02043: All games in cps2.c: All Parent Sets for CPS2 Games Abort

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
02043	Crash/Freeze	Critical (emulator)	Always	Jul 25, 2008, 05:40	Jul 31, 2008, 08:57

Tester	FrizzleFried	View Status	Public	Platform	MAME (Official Binary)
Assigned To		Resolution	Fixed	OS	Windows XP/Vista 32-bit
Status [?]	Resolved			Driver	capcom/cps2.cpp
Version	0.126u2	Fixed in Version	0.126u3	Build	Athlon
		Fixed in Git Commit		Github Pull Request #

Summary	02043: All games in cps2.c: All Parent Sets for CPS2 Games Abort
Description	Every CPS2 game I've tried loads with black screen for a couple seconds then aborts to the front end/command prompt. Clones don't appear to be affected. Games I've tried: Xmen vs SF Xmen: Children of the Atom Both Dungeons & Dragons games 1944 Street Fighter Alpha (all of them) Marvel vs Capcom
Steps To Reproduce	Simply attempt to run a parent CPS2 ROM.
Additional Information	Here are a few backtraces - they all seem to indicate sound problem causing the crash (as Mamesick mentions). I was able to crash a clone of SFA (sfar1) as well, so it might not be specifically a parent issue. - 1944 Program received signal SIGSEGV, Segmentation fault. 0x77f87eeb in ?? () (gdb) bt full #0 0x77f87eeb in ?? () No symbol table info available. #1 0x7800efdf in fcloseall () from C:\WINNT\system32\msvcrt.dll No symbol table info available. #2 0x0120fe57 in qsound_update (param=0xa03a720, inputs=0x0, buffer=0x7bdce00, length=481) at src/emu/sound/qsound.c:365 chip = (struct qsound_info ) 0xa03a720 i = 16 j = 4656612 rvol = 445948416 lvol = 29 count = 2292600 pC = (struct QSOUND_CHANNEL ) 0xa03aaa8 datap = {0x7bddce0, 0x7be02a8} Backtrace stopped: frame did not save the PC - sfa Program received signal SIGSEGV, Segmentation fault. 0x77f87eeb in ?? () (gdb) bt full #0 0x77f87eeb in ?? () No symbol table info available. #1 0x7800efdf in fcloseall () from C:\WINNT\system32\msvcrt.dll No symbol table info available. #2 0x0120fe57 in qsound_update (param=0x983a838, inputs=0x0, buffer=0x983adf8, length=481) at src/emu/sound/qsound.c:365 chip = (struct qsound_info ) 0x983a838 i = 16 j = 4656612 rvol = -1965839360 lvol = 7159 count = 2292600 pC = (struct QSOUND_CHANNEL ) 0x983abc0 datap = {0x7bdb828, 0x7bdddf0} Backtrace stopped: frame did not save the PC - sfar1 Program received signal SIGSEGV, Segmentation fault. 0x77f87eeb in ?? () (gdb) bt full #0 0x77f87eeb in ?? () No symbol table info available. #1 0x7800efdf in fcloseall () from C:\WINNT\system32\msvcrt.dll No symbol table info available. #2 0x0120fe57 in qsound_update (param=0x7bddc00, inputs=0x0, buffer=0x983a948, length=481) at src/emu/sound/qsound.c:365 chip = (struct qsound_info ) 0x7bddc00 i = 16 j = 4656612 rvol = -1965839360 lvol = 7159 count = 2292600 pC = (struct QSOUND_CHANNEL ) 0x7bddf88 datap = {0x7bdedf8, 0x7be13c0} Backtrace stopped: frame did not save the PC
Github Commit

Flags
Regression Version	0.126u1
Affected Sets / Systems	All games in cps2.c

Attached Files	fix.diff (1,282 bytes) Jul 25, 2008, 06:27 [Show Content]
Attached Files	qsound.diff (350 bytes) Jul 27, 2008, 20:49 [Show Content]

No.01783 Mamesick Senior Tester Jul 25, 2008, 06:28	Confirmed. Something has been broken in qsound.c I attached a temporary fix, maybe who broke it has a better one.
No.01787 Fujix Administrator Jul 25, 2008, 07:50	My 0.126u2 doesn't have the problem. hmm.
No.01789 john_iv Senior Tester Jul 25, 2008, 08:22	Non-repro also in MameUI64 .126.2.
No.01790 etabeta Developer Jul 25, 2008, 09:00 edited on: Jul 25, 2008, 09:03	not confirmed on my SDLMAME 126u2 (with u1 OSD layer), but it seems Tafoid was able to reproduce it http://www.mameworld.info/ubbthreads/showthreaded.php?Cat=&Number=158687
No.01792 Mamesick Senior Tester Jul 25, 2008, 13:44	Yes, it looks like it's a random thing... I tested SFA and sometimes it crashes after the game info screen, sometimes it boots perfectly.
No.01794 FrizzleFried Tester Jul 25, 2008, 14:28	To be 100% clear, NONE of the titles I have tried the parent roms of have made it to the title screen. They all start to load (black screen) then they abort back to either the front end or the command line. When I try their clones, they work fine.
No.01795 Tafoid Administrator Jul 25, 2008, 14:36 edited on: Jul 25, 2008, 15:00	I'm building a symbols build in hopes of getting some backtraces for these crashes. Mamesick listed a DIFF that you could try, FrizzleFried - but that might not be the official fix. EDIT: Set to acknowledged. This is a straight/normal build (not Athlon optimized as I've used before) - but I'm still on an Athlon. I noted OP has Athlon as well - is there a possible relation? Does Mamesick have an Athlon? I'd assume the majority of testers have pentium....
No.01796 Mamesick Senior Tester Jul 25, 2008, 15:23	I'm on a P4 3.00 GHz and I experienced the crashes.
No.01804 mahlemiut Developer Jul 25, 2008, 22:19	ddsom and sfar1 crash for me, but the others mentioned work fine. CPU is an AthlonXP 3000+.
No.01824 Tafoid Administrator Jul 27, 2008, 16:42	There is enough evidence here to label this a confirmed bug. As to why or why not people are experiencing it and some are not - It will be sorted out soon enough :)
No.01827 Firewave Senior Tester Jul 27, 2008, 19:40 edited on: Jul 27, 2008, 20:49	This is an issue with uninitialized memory (my first guess - but I found the memset...). It's an not so obvious issue. Look at this in qsound_start(): memset(chip, 0, sizeof(chip)); It does initialized the whole structure with the size of a pointer. It should be the same argument as the auto_malloc() call, which is actually the structure itself -> "sizeof(*chip)". I attached another diff to this report.
No.01836 aaron Developer Jul 29, 2008, 01:39	Good catch. I checked this in. Can someone confirm this is the real fix?
No.01837 FrizzleFried Tester Jul 29, 2008, 05:50 edited on: Jul 29, 2008, 05:50	I can confirm that gsound.diff above fixes the issue I reported. I tested 1944 and both Dungeons & Dragons games with success. Thank you!
No.01856 aaron Developer Jul 31, 2008, 08:57	Fixed by Firewave