Hangs with anything other than pure interpreter in Majora's Mask

[blargity]

I realize this has been discussed a bit in the 0.4 days, but I thought I'd mention it here with active development occurring on the 64bit branch.

Majora's Mask hangs, usually after only a few minutes with anything other than pure interpreter enabled. I know Richard42 was primarily interested in segfaults - though I'm not sure of Mupen's architecture, is it possible that the emulator process is dying in the background, and not dumping anything to stdout/stderr? I'm using the GUI frontend.

I can answer any questions about my setup. Certainly I've noticed this behavior since vanilla 0.5, so it's been around for a while. It happens regardless of what plugins I use.

 

[blargity]

Sorry for replying to myself, but upon further inspection the situation is somewhat different than I originally thought.

Both the 1.1 and SVN versions of the AMD64 branch hang within a few minutes of gameplay, not always at the same place, regardless of whether or not Dynamic Recompilation is enabled. This is a definite regression from Mupen 0.5. This is a 32bit build.

Poking around with GDB, if Dynamic Recompilation is enabled the hang seems to occur in gen_interrupt. The emulator thread is either in gen_interrupt or in a place it doesn't have debugging symbols, and since gen_interrupt() acts entirely on globals I'm not sure of the state at the time - I'm a gdb noob, sorry for the lack of info.

If you can make it to Clock Town at all then you've managed to skip past this bug.

 

[Richard42]

Thanks for reporting the bug. The dynamically recompiled code is difficult to debug. The gen*() functions are the C functions which emit the compiled x86 instructions to emulate the r4300. The hangup is probably not in the gen_interrupt() function itself.

If I can reproduce the bug I should be able to fix it. Is there anything in particular that you do in the game which seems to trigger it? I'll take a look at it this week.

 

[blargity]

I played with this bug for far longer than I really wanted to last night ;-)

The bug occurs with the pure interpreter on as well. I ran a memtest on my machine to verify it wasn't a hardware issue, and when that came up neg I would up grabbing a different copy of the ROM. The trouble vanished.

So while this still may qualify as a regression (the "broken" rom plays fine in 0.5), we may have crossed into "too damn hard to debug". Now I'm just trying to find a video plugin that will actually render most of the game ;-)

Thanks fr the quick response Richard. If you decide t try to track it down, the hang almost always occurs during the opening sequence, when you're just pressing buttons to move the cut scene forward.

 

[Richard42]

I played through a whole bunch of Majora's mask this afternoon in the debugger with the 64-bit build and the dynamic recompiler and never encountered any problems. The ROM that I played gave this data:

CRC: 5354631c 3a2def0
name: ZELDA MAJORA'S MASK
Manufacturer: Nintendo
Cartridge_ID: 535a
Country : United States
size: 4096
PC= 80080000
md5 code:2A0A8ACB61538235BC1094D297FB6556

As long as there is a ROM dump for this game which runs properly it's probably not worth the time to track down the problem. It could be related to the bad dump.

 

[azzamite]

How did you get the info?

HI, I have the same problem playing Majora's Mask.
The game ramdomly hangs specialy while I reach
the weird small tree before reaching Clock Town the
first time, and falling the clift before being converted
to a scrub. So I have to save the state every 5 minuts...

Where do you get that info about the ROM from?
I do right click on the rom and properties but nothing
of what you put appears, only
Name (blank)
Size 256.0 MBits
conuntry Europe (0x50)
Ini Code E97955C6-BC338D38-C50
MD5 (doesn't calculate) 13fab67e603b002ceaf0eea84130e973

 

[Richard42]

Where do you get that info about the ROM from?

You can run mupen64 from the console, play the game once, and then the info will be dumped out.