gnugo-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] Crash: GNU Go 3.6 pre 2


From: Gunnar Farnebäck
Subject: Re: [gnugo-devel] Crash: GNU Go 3.6 pre 2
Date: Fri, 24 Sep 2004 03:17:07 +0200
User-agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/21.3 (sparc-sun-solaris2.9) MULE/5.0 (SAKAKI)

Trevor wrote:
> I just found a crash when running regress.pl on GNU Go 3.6 pre 2:

Ouch.

> perl regress.pl --goprog "..\\interface\\gnugo.exe --mode gtp --quiet
> -t -w -d0x1000" --verbose 1 -a 1

This runs fine for me on Linux, at least up to trevora.tst. As a minor
note "--quiet" is redundant nowadays.

> Here's the stack trace: (notice bogus move value passed to owl_reasons)

I suppose you mean bogus color value.

> test_owl_attack_move(int 4453113, int 37, int 38, int 0) line 4564 + 56 bytes
> owl_reasons(int 4314407) line 4680 + 48 bytes
> collect_move_reasons(int 2) line 296
> do_genmove(int *, int, float, int *) line 424

This doesn't make much sense. Apparently the color parameter passed to
collect_move_reasons() manages to change before it's passed on to
owl_reasons(). Either the backtrace is inaccurate or we have a case of
serious memory corruption.

> I believe the crash is on blunder:21, though running blunder.tst alone 
> doesn't cause the crash.  The crash is on the following DEBUG line, so the
> problem probably requires appropriate debug flags to reproduce:
> 
>     DEBUG(DEBUG_OWL, "owl: %1m attacks %1m (%s) at move %d\n",
>         pos, dr, result_to_string(DRAGON2(pos).owl_attack_code),
>         movenum+1);
> 
> I've not attempted to diagnose the problem further.

Please do, if you can. I can't reproduce the crash.

/Gunnar




reply via email to

[Prev in Thread] Current Thread [Next in Thread]