Re: [Bug-gnubg] Command-line bug ?

[Christian Anthon]

Sounds like a more educated fix, however, it doesn't quite work:

You have already rolled the dice.
  1. Cubeful 0-ply    2/af 1/af                    Eq.:  +0,722
     0,861 0,000 0,000 - 0,139 0,000 0,000
      0-ply cubeful prune [ekspert]
  2. Cubeful 0-ply    2/1 2/af                     Eq.:  -1,000 ( -1,722)
     0,144 0,000 0,000 - 0,856 0,000 0,000
      0-ply cubeful prune [ekspert]

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread -1220225344 (LWP 9941)]
0x080a4426 in RolloutGeneral (apBoard=0xbfd300e0, apOutput=0xbfd300c0,
apStdDev=0xbfd300a0, aarsStatistics=0x0,
  apes=0xbfd30080, apci=0xbfd30060, apCubeDecTop=0xbfd30040,
alternatives=2, fInvert=1, fCubeRollout=0,
  pfProgress=0x80e33b1 <RolloutProgress>, pUserData=0x8846488) at
rollout.c:1095
1095        if ((pes->et != EVAL_ROLLOUT) || (prc->nGamesDone == 0)) {
(gdb) bt
#0  0x080a4426 in RolloutGeneral (apBoard=0xbfd300e0,
apOutput=0xbfd300c0, apStdDev=0xbfd300a0, aarsStatistics=0x0,
  apes=0xbfd30080, apci=0xbfd30060, apCubeDecTop=0xbfd30040,
alternatives=2, fInvert=1, fCubeRollout=0,
  pfProgress=0x80e33b1 <RolloutProgress>, pUserData=0x8846488) at
rollout.c:1095
#1  0x0807b3ce in CommandRollout (sz=0xbfd30f39 "") at gnubg.c:4511
#2  0x08076b80 in HandleCommand (sz=0xbfd30f34 "=1", ac=0x8281b40) at
gnubg.c:3019
#3  0x08076f21 in LoadCommands (pf=0x884b1b8, szFile=0x8847c66 "com")
at gnubg.c:4742
#4  0x080770b7 in CommandLoadCommands (sz=0x8847c66 "com") at gnubg.c:4825
#5  0x08076b80 in HandleCommand (sz=0x8847c66 "com", ac=0x8284fe0) at
gnubg.c:3019
#6  0x08076d82 in HandleCommand (sz=0x8847c5d "commands",
ac=0x8281b40) at gnubg.c:3023
#7  0x0807c90d in main (argc=1, argv=0xbfd34574) at gnubg.c:7798

Out of time for now.

Christian.

On 8/23/06, Jim Segrave <address@hidden> wrote:
> On Wed 23 Aug 2006 (13:58 +0100), Jonathan Kinsey wrote:
> > address@hidden wrote:
> > [snip]
> > > On the bug-tracker there's a patch proposed by Michel Delpreli : Jon
> > > said it didn't solve the problem, but if you understand the intended
> > > modification maybe you can get a clue at what's happening ...
> >
> > I've just retried the patch and it still fails for me, here's the
> > command file I ran and the output I got (entering "2 1" as the roll):
> >
> > ------cmd file--------------
> >
> > set rng man
> > set player 0 human
> > set player 1 human
> > new session
> > set board GwAAYAMAAAAAAA
> > roll
> > hint
> > rollout =1 =2
> >
> > ------gnubg output----------
> I've just committed a fix for this. In gnubg.c, CommandRollout assumed
> that the moves in the movelist already had an evaluation context and,
> worse, that the context was a rollout context as opposed to something
> else.
>
> It now copies the current rollout context into each move being
> evaluated and correctly gives -1 as the cubeful equity of the
> ill-advised move.
>
> --
> Jim Segrave           address@hidden
>
>
>
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-gnubg