[Bug tree-optimization/58946] [4.9 Regression] internal compiler error: in operator[], at vec.h:722
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58946 --- Comment #4 from Jakub Jelinek jakub at gcc dot gnu.org --- Author: jakub Date: Mon Nov 4 10:29:42 2013 New Revision: 204348 URL: http://gcc.gnu.org/viewcvs?rev=204348root=gccview=rev Log: PR tree-optimization/58946 * tree-ssa-reassoc.c (maybe_optimize_range_tests): Update all bbs with bbinfo[idx].op != NULL before all blocks with bbinfo[idx].op == NULL. * gcc.c-torture/compile/pr58946.c: New test. Added: trunk/gcc/testsuite/gcc.c-torture/compile/pr58946.c Modified: trunk/gcc/ChangeLog trunk/gcc/testsuite/ChangeLog trunk/gcc/tree-ssa-reassoc.c
[Bug tree-optimization/58946] [4.9 Regression] internal compiler error: in operator[], at vec.h:722
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58946 Jakub Jelinek jakub at gcc dot gnu.org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |FIXED --- Comment #5 from Jakub Jelinek jakub at gcc dot gnu.org --- Fixed.
[Bug tree-optimization/58946] [4.9 Regression] internal compiler error: in operator[], at vec.h:722
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58946 octoploid at yandex dot com changed: What|Removed |Added CC||pinskia at gcc dot gnu.org --- Comment #1 from octoploid at yandex dot com --- Started with r204194.
[Bug tree-optimization/58946] [4.9 Regression] internal compiler error: in operator[], at vec.h:722
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58946 --- Comment #2 from Andrew Pinski pinskia at gcc dot gnu.org --- (In reply to octoploid from comment #1) Started with r204194. Yes I expected failures to happen in reassoc due it not being well tested before and now my patch is forcing the path inside reassoc being a lot more than before. Most likely you could get a test case that exposes this before my patch.
[Bug tree-optimization/58946] [4.9 Regression] internal compiler error: in operator[], at vec.h:722
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58946 Jakub Jelinek jakub at gcc dot gnu.org changed: What|Removed |Added CC||jakub at gcc dot gnu.org Target Milestone|--- |4.9.0 --- Comment #3 from Jakub Jelinek jakub at gcc dot gnu.org --- This is a bug in my reassoc changes, update_ops uses is_reassociable_op and has_single_use predicates and expects to match exactly what get_ops did, but if some SSA_NAME has say two imm uses during get_ops, one of those could be gone because of earlier update_ops or GIMPLE_COND changes processing and then update_ops can recurse on something get_ops has not. Will think about it tomorrow.