[Bug tree-optimization/45991] New: [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[Bug tree-optimization/45991] New: [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45991

           Summary: [4.6 Regression] ICE: verify_stmts failed: Invalid
                    address operand in in TARGET_MEM_REF. with
                    -fstrict-overflow
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: zsojka@seznam.cz


Created attachment 22024
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22024
reduced testcase

Command line:
$ gcc -O -fstrict-overflow pr45991.c

Compiler output:
$ gcc -O -fstrict-overflow pr45991.c 
pr45991.c: In function 'foo':
pr45991.c:8:6: error: Invalid address operand in in TARGET_MEM_REF.
# VUSE <.MEM_25>
b_10 = MEM[symbol: A[20], index: ivtmp.9_20, offset: 0B];

pr45991.c:8:6: internal compiler error: verify_stmts failed
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Tested revisions:
r165354 - crash
r163636 - OK
4.5 r163761 - OK

[Bug tree-optimization/45991] [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45991

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2010.10.13 04:29:26
                 CC|                            |spop at gcc dot gnu.org
   Target Milestone|---                         |4.6.0
     Ever Confirmed|0                           |1

--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> 2010-10-13 04:29:26 UTC ---
It is caused by revision 164790:

http://gcc.gnu.org/ml/gcc-cvs/2010-09/msg01089.html

[Bug tree-optimization/45991] [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45991

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug tree-optimization/45991] [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45991

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-04 12:17:02 UTC ---
Mine.

[Bug tree-optimization/45991] [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45991

--- Comment #3 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-04 12:29:20 UTC ---
The problem is that fold figures sth out that tree-affine doesn't.  The
latter doesn't simplify

{
  type = long unsigned int
  offset = 0
  elements = {
    [0] = &A[(long unsigned int) (long unsigned int) ((i_3(D) + -1) * 16) + 20]
* 1, 
    [1] = (long unsigned int) ((i_3(D) + -1) * 16) *
340282366920938463463374607431768211455, 
    [2] = ivtmp.9_20 * 1
  }
}

to just &A[20] + ivtmp.9_20.

Now we can avoid the ICE by gimplifying the address properly which we don't
because of a typo I made in an earlier change.

[Bug tree-optimization/45991] [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45991

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-04 14:07:54 UTC ---
Author: rguenth
Date: Thu Nov  4 14:07:48 2010
New Revision: 166312

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166312
Log:
2010-11-04  Richard Guenther  <rguenther@suse.de>

    PR tree-optimization/45991
    * gimplify.c (force_gimple_operand_1): Use the provded test
    function in the initial test.

    * gcc.dg/pr45991.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/pr45991.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/gimplify.c
    trunk/gcc/testsuite/ChangeLog

[Bug tree-optimization/45991] [4.6 Regression] ICE: verify_stmts failed: Invalid address operand in in TARGET_MEM_REF. with -fstrict-overflow

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45991

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED

--- Comment #5 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-04 14:08:32 UTC ---
Fixed.