Fix futility pruning
authorH.G.Muller <hgm@hgm-xboard.(none)>
Mon, 2 Jun 2014 17:42:38 +0000 (19:42 +0200)
committerH.G.Muller <hgm@hgm-xboard.(none)>
Mon, 2 Jun 2014 17:42:38 +0000 (19:42 +0200)
commita494f710fd8564278ece8d843573e06b05bb1ec8
tree4b96cb4579d1fa7e4211e0c5739a18b8dfc87ab2
parentb3826cf0103b476b9500318a6c6b974ebf2a14af
Fix futility pruning

Move generation was stopped at d=1 when futile victims were reached,
but after deepening to d>=2 the victim for which this first happened
was never tried again. In addition, the upper-bound score was not adapted
to the estimated gain, meaning it could remain at the initial -INF,
creating the permanent impression that the current position was a checkmate.
hachu.c