01815: All Sets: Saved machine states do not properly adjust to changes in -samplerate

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
01815	Save/Restore	Minor	Always	May 19, 2008, 21:23	May 22, 2008, 05:15

Tester	Tafoid	View Status	Public	Platform	MAME (Self-compiled)
Assigned To	aaron	Resolution	Fixed	OS	Windows 2000
Status [?]	Resolved			Driver
Version	0.125u1	Fixed in Version	0.125u2	Build	Athlon
		Fixed in Git Commit		Github Pull Request #

Summary	01815: All Sets: Saved machine states do not properly adjust to changes in -samplerate
Description	Simply, if you decide to run a game at a sample rate that is different than that which the STATE was made, the pitch of the audio is offset based on the difference in HZ between the original STATE sample rate and the newly issued sample rate. I've tested drivers which are set up for Save States (-autosave) as well as sets that are not set up for States (Using LeftShift + F7) - they've all reacted similarly. All tests were without any video or audio syncing.
Steps To Reproduce	For this example, we'll use the set PACMAN - Make sure your MAME.INI is default in samplerate (48000). Delete STA\PACMAN\auto.sta if that is present (or move aside if you are saving scores). - TYPE: *mame -autosave pacman* - Let the emulation start and boot up, you can exit MAME anytime. - NOW TYPE: *mame -autosave pacman -sr 36000* Observe the audio output - the pitch is much lower than expected! This also holds true for a STATE which is created at a lower sample rate, except, the pitch is increased in this case. Simply reverse the MAME commands above (36000 first.. then default 48000 after). This also causes an audio artifact that shows itself in rapid duplication of audio bits.. with obvious stuttering.
Additional Information	I had a discussion about this on MAME Testers forum nearly a year and a half ago. I tried to reference the forum posts but they've already been moved off. At the time, I was told that this was not really a bug. I still insist that it is indeed a bug and which even has a version in which the regression first occurs.
Github Commit

Flags
Regression Version	0.103u4
Affected Sets / Systems	All Sets

Attached Files

No.01012 Haze Senior Tester May 20, 2008, 11:06	Well it's clearly a bug, whether it's a bug by design or not is rather irrelevant.