You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There appears to be a segfault in the variant "coffeehouse" when null moves are generated. It happens frequently on the fishnet worker I use, but is also reproducible from (most) of the same positions on fairyground. The variant definition used can be found below:
[coffeehouse:crazyhouse]
mustCapture = true
An example can be reproduced on fairyground by setting multipv to 2, hash to 256 MB, threads to 4, and an initial move for White of f2f3 then starting analysis. After reaching a search depth of around 9, a null/invalid move is reached in the second line and the engine subsequently crashes. Interestingly this particular move doesn't force a capture.
I don't have much more insight into the cause, except that raising the hash size tends make the errors less frequent. Reading through the current issues, it's possible that this is also related to #800 , since this definition inherits from ZH. Although I can't do it today, I can test a definition without inheritance soon to see if that works around the problem.
Edit: I tried inheriting directly from chess by adding the rules for Crazyhouse to the definition (see below), and it seems to have worked around the issue.
Thanks. Can you reproduce this with native Fairy-SF? So far I failed to reproduce it regardless of multipv, hash, and thread setting, even with debug=yes.
There appears to be a segfault in the variant "coffeehouse" when null moves are generated. It happens frequently on the fishnet worker I use, but is also reproducible from (most) of the same positions on fairyground. The variant definition used can be found below:
An example can be reproduced on fairyground by setting
multipv
to 2,hash
to 256 MB,threads
to 4, and an initial move for White off2f3
then starting analysis. After reaching a search depth of around 9, a null/invalid move is reached in the second line and the engine subsequently crashes. Interestingly this particular move doesn't force a capture.I don't have much more insight into the cause, except that raising the hash size tends make the errors less frequent. Reading through the current issues, it's possible that this is also related to #800 , since this definition inherits from ZH. Although I can't do it today, I can test a definition without inheritance soon to see if that works around the problem.
Edit: I tried inheriting directly from
chess
by adding the rules for Crazyhouse to the definition (see below), and it seems to have worked around the issue.The text was updated successfully, but these errors were encountered: