RE: [gnugo-devel] A crash

[Portela Fernando]

Hi,

Arend wrote:
> Did you trace down in which situation this initialization is necessary?
> valgrind agrees it's needed, but I don't see where, and I worry that
> there is a bug lurking somewhere else.

I can remember it had to do with the use of xpos in the macro at line
3003 of readconnect.c in recursive_break(). So it was a case where no moves
were found or tried and distance was short. Possibly, the line should
actually read :

    READ_RETURN_HASH_NG(BREAK_IN, str, depth - stackp, 
                        goal_hash, move, NO_MOVE, WIN);

I'm not familiar with the code, so I fixed the problem by initializing the
variable and went to preventively do the same in the other function. The
above line is maybe the proper fix, you certainly can tell better than me.

nando