Re: [gnugo-devel] Last night's CVS (14 FAILS 14 PASSES)

[Arend Bayer]

> There seem to be a few platform specific differences, still.
> See, for example:
> http://www.public32.com/regress/2002-01-29/?unexpected=1
> 
> Problem Status 
> global:14 FAILED 
> trevor:200 PASSED 
> 
> This build should have had no unexpected results.

Are you both running with default cache size? I once observed a couple of
unexpected results with cache size set to 96 (see
http://mail.gnu.org/pipermail/gnugo-devel/2001-December/000830.html,
http://mail.gnu.org/pipermail/gnugo-devel/2001-December/000843.html).

Btw, isn't this cache size dependence a behaviour to be expected due to he
owl node limit? As far as I can tell, the number of variations is not
increased when an do_owl_attack/defense position is found in the cache.

Arend