[Bug tree-optimization/51583] New: One more missing force_gimple_operand in SRA

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51583

             Bug #: 51583
           Summary: One more missing force_gimple_operand in SRA
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: jamb...@gcc.gnu.org
        ReportedBy: jamb...@gcc.gnu.org


Created attachment 26113
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26113
Testcase

I noticed that part of my fix to PR 50622 somehow got lost which can
lead to ICEs in some obscure cases, for example when compiling the
attached testcase.  It fails on i686 and x86_64, on both the trunk and
the 4.6 branch with:

test.c: In function ‘foo’:
test.c:34:1: error: invalid rhs for gimple memory store
s$u$v

t.u.v;

# .MEM_20 = VDEF <.MEM_14>
s$u$v = t.u.v;

test.c:34:1: internal compiler error: verify_stmts failed

I am about to bootstrap and test a fix.