Bugfix in owl.c

[nando]

I was actually about to call this a simple enhancement rather than a bugfix, since it only prevents some cluttering of the SGF output. But the node counts changes

Total nodes: 1691326785 3342291 12618862
Total time: 7849.60 (7856.47)
Total uncertainty: 46.88

... suggest that it could have an actual influence on move generation (although it didn't in the regression run)

-- nando

[gunnar]

Seems weird that it should affect the node count. Might be worth investigating.

[nando]

You're right, it's weird. I'm going to dig a bit deeper...

[nando]

Ok, got it.

The semeai code in owl.c sometimes calls do_owl_attack(). Without the patch, the global variable count_variations may get incremented more than it actually should. But since OWL checks its node limits by (among others) comparing local_owl_node_counter to owl_node_limit, it doesn't matter.

Additionally, the semeai code checks its node count limits by comparing a certain node count variable to the semeai_node_limit one. And that node count global variable name is ...

Time for a local_semeai_node_counter ?

[arend]

(Only file name changed) Fixes node counting in semeai reading