo1-1/partial_movie_file_list.txt: Invalid data found when processing input

I think manim is trying to do something with partial movie files even when it's been instructed to only save the last frame, meaning there are no partial movie files to process in the first place.

[huguesdevimeux]

This looks all good, but can you do this for the other ffmpeg commands as well ?

[kolibril13]

Looks good to me as well.
However I wonder why the docs build saies:
'/home/docs/checkouts/readthedocs.org/user_builds/manimce/checkouts/546/docs/source/media/videos/480p30/Example3DNo1-1.mp4']' returned non-zero exit status 1.
The strange thing is, that there is only Example3DNo1 and no Example3DNo1-1.
I will make a commit and temporary remove Example3DNo1, so maybe this will solve this.

[behackl]

Looks good to me as well.
However I wonder why the docs build saies:
'/home/docs/checkouts/readthedocs.org/user_builds/manimce/checkouts/546/docs/source/media/videos/480p30/Example3DNo1-1.mp4']' returned non-zero exit status 1.
The strange thing is, that there is only Example3DNo1 and no Example3DNo1-1.
I will make a commit and temporary remove Example3DNo1, so maybe this will solve this.

@kolibril13 No! Remember: the documentation slightly changes file names in order to make them unique; if you remove this one example, I am rather sure another one will fail.

Please revert your commit again. I fear that we need to find a tailored solution for the documentation, or we will need to fix the underlying issue first...

[kolibril13]

Ok, I will read this example, now Example3DLightSourcePosition is breaking.
Its interesting that first of all the 3d scenes seems to be affectd

[behackl]

Because the files in 3d_scenes are rendered first.

[kolibril13]

Because the files in 3d_scenes are rendered first.

Ok, I am only hoping for an easy fix :)
I will once run it without the 3d scenes and read them as soon as the docs run should fail.

[behackl]

You can check out one of the last successful builds (https://readthedocs.org/projects/manimce/builds/12105792/) and search for "invalid data found" in the last log field. The error occurs 30 times.

[leotrs]

Aha! I think I may have inadvertently uncovered the very bug that sent me down this path in the first place. Imma fix the docs and report back.

[leotrs]

All checks green now. This was super hard to debug, mainly because of the mess of the config system and scene file writer.

The tests pass on CI, but not locally, which is weird. Locally, I'm seeing problems with tests/test_pango.py, so think I may just have screwed up my setup. Can anyone please checkout this branch and run the tests locally?

[behackl]

The changes LGTM, the pipeline passes and the examples I checked sporadically in the rendered documentation also seem okay.

Thanks for your efforts!

[leotrs]

Ok this broke the --disable_caching flag (thanks @eulertour for realizing and reverting).

In order to keep working on this, I want to implement tests first.