------- Comment #21 from jakub at gcc dot gnu dot org 2010-06-25 09:38 ------- True. Most of the fields of the pointer are still uninitialized (NULLIFY only clears the data field). I'm afraid when NULLIFY doesn't clear the whole struct there will be always a possibility valgrind might complain, the tree optimizers see the comparison as not having any side-effects, not possibly trapping etc., so some pass might exchange the two comparisons somewhere. But worse case it would be just a valgrind warning, conditional jump based on uninitialized values which in the end wouldn't have any runtime visible effect.
This patch doesn't try to do anything about that, just improve unswitching such that it chooses conditions that it can usefully unswitch on (and not unswitch on always false/true conditions). -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43866
