The following FAILs appeared on mainline on hppa64-hpux between 20050408 and 20050409. FAIL: gcc.c-torture/execute/931004-2.c execution, -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions FAIL: gcc.c-torture/execute/931004-2.c execution, -O3 -fomit-frame-pointer -funroll-loops (The test continues to pass at other optimization levels.)
I see this too, for cris-elf, but for 931004-6.c as well. I'm going to assume it's the same bug (and relevant in this PR) for two reasons: due to the failure being at -O3: FAIL: gcc.c-torture/execute/931004-2.c execution, -O3 -fomit-frame-pointer -funroll-loops FAIL: gcc.c-torture/execute/931004-2.c execution, -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions FAIL: gcc.c-torture/execute/931004-6.c execution, -O3 -fomit-frame-pointer -funroll-loops FAIL: gcc.c-torture/execute/931004-6.c execution, -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions (the gcc.log says "program stopped with signal 6." hence an abort call) and because of the timeframe for the appearance of the failure: the tests worked with LAST_UPDATED "Fri Apr 8 17:09:07 UTC 2005" and started failing with "Sat Apr 9 01:52:04 UTC 2005".
Appeared on hppa64-hpux between 2005-04-09 01:34 UTC and 2005-04-09 01:38 UTC. I.e., caused by tree-cleanup-branch merge.
Started to pass again on hppa64-hpux between 20050422 and 20050423, could it be confirmed whether they pass again on cris-elf as well?
Still failed with LAST_UPDATED "Fri Apr 22 23:44:55 UTC 2005". Confirmed to pass with "Sat Apr 23 12:51:56 UTC 2005".
Most likely fixed by: +2005-04-22 Diego Novillo <dnovillo@redhat.com> + + * tree-cfg.c (tree_duplicate_bb): Call create_new_def_for + for newly created PHI nodes. + Call create_new_def_for for every new V_MAY_DEF and + V_MUST_DEF on the copied statement. + (struct ssa_name_map_entry): Remove. + (ssa_name_map_entry_hash): Remove. + (ssa_name_map_entry_eq): Remove. + (allocate_ssa_names): Remove. + (rewrite_to_new_ssa_names_def): Remove. .... Then. Closing as fixed.