Re: [gnugo-devel] semeai tracing fixes

[bump]

> Dan, I assume when update_status() is called from new_semeai, it should
> set the safety to "ALIVE_IN_SEKI" not "ALIVE", right? (See below.) --
> That's what originally started confusing me.

Yes, I guess that's right.

I think it may be an improvement if we have new_semeai() never
change the status of a dragon from CRITICAL to something else.
This would at least prevent GNU from tenuki'ing in a dangerous
situation.

Such a change has the following breakage:

./regress.sh . neurogo.tst 
11 unexpected PASS!
./regress.sh . lazarus.tst 
15 unexpected FAIL: Correct 'Q15|T5|H5|H3', got 'S15'
./regress.sh . nngs.tst 
820 unexpected FAIL: Correct 'J13|L9', got 'D14'
./regress.sh . 13x13.tst 
39 unexpected FAIL: Correct 'H4|J4', got 'G3'
./regress.sh . ninestones.tst 
500 unexpected PASS!
510 unexpected FAIL: Correct 'E19', got 'A8'

Although this is not impressive, consider.

lazarus:15 seems to be a caching problem. This doesn't happen
if the test lazarus:14 is omitted.

In nngs:820, D14 is relevant to the semeai between the
two dragons in the middle left. It is not clear that D14
is not the right move in this position. Perhaps the test
should be rewritten using restricted_genmove.

In 13x13:39, G3 is a bad move.

Although this change breaks ninestones:510 we would not
be in that position if A8 had been played earlier at the
position in ninestones:500.

Dan