------- Comment #2 from jakub at gcc dot gnu dot org 2006-11-13 10:16 ------- Created an attachment (id=12601) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12601&action=view) gcc43-pr29581-tests.patch
Testcases that show this bug even on the trunk (taken from vect-8{5,6,7,8}.c, removed vector related stuff from it, added -O2 -ftree-loop-linear). With my preliminary fix for this PR all these seem to be fixed (on a release only checking gcc), unfortunately with checking enabled I'm getting tree checking failures like: error: statement makes a memory store, but has no V_MAY_DEFS nor V_MUST_DEFS perfecttmp.42 = perfectiv.40_5 + 16; I called update_stmt, o not sure what I'm still missing. BTW, with -ftree-loop-linear also fortran forall_1.f90 and scalarize.f90 are miscompiled, but that problem isn't fixed by my patch, so perhaps there are other bugs. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29581