09401: SDL sound unaffected by fast forward

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
09401	Sound	Minor	Always	Mar 5, 2026, 09:58	30 days ago

Tester	Janez	View Status	Public	Platform	SDLMAME
Assigned To	R. Belmont	Resolution	Fixed	OS	Linux (64-bit)
Status [?]	Resolved			Driver
Version	0.286	Fixed in Version	0.287	Build	x64
		Fixed in Git Commit	c48c188	Github Pull Request #

Summary	09401: SDL sound unaffected by fast forward
Description	sdl is the default sound output method, if you press the fast forward button the video accelerates but the audio doesn't. So, after a while, you get a video-audio delay. It doesn't happen with the other sound output methods; portaudio, pulse and pipewire are not affected. It never happened before, so I guess the new upgrade from SDL2 to SDL3 is involved.
Steps To Reproduce	- Start any game with the sdl sound output method (the default): mame tehkanwc -sound sdl - Press the fast forward button (Page Down for SDL MAME) for a couple of seconds. - You'll see the video accelerate, but the audio will not.
Additional Information
Github Commit

Flags	Regression
Regression Version	0.286
Affected Sets / Systems

Attached Files

No.24207 Robbbert Moderator Mar 5, 2026, 10:12	I don't use SDL so unable to confirm, but it's acknowledged. Hopefully someone who does use SDL can check it out.
No.24235 holub Tester Mar 10, 2026, 00:42	Confirming issue with sdl3; sdl2 - works fine
No.24270 hap Developer 31 days ago edited on: 31 days ago	I didn't compile SDLMAME, but I see what's going on in the sdl3_sound module. It doesn't use sink_callback, it sends samples immediately through SDL_PutAudioStreamData instead of abuffer.
No.24275 Robbbert Moderator 31 days ago edited on: 31 days ago	This commit references this report: https://github.com/mamedev/mame/commit/c48c18802e312b3ce5fea4502a369a5ae564fb39 Can someone with SDL say if it changes anything?
No.24277 hap Developer 30 days ago edited on: 30 days ago	RB is SDL user, so... Looking at the diff, the infinite buffer queuing is indeed fixed. BTW the workaround introduces latency that sticks around after fast-forward (or other events that may cause overruns), 0.16s by default. (which itself can be worked around with by pausing/unpausing, like previously with the recently fixed abuffer sound crackling bug)