Re: [gnugo-devel] persistent reading cache inconsistencies

[Evan Berggren Daniel]

On Mon, 30 Sep 2002, Gunnar Farneback wrote:

> Evan wrote:
> > attack_and_defend has a comment that talks about inconsistencies
> > introduced by the reading cache.  I think I have a patch to fix it (if
> > it's inconsistent, redo the reading).  How do I go about reproducing such
> > an error to check if it works?
>
> The only way I can see is to set up an assertion somewhere in the code
> where an inconsistency can be observed and run through the regressions
> (or replay some game, or whatever) and hope that it catches something.
> These things have been observed in the past but I don't remember where
> and it wouldn't be worth much if I did since both the reading code and
> persistent caching has changed much since them.

OK, I think I'll do that.  It would be nice to have a regression test,
though, so I'm running matcher_check and watching for differences (hence
the patch...)

I figure once I find a game, loading to move n - 5 and then doing a
sequence of gg_genmove, black x, etc should catch it.

Thanks

Evan Daniel