[FIXED]Revisions 488 and above crashing on emulation startup.
#1
Hi there.

I've been having a bit of an issue with Playground lately, that is the emulator crashes right after "Run CD/DVD" or "Execute", starting, as i said in the subject, from revisions 488 included up to current (514 as i'm writing here), on any Release target exes (vm and vtlb). 487 and below are fine.
Debug targets also work up to current, but are way slower, obviously.

Crashes in two different ways: either with an "unhandled Win32 exception" or simply instantly closing with no messages (this one mostly with Release vm nondev).

The emu's logfile end with these line on crashing, so no help from there:
Code:
MTGS > Thread Started, Opening GS Plugin...
MTGS > GSopen Finished.

Finally, it happens on either Express or Retail versions of VS2008 (checked myself).

Any idea please?
Rig: [email protected]/2gb DDR2-5300/MSI HD4670 512M/Intel P5K Pro/XP Pro SP3 & Win7 RC1
Reply

Sponsored links

#2
First of all, you OS name, game name, graphic plugin name. Then turn off MGTS and turn on all loggin options. And see what's happened.
Reply
#3
WinXP Pro 32.
Not related to games, happens even with CDVDNull / running the bios (Run->Execute).
Happens with any graphic plugin (gsdx, zerogs, whatever), any version.

Switched to full interpreter mode (mtgs off as well), default options in Advanced, no speedhack, full logs.
Still the same issue.
Nothing shows up at all in the logfiles.
Rig: [email protected]/2gb DDR2-5300/MSI HD4670 512M/Intel P5K Pro/XP Pro SP3 & Win7 RC1
Reply
#4
kill ur ini files and try again
Fighting for peace is like ***** for virginity ...

Failure is not an option | It is bundled with your Microsoft product
Reply
#5
No better.
Rig: [email protected]/2gb DDR2-5300/MSI HD4670 512M/Intel P5K Pro/XP Pro SP3 & Win7 RC1
Reply
#6
I had a similar problem recently, It was solved by deleting all of my plugins and downloading them anew (from official playground release, GSDX, etc. - nothing old)
Reply
#7
Nope, this didn't help either.
Rig: [email protected]/2gb DDR2-5300/MSI HD4670 512M/Intel P5K Pro/XP Pro SP3 & Win7 RC1
Reply
#8
Tried a clean rebuild (CTRL+ALT+F7)? Partial rebuild should be disabled but manually forcing it to rebuild won't hurt.
Reply
#9
Done that several times as well. No change.
Rig: [email protected]/2gb DDR2-5300/MSI HD4670 512M/Intel P5K Pro/XP Pro SP3 & Win7 RC1
Reply
#10
Have you tried one of the 3rd party builds by > or the like?
If those have the same crash problem then we've definitely isolated it to being a general coding error on my part somewhere (that somehow only exploits itself on your setup). If those builds work then it's back to troubleshooting your compilation environment.

Edit: Also, I have a hunch on this, and I think it's related to a global array not being defined as PCSX2_ALIGNED16() -- the problem is knowing which one needs to be aligned. Only arrays that need to be accessed by certain types of accelerated SSE copies need to be aligned, and the trouble is that most of the time the arrays end up being aligned to 16 bytes anyway, just because of the way the compiler aligns data by default. But every once in a while, after a set of code changes, the alignment of some arrays can shift and suddenly you have a mysterious crash that didn't exist just moments before. Wink

Additionally, Debug build tends to always align data to at least 16 bytes due to the padding it uses to trap data over/underruns. Thus explaining why Debug builds work while release modes do not (and also making it really annoying to troubleshooot, since it works in the effing debugger!).

I know memcpy_raz_ can crash the emulator on unaligned data. I'll have to see if there are any other functions that assume/relay on aligned arrays.
Jake Stine (Air) - Programmer - PCSX2 Dev Team
Reply




Users browsing this thread: 1 Guest(s)