[Pachi] A bug report

Evan Daniel evanbd at gmail.com
Thu Apr 28 17:33:47 CEST 2016


Did this occur while using tools/twogtp.py?

That script does not send pass moves to the other program. So when one
player passes, the other just sees two successive genmove commands for
the same color.

This is also a bug in Pachi; it should happily generate a move for the
color requested, regardless of non-alternating play. GTP spec:

> fails: never
> comments: Notice that ``pass'' is a valid vertex and should be returned if the engine wants to pass. Use ``resign'' if you want to give up the game. The controller is allowed to use this command for either color, regardless who played the last move.

Anyway, this commit makes twogtp handle the passes correctly; I don't
yet have a fix for Pachi but might do that soon.

https://github.com/evand/pachi/commit/8d494af1b223091dc72f57bfe332e47c8b8840a1

Evan Daniel

On Thu, Apr 14, 2016 at 11:19 AM,  <magnet_guo at sina.cn> wrote:
> Hi:
>
> As we are running pachi, a bug occurred randomly, which means might occur 20
> minutes or 20 hours or never.
>
>
> The stderr says:
>
>
> Fatal: Non-alternating play detected 1 1
>
>
> This really bothers us a lot because we plan to test for 2000 games each
> setup, and we have to reload the test by hand once the error occurred.
>
>
> What can we do to avoid it?
>
>
> Thanks!
>
>
> --------------------------------
>
> Guo
>
>
>
> _______________________________________________
> Pachi mailing list
> Pachi at v.or.cz
> http://rover.ms.mff.cuni.cz/mailman/listinfo/pachi
>


More information about the Pachi mailing list