Re: [gnugo-devel] Semeai

[Daniel Bump]

> > After a lot of mucking around I convinced myself that
> > the Hashnode next pointer was getting trashed and it seemed
> > the function hashtable_partially_clear() itself was
> > responsible. So I changed one line of that function to:
> > 
> >     hashtable_unlink_closed_results(node, 
> >                           (1 << OWL_ATTACK |
> >                           1 << OWL_DEFEND  |
> >                           1 << SEMEAI), 3, statistics);
> > and the crashes seem to have stopped. I'm most of the
> > way through the third regression batch (running strategy3.tst
> > right now).
> 
> This doesn't make all that much sense to me. It would be valuable to
> know exactly where the pointer got trashed.

I tried to figure out where the pointer got trashed but
didn't pin it down. But I believe it happened when
hashtable_unlink_closed_results was called from
hashtable_partially_clear.

> It's necessary to also bump MAX_ROUTINE. Otherwise there may be
> indexing out of bounds and noone knows what might happen. I've added
> an assertion in hashtable_unlink_closed_results() to detect this
> problem. It wouldn't surprise me if this was in fact the cause for the
> crashes.

In other words, MAX_ROUTINE should be SEMEAI. Perhaps you are
right, that was the problem.

> So why don't we implement support for two strings then? The
> connection reading will need it too.

I put up the revised (noncrashing) version of my patch
at devel.html. But I suppose it will need to be revised
again using your hashing patch.

Dan