Re: [Gnushogi-devel] gnushogi xboard support

[Yann Dirson]

On Wed, Oct 23, 2013 at 10:40:12PM +0200, Yann Dirson wrote:
> > > Note I just pushed a new patch, about this ElapsedTime/polling issues,
> > > for comment.  To win32 branch just because it does not commute to master
> > > without conflict, and will be rewritten anyway.
> > >
> > This does seem closer to the original FIONREAD stuff. I guess I should
> > have put the WIN32 polling there in the first place, except that I
> > originally had no idea that Elapsed time would already do the polling I
> > needed to get out of ponder. I am kind of used to engines not immediately
> > responding to 'quit', when they are thinking. XBoard sends a SIGKILL a
> > configurable delay after 'quit', which handsomely takes care of that
> > problem. :-) It seems that GNU Shogi mainly has this because it has a
> > 'both' mode where it plays itself, and would thus always be thinking.
> 
> In fact, before 4a9acd2e the FIONREAD version was also used for
> xshogi, despite the hidden curses dependency there.  This probably
> explains why I could not reproduce any tight loop when checking after
> 74ba6a1c6, and still I got hit yesterday by this issue.  Will push a
> real fix ASAP.

Pushed a 2-commit fix/refactoring to branch "poll" for testing, and
rebased win32+xboard+minishogi on top of that.

If you can confirm the win32 was not broken by all those successive
rebases, I'd like to push it to master ASAP.

If we had a win32/config.h too, and if win32/mkmakefile was confirmed
OK, that would be good additions as well.


With a little work on the versionning issue to get rid of that "pl0",
plus the minishogi patches in their current form, we could be ready to
issue a beta for 1.4.1.