Re: [Bug-gnubg] Fw: new gnu bug

[Jim Segrave]

On Sun 22 Jun 2003 (00:25 +0200), Nardy Pillards wrote:
> From: "Joseph Heled" <address@hidden>
> Sent: Saturday, June 21, 2003 11:02 PM
> 
> 
> > Hugh Sconyers wrote:
> > > From: "Hugh Sconyers" <address@hidden>
> > > To: "Joern Thyssen" <address@hidden>
> > > Sent: Thursday, June 19, 2003 5:34 PM
> > >>hi
> > >>
> > >>found a bug in gnu.
> > >>
> > >>side on roll has 3-6 3-5 3-4 2-3 2-2 2-1 pc = 57
> > >>has the cube on 2 money game
> > >>
> > >>side not on roll has 1-25 pc = 25
> > >>click on evaluate
> > >>get this error
> > >>gnubg.exe
> > >>line 1706
> > >>expression men Off  <= 8
> > >>then program aborts
> > >>hugh
> > >>
> 
> > Can you send me the GNU position ID please.
> > 
> > Thanks, Joseph
> > 
> 
> pos id: AQAAbLs7AAAAAA
> match id: AQEAAAAAAAAA
> 
> (but i get pc=56 with 3on6, 3on5, 3on4, 2on3, 2on2 and 2on1)
                         18 + 15 + 12    + 6   +4       +2
That adds up to 57
                        
> and no error

I get a core dump under FreeBSD with a devel branch build from about a week
ago.

The interesting part of the backtrace:

#0  0x28782b58 in kill () from /usr/lib/libc.so.4
#1  0x287c40d2 in abort () from /usr/lib/libc.so.4
#2  0x287a0103 in __assert () from /usr/lib/libc.so.4
#3  0x08069397 in menOffNonCrashed (anBoard=0x678e, afInput=0xbfbfcee0)
    at eval.c:1718
#4  0x080693c5 in CalculateContactInputs (anBoard=0xbfbfd0c0,
    arInput=0xbfbfcbc0) at eval.c:1734
#5  0x0806c015 in DumpContact (anBoard=0xbfbfd0c0, szOutput=0xbfbfd1a5 "",
    bgv=VARIATION_STANDARD) at eval.c:3748
#6  0x0806c45e in DumpPosition (anBoard=0xbfbfd0c0,
    szOutput=0xbfbfd190 "Evaluator: \tCRASHED\n\n", pec=0x815a9c4,
    pci=0xbfbfd080, fOutputMWC=0, fOutputWinPC=0, fOutputInvert=0)
    at eval.c:3907
#7  0x0806045b in CommandEval (sz=0x9f45c54 "") at gnubg.c:3011

menOff is 14 in menOffNonCrashed (which is reasonable)

ClassifyPosition() has decided this is a contact position at the top of
DumpPosition(), but the contact evaluater doesn't like having that many men
off.

This is not unique - for an experiment, I started a game, gnubg made a
first move. I then edited the position by removing 9 of gnubg's pieces
(chosen arbitrarily). It was then my move and I asked for an evaluation but
got a core dump.

-- 
Jim Segrave           address@hidden