GenoM wrote:Strange, is it only on my PC ZCT is hanging in memory eating CPU resources after the game ends? Another problem: under Arena 1.1 I failed to run 2cpu-version...
PS:
there is the same problem (without hanging in memory) with Jim's builds of 22.1th Crafty too...
Hard to say what is happening with the hanging, unless it is SMP and the engine is still loaded (the idle processors will sit there waiting for work, taking up CPU time). Under my tests (if you disregard the few bugs that I have recently added and then fixed), using Xboard and non-SMP, ZCT has been 100% stable and well behaved.
Zach Wegner wrote:Hard to say what is happening with the hanging, unless it is SMP and the engine is still loaded (the idle processors will sit there waiting for work, taking up CPU time).
Zach Wegner wrote:Thanks for the compiler output, Fonzy. It did help me catch a couple subtle bugs, but nothing pertaining to bookmaking. I saw that you had some more bookmaking output before between debug mode and release, and then edited it out. Could you post that again, or something like it? That looked like a genuine bug...
While stepping in debug mode I noticed that the pgn moves were just skipped.
The line "notation coord" in zct.ini does not work.
It has to be "notation coordinates" else it treats input as SAN.
F. Bluemers wrote:While stepping in debug mode I noticed that the pgn moves were just skipped.
The line "notation coord" in zct.ini does not work.
It has to be "notation coordinates" else it treats input as SAN.
Hehe, actually the reverse is true. When it's in coord mode, I realize it won't read PGNs correctly because it's looking for coordinate moves. I suppose this is a bug and I should switch to SAN mode when reading PGNs. But if you put "coordinates" it won't recognize it and continue in SAN mode. It will still accept coordinate moves from pawns (as they don't require a piece specifier in SAN), but you can see that it prints moves in SAN.
Also, you don't need to use that line anymore, it was a bug workaround for UCI mode. Of course you still can if you like coord input.
Hmm.. I don't know what is going on there. Probably something to do with the Windows-Unix compatibility stuff. I guess it is running under Cygwin? In any case, it is just not killing the child processes.
Zach Wegner wrote:The book bugs are fixed, and ZCT now accepts unlimited input.
Source updated at SF.
thanks
applied a little fix in input.c getln():
realloc(buffer, size);=>buffer=realloc(buffer, size);
in globals.h extern char *input_buffer;//[BUFSIZ];
and the fix in http://64.68.157.89/forum/viewtopic.php ... dee0071057