Scene Caching Tries to Access the Wrong File in Certain Circumstances

[sudgy]

Consider the following scene:

class Issue3DCache(ThreeDScene):
    def construct(self):
        self.begin_ambient_camera_rotation(rate=0.1)
        self.wait()
        self.wait()

when I run manim once, it makes a black screen as expected, but when I run it again, it crashes saying that it can't find the partial movie file.

When running manim test.py Issue3DCache -pl the first time:

Animation 0 : Partial movie file written in {'.../480p15/partial_movie_files/Issue3DCache/2314591449_1-0_2415171488.mp4'}
Animation 1 : Partial movie file written in {'.../480p15/partial_movie_files/Issue3DCache/1347326907_1-0_2415171488.mp4'}

When running it again:

Wait 0 : Using cached data (hash : 2314591449_1-0_2415171488)
Animation 1 : Partial movie file written in {'.../480p15/partial_movie_files/Issue3DCache/2314591449_1-0_2415171488.mp4'}
Impossible to open 'file:.../480p15/partial_movie_files/Issue3DCache/865043704_1-0_2415171488.mp4'
.../480p15/partial_movie_files/Issue3DCache/partial_movie_file_list.txt: No such file or directory

[huguesdevimeux]

Hi, thanks for the report. Did you try to run this with --disable_caching flag ?

[sudgy]

Yeah, I did, and it works fine when using that flag.

[sudgy]

Well, I just found somewhere else it breaks without rotation. Let me try to isolate what's breaking it.

[sudgy]

Here's a reduced version of another case I found:

class Issue3DCache2(Scene):
    def construct(self):
        self.wait()
        m = Square()
        self.add(m)
        #self.wait()

Run it the first time without the commented wait, and the second time with it. Then a similar problem occurs. However, unlike before, the wait has to be added between runs.

[huguesdevimeux]

Alright, thanks so much for that.
Last request (I swear !) Could you run a borken scene with -v DEBUG flag (and without --disable_caching flag)and post the logs ?

[sudgy]

output.txt

This was the second one (without the rotation), even though the name is still Issue3DCache. It doesn't seem to be that much different.

[sudgy]

I was wanting to jump to an earlier commit so I wouldn't have to deal with this problem any more, and I found that the bug was introduced in commit 1a516db.

[huguesdevimeux]

Maybe this is fixed by #489 . I tested it, and it worked. Could you confirm ?

[sudgy]

If it worked for you, I'm fine trusting that. I don't really know any way to test it right now short of uninstalling my copy and then installing from your fork, and that seems like more trouble than it's worth.

[sudgy]

This has been closed, but it's not actually fixed yet. I'm guessing somebody got confused and thought that #489 was #498.