Re: [Bug-gnubg] Anomolies in threaded code

[Jonathan Kinsey]

Jim Segrave wrote:
> Running under gentoo linux 2.6.18 on an Intel Core Duo in 64 bit mode.
> 
> I built gnubg from the sources as of 23:00 GMT 17 Feb. Analysing a
> match with about 760 moves goes almost exactly twice as fast as with
> threads set to one.
> 
> However:
> 
> In both gui and cli modes, the 'Analysing Move' count shows very odd
> values - this being a sample. I've seen a digit slip past the '/',
> turning a match with 400 or so moves into a match with 1400 or so
> according to this display.
> 
> Analysing match; move:
> Analysing match; move: 1/767
> Analysing match; move: 2/767
> Analysing match; move: 3/767
[snip]
> Analysing match; move: 15/767
> Analysing match; move: 17/767

I wonder if the progress is getting displayed from somewhere else than
UpdateProgressBar() (on linux)?

> I sometimes encounter this message, I can get it when simply starting
> gnubg in CLI mode and giving the quit command. I believe at least once
> it terminated a match analysis half way through:
> 
> GThread-ERROR **: file gthread-posix.c: line 160 (): error 'Device or 
> resource busy' during 'pthread_mutex_destroy ((pthread_mutex_t *) mutex)'
> aborting...
> Trace/breakpoint trap

Strange, In MT_Close() the threads should be destroyed before the
mutexes - a trace with the line might help track down what is causing
the error message.

Jon

signature.asc
Description: OpenPGP digital signature