3.6.0: AudioChannel.Position not working correctly for all sound types

[ivan-mogilko]

According to the manual AudioChannel.Position:

What position represents depends on the FileType of the audio clip:

MIDI - the beat number
MOD/XM/S3M - the pattern number
WAV/VOC - the sample number (e.g. in a 22050 Hz sound, 22050 = 1 second)
OGG/MP3 - milliseconds offset

This was never fully reimplemented in 3.6.0 after audio subsystem rewrite.

Latest state:

• - MIDI - not working;
• - MOD - not working;
• - WAV - done;
• - OGG/MP3 - done

Same refers to the AudioChannel.Seek function.

[ericoporto]

Is the midi and mod not working a non-implemented part in SDL_Sound too, like it was in the more case?

[ivan-mogilko]

Is the midi and mod not working a non-implemented part in SDL_Sound too

No, it's not something non-implemented, but simply a restriction of the interface design. SDL_Sound's interface is fully opaque so there's no way to know the actual sound format, or do any format specific operations. We may seek and get position in milliseconds for all these sound types, but not in the special position kind required by AGS.

With the WAV it is possible to get sample count since we know duration in ms and frequency, but I don't yet know if anything similar is possible to do with MIDI and MOD.

[ericoporto]

I imagine with MIDI it should be possible, but the pattern number for mod is an odd one, I need to look into the code to see what that really means.

[ivan-mogilko]

If this cannot be done with the available data, then it may be worth asking SDL_Sound author(s) if they see this acceptable to add format-specific functions into the library API. Or as some kind of extension.

[ericoporto]

ok, read a bit on the old code, there were jgmod and aldumb builds of ags apparently, am considering only aldumb. Aldumb clips returns current order for position. From reading the loaded data, I think the mod clip is made of orders which index a specific pattern, which is it made of different ordered samples in different channels.

Looking into SDL_sound, the modplug used there doesn't have an interface internally too. It has an interesting comment on seeking using milliseconds, which is why I think ags/aldumb used the order instead.
https://github.com/icculus/SDL_sound/blob/495e948b455af48eb45f75cccc060498f1e0e8a2/src/libmodplug/modplug.h#L40-L46