Mixer is not accurate in playing audio at a rate.

[qaqFei]

Environment:

If possible, please include the output of pygame.print_debug_info() from your program in your bug report. It looks something
like this:

pygame-ce 2.5.2 (SDL 2.30.8, Python 3.12.8)
Platform:		Windows 10
System:			Windows
System Version:	10.0.19042.508
Processor:		Genuine Intel(R) 0000 @ 1.80GHz
Architecture:	Bits: 64bit	Linkage: WindowsPE

Python:			    CPython 3.12.8 (tags/v3.12.8:2dc476b, Dec  3 2024, 19:30:04) [MSC v.1942 64 bit (AMD64)] on win32
pygame version:		 2.5.2
SDL versions:		2.30.8.0
SDL Mixer versions:	2.8.0.0

Mixer Driver:		unknow

If you can't get the debug output, any of the environment details included in it that you do know would be useful
in diagnosing the issue & helping you.

Other environment details, not included in print_debug_info(), that might help:

• Relevant hardware (e.g. if reporting a bug about a controller, tell us the brand & name of it):

Current behavior:
The latency of the Mixer get_pos gradually increases, disappearing when the Mixer is reset (i.e., get_pos reset to zero)

Screenshots

directsound:

dxs.mp4

sdl2_mixer:

sdl2-mixer.mp4

Steps to reproduce:

Please explain the steps required to duplicate the issue, especially if you are able to provide a sample application.
if the bug is caused by a specific file (image, font, sound, level, please upload it as an attachment

1. Load a long audio file
2. Play it
3. Wait playing...

Test code

If possible add a simple test program that shows the problem described in this report.

import time
from pygame import mixer

mixer.init()

mixer.music.load("a_long_music.mp3")
mixer.music.play()
while not mixer.music.get_busy(): ...

start_time = time.time()

while mixer.music.get_busy():
    # There is no error here.
    print(time.time() - start_time, mixer.music.get_pos() / 1000)

# If the length of that audio is 3600s, the variable "use_time" will be greater than 3600s
use_time = time.time() - start_time

Stack trace/error output/other error logs

paste other relevant logs or stack traces here, if applicable

[qaqFei]

When I turned up the buffer of the mixer, the delay disappeared decrease

[qaqFei]

I'm guessing it's get_pos using the now_time-start_time timing method, but there's a delay in updating the audio buffer

[damusss]

I'm not home so I'm unable to read everything except for the title. I'm here to tell you that get pos for weird design choices does not return the music position but only the elapsed time from when you called play(). this is because the correct sdl function was added later. it might change in future versions so you might want a custom way to keep track of the music position.

[Starbuck5]

I tested your scenario with https://github.com/pygame-community/pygame-ce/blob/main/examples/data/surfonasinewave.xm,
Windows 10, did not see any growing difference between the position of the track and the time elapsed.

What happens if you test with this file?

The fact that this changed when you changed your buffer size does not give me a good feeling about this issue being something that can be productively addressed.

[qaqFei]

use_time will be different from the actual length of the music, but I can't remember if it's bigger or smaller

[Starbuck5]

Have you tested on this file?

Also have you tried speakers vs headphones, maybe it's your device that is being weird.

[qaqFei]

There is a real error, about 500ms every 300s, but it only happens in MP3 format
i convert vm to mp3 using ffmpeg.

[qaqFei]

The audio speed is about 1.0016 times, which is very important in music games

[Starbuck5]

Further questions:

• Have you tried older pygame-ce versions? Is this consistent throughout? pygame-ce 2.5.2 uses SDL2_mixer 2.8.0, and according to their release notes (https://github.com/libsdl-org/SDL_mixer/releases/tag/release-2.8.0 ) they switched from dr_mp3 to minimp3 in that release.
• You say there is a drift between get_pos and time.time(), but which is right relative to the music? Is the music playing slower than it should, or is get_pos wrong?
• If time.time() is right, or get_pos() is right (relative to the music), why not use the one that is correct for your use case?
• Does it change the result if you don't call get_pos() over and over with no delay? I looked at the source of Mix_GetMusicPosition and it locks the audio every time it is called. Perhaps you are slowing down the playback yourself. https://github.com/libsdl-org/SDL_mixer/blob/dc9b96f12d9f8d8893f7f8493d96c577140f7873/src/music.c#L1039-L1055

[qaqFei]

When I only get his duration, it turns out to be wrong

[qaqFei]

ffmpeg says it has 302.52s, so sdl2-mixer is wrong.

[qaqFei]

It's also wrong when I use 2.7.2

[qaqFei]

I'm guessing it's also a problem with the sample rate, but it seems like a possibility