On Fri, Feb 25, 2011 at 3:12 PM, Simon Waters
<address@hidden> wrote:
I don't think there is a lot of dependency on file I/O in the GNU Chess
5.08 code base. Opening book code is the main one (which for most
opponents can simply be omitted - which you can find in the code as
there is a "book off" option that uses a flag throughout to do the right
thing).
I was able to successfully do an initial compile and run it with the Native Client sel_ldr tool (which allows for running a command-line version). I didn't need to do anything with the file I/O, although I had to run without any opening book. Similar projects have used a technique where they hard-code the file as a C-language data structure and just directly access that as though it were a file. Eventually we should have
We have moved our attention to a code base derived from Fabien's Fruit
chess engine.
Do this mean that there will be a gnuchess version 6 coming at some point?
You probably want to focus on the Winboard/Xboard chess interface aspect
as in that mode the code should flush standard out, and talk a
(reasonably) well defined chess language which would make using the JS
front end with other chess engines in future a lot easier.
I agree that using the xboard interface is a smart move.
Main dependency headache I can imagine is the code using threading for
move input. You can probably find the version before that in the
changelog, but a lot of changes have happened since that was
implemented, but it might be side-steppable if that is an issue.
I think I probably ran into some threading issues on my initial attempt. gnuchess worked, but it didn't search very deep at all. A normal build of gnuchess works for maybe 5 seconds per move, but the Native Client build essentially moved instantly...
Cheers,
-Matt