I reported the problem to wms, and have received a reply:

"It is possible that the bug is still there. I can't recall exactly what I fixed in kgsGtp and when. Please let me know if the problem recurs wher both players are using the latest kgsGtp."

This is less than I was hoping for, but seems reasonable.  He does
not want to spend his time looking for a bug in an obsolete version
of kgsGtp.

So, for all future KGS bot tournaments, entrants should download and
use the current version of kgsGtp
(from http://www.gokgs.com/download.jsp ) shortly before the tournament.

Nick


On 04/06/2012 23:14, Jean-loup Gailly wrote:
 > Play resumed, and pachi2 captured the dead white stone, but Zen19S
passed instead of capturing the dead black stone.

Actually pachi did not capture the dead white stone, it simply passed
because it would still win
even when both C2 and A8 are considered alive. The problem is that KGS
ended the game
immediately after this one pass, instead of sending
a kgs-genmove_cleanup command to Zen.

I reported this problem to wms in Aug 2011. wms made a change in the
kgsGtp client, but not
in the server. It seems that the change did not fix the problem. Zen was
using client 3.5.8
which has the change. The change uses"I got a pass back from
kgs-genmove_cleanup" as the
indicator that the game is finished. Pachi did get that
kgs-genmove_cleanup and did pass, so
everything was fine from Pachi's point of view. But Zen did not get a
kgs-genmove_cleanup command,
so it did not get an opportunity to capture the dead black stone at C2.

I think the only way to fix the problem reliably is to require two
passes even in cleanup up mode.
Both players need an opportunity to capture all dead stones, so both
need to receive
kgs-genmove_cleanup commands until they pass. Which implies that the
game ends with two passes.

My apologies to the Zen team for thinking initially that Zen received a
kgs-genmove_cleanup
command. And congratulations for winning the tournament, this is well
deserved!

Jean-loup



--
Nick Wedd
[email protected]
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to