09165: 4in1boot: sound randomly 'dies' (music stops, odd broken / out of sync sound effects still play)

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
09165	Sound	Minor	Always	12 days ago	8 days ago

Tester	Haze	View Status	Public	Platform	MAME (Official Binary)
Assigned To		Resolution	Open	OS	Windows 10/11 (64-bit)
Status [?]	Acknowledged			Driver	kaneko/snowbros.cpp
Version	0.277	Fixed in Version		Build	64-bit
		Fixed in Git Commit		Github Pull Request #

Summary	09165: 4in1boot: sound randomly 'dies' (music stops, odd broken / out of sync sound effects still play)
Description	When playing games from this 4in1 the sound will fail in an unusual way. this is most noticeable in the Pacman game. Typically when this happens the music will hang on a single note, or go silent. Some sound effects still play, but they're corrupt, and often delayed. This is semi-random, and can happen as early as the first stage, but also sometimes much later although it's pretty much guaranteed to happen eventually. It isn't clear if this is a regression (as I don't remember it happening before the YM/FM rewrite) but given the semi-random nature of the issue it might just have not been noticed previously. While the game seems to be derived from the SemiCom titles in the same driver (especially Hyper Pacman / Twinkle) I haven't observed the same problem with those as of yet.
Steps To Reproduce	play 'pacman 2' in 4in1boot until the sound dies.
Additional Information
Github Commit

Flags
Regression Version
Affected Sets / Systems	4in1boot

Attached Files

No.23257 Robbbert Moderator 12 days ago	Unable to replicate this.
No.23258 Haze Senior Tester 12 days ago	it definitely happens 3:28:05 (2 years ago) 1:04:10 (last week) also reproduced it off stream a few times
No.23260 Robbbert Moderator 11 days ago edited on: 11 days ago	Just tried again, went as far as level 23 in the pacman game, no loss of sound.
No.23262 Robbbert Moderator 8 days ago	Which sound output are you using (auto,dsound....). What are your other sound-related settings? I've acknowledged this, since nobody else has been able to confirm it.
No.23265 Haze Senior Tester 8 days ago edited on: 8 days ago	I can get it to trigger on a stock build without changing anything, so whatever the defaults are on Windows these days. It seems entirely random though, both times on stream it happened in the first few screens but as you note, sometimes you can play it for a long time and get no reproduction. Annoyingly the times it's happened early I've not been using -record to try and capture and inp, but I'm wondering if it's actually memory corruption somewhere in the driver because it seems too unpredictable and strange in the end result (which would likely make an inp capture useless anyway) The argument against memory corruption of the Z80 program is that the Z80 does still seem to be running, it does still seem to have responding to sound effects albeit with a high delay, which suggests the CPU has become swamped, probably with IRQs, meaning it rarely gets chance to process the soundlatch. The IRQs come from the YM timer, which would suggest the YM timer has ended up in an invalid state and is spamming the Z80 with too many IRQs for it to do much else (which would also explain why the music stopped working, as if the YM timers broke, the music would also no longer work) I guess it could still be memory corruption if the memory with the timer details is getting stomped on though. It might be worth running the driver with one of the more aggressive memory analysis tools to detect for any possible corruption. There's a small chance it's protection, there is a read that's set to return a fixed value in the driver, but most of the Semicom type games have that, maybe to make board swaps more annoying. This doesn't have the usual Semicom protection of 'code in RAM' though, as it's a bit more bootleggy overall than the original Semicom games. I don't think it can be down to the MAME output modules anyway, otherwise I'd be seeing it on far more games,