Computer freeze when using MPV 0.41 on an old machine

[cybersnow02]

mpv Information

I was using MPV 0.38 and everything worked well. Then I decided to compile the newest 0.41. I started experiencing hard computer freeze when watching MPEG4 video. I freeze the whole computer in less than 5 minutes.

This is a old computer using Intel Core 2 Duo wit a NVidia 470 series Card. I force VDPAU for the backends.

I know 0.38 was the last version not enforcing GPUNext new backends. I suspect 0.39+ will cause this issue. I also recompiled the latest FFMPEG libs.

Other Information

- Linux version: Debian Buster 10
- Kernel Version: 6.18 custom build
- GPU Model: Nvidia GT 470
- Mesa/GPU Driver Version:
- Window Manager and Version: xfce 4.20
- Source of mpv:
- Latest known working version: 0.38
- Issue started after the following happened: playing mpeg4 video

Reproduction Steps

Compile 0.41 and test with this specific hardware setup

Expected Behavior

should work as 0.38 was.

Actual Behavior

it hard freezes the whole computer without let me diag anything (no log at all nor kernel crash)

Log File

I cant

Sample Files

No response

I carefully read all instruction and confirm that I did the following:

• I tested and confirmed that the issue exists with the latest release version or newer.
• I provided all required information including system and mpv version.
• I produced the log file with the exact same set of files, parameters, and conditions used in "Reproduction Steps", with the addition of --log-file=output.txt.
• I produced the log file while the behaviors described in "Actual Behavior" were actively observed.
• I attached the full, untruncated log file.
• I attached the backtrace in the case of a crash.

[Dudemanguy]

You might just be experiencing the freeze described here. #17960 It was fixed by that PR as well. Can you try mpv master?

[Traneptora]

Can you run mpv with mpv --log-file=mpv.log and then attach mpv.log?

[CounterPillow]

Also try --gpu-api=opengl

[kasper93]

Also try --gpu-api=opengl

CounterPillow was faster, but I also would bet it's Vulkan change. It's probed first now and if you say whole system freeze than, it's likely driver/kernel issue.

[cybersnow02]

I tried --gpu-api=opengl without success. Still freezing. As for the log --log-file=mpv.log, I can't make it to write to the log file because the computer hard freeze and nothing is written to the HDD, so it is useless.

As for the issue #17960 , I don't know if this is related because I don't use pipewire. Anyway, I will try to compile the latest GIT master to do some test. I forgot to say that it seems to be related to a looping A-B condition. It only crash when doing a loop I think. I need to do more test. It cannot be a driver/kernel issue because it is working fine with all kind of video software including MPV 0.38. I can only reproduce this on MPV 2 last versions.

[kasper93]

including MPV 0.38. I can only reproduce this on MPV 2 last versions.

The quickest to find the issue, would be to bisect mpv and find where exactly it regressed.

[cybersnow02]

Ya... I was looking at it but it requires to exclude the whole GPUNext backends which is not a good news... Beside that, It could be related to FFMpeg 7.x and 8.1.0 libs. I can see alot of fixes related to memory and the latest FFMpeg 8.1.1 and 8.1.2 show it might be related to possible crashes like this. Let's see... It seems the FFMpeg team is now finding important bug fix and some of them are there for a long time without people ever noticed them. Look like the AI codes analyzers are maybe being used to clean up the code...

I know it might look stupid but can you make the possibility to exclude the VO-GPU-NEXT from the compilation (make it CMake optional)? All I need is VDPAU and OpenGL at last... I already removed alot of lines of code to exclude GPU-NEXT and compiled it successfully. So why not keep it compatible/lighter for old systems?

[CounterPillow]

The problem isn't that your system is old, the problem is that you're running an incredibly fucked up franken-system that evidently is barely holding together.

If you want to use non-gpu-next, use --vo=gpu.