Every test in SPEC CPU2000 causes cc1 to segfault when built with:
-O2
-fno-tree-ch
-fno-tree-copy-prop
-fno-tree-dominator-opts
-fno-tree-store-copy-prop
-fno-tree-vrp
Yes, a normal person wouldn't do that, but in theory we'd like the optimization
passes to not rely on each other. Minimizing a test case proved to be
frustrating because even a change in whitespace could cause a compile to
suddenly succeed. The following tests in gcc/testsuite/gcc.dg fail in the same
way with these options for powerpc-linux. I tried one of them with 24 cross
cc1s and most failed in the same way.
20000906-1.c
20020122-2.c
20020426-2.c
20040916-1.c
20050111-2.c
20050702-1.c
20070507-1.c
The failure is in tree-ssa-pre.c:2002 with i=20, postorder[i]=283741360.
--
Summary: cc1 segfaults when multiple tree opts disabled
Product: gcc
Version: 4.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: janis at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33649
