http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56294
--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> 2013-02-13 10:56:15 UTC --- With the debug counters I have identified the first function where the difference in behavior of intra-SRA behavior with -g vs -g0 caused the comparison failure but could not spot any differences in non-debug statements. Adding -fno-inline makes the comparison failure go away (-fno-early-inlining does not help) so at this point I suppose it's tree-inline that generates different stuff when there are references to SSA names from debug statements.
