I have not looked at it for years, but I vaguely recall the board state
may have had the turn backwards, maybe like it was for the other player
and fibs rejected it
On 3/16/26 7:15 PM, Øystein Schønning-Johansen wrote:
I'm not sure. I think I have to know more about the GammonBot code to
answer that question. But I notice some differences in how FIBS sends
a rawboard and how GNU Backgammon generates a rawboard for the
external player.
-Øystein
tir. 17. mars 2026 kl. 00:11 skrev Tom Moulton <[email protected]>:
Would the bug/issue found possibly cause the match resume bug the
GammonBots have?
On 3/16/26 6:08 PM, Øystein Schønning-Johansen wrote:
I finally managed to solve my issue. But it took me a few hours
of backtracking GNU Backgammon code. I am now happy again -
frustration level is getting lower.
-Øystein
man. 16. mars 2026 kl. 16:15 skrev Øystein Schønning-Johansen
<[email protected]>:
Oh! I have found the source of my problems:
https://cgit.git.savannah.gnu.org/cgit/gnubg.git/commit/play.c?id=bada21ae6315ddc49827ee789a0249d1fabcd2ef
When generating FIBSBoards for the external player, GNU
Backgammon will alway set player 0 on turn. This change dates
back to 2003 - 23 years ago. So this is why I struggle to
have an external player working both as player 0 and player 1.
I'll see if I can find a simple way to recreate the bug! (and
maybe suggest a patch).
-Øystein
tir. 24. feb. 2026 kl. 17:44 skrev Guido Flohr
<[email protected]>:
> On 22 Feb 2026, at 15:37, Øystein Schønning-Johansen
<[email protected]> wrote:
> So, has anyone else worked with these two systems, FIBS
server and external player. Guido? Did you work with
these two?
Probably not. At the time, I had written my own version
of FIBS by reverse-engineering, and I’m pretty sure that
I have implemented the behaviour of gnubg, not that of
FIBS. At least, I don’t remember nuances as you described
them here.
Cheers,
Guido
