[Bug middle-end/48685] New: [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[Bug middle-end/48685] New: [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[doko at ubuntu dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

           Summary: [4.5/4.6/4.7 regression] ICE in gimplify_expr, at
                    gimplify.c:7034
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: doko@ubuntu.com


[forwarded from http://bugs.debian.org/623161]

works with 4.4.6, fails with 4.5, 4.6 branches and trunk

$ cat foo.c 
int main() {
        unsigned long int pixel;
        ((void)((1 == 2) ? ((void)(0)) : ((pixel) = 0)));
        return 0;
}

$ gcc-snapshot-20110419/build/gcc/xgcc -Bgcc-snapshot-20110419/build/gcc/ foo.c
foo.c: In function 'main':
foo.c:3:10: internal compiler error: in gimplify_expr, at gimplify.c:7034
Please submit a full bug report,

[Bug middle-end/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011.04.19 15:04:39
            Version|unknown                     |4.5.2
   Target Milestone|---                         |4.5.3
     Ever Confirmed|0                           |1
      Known to fail|                            |4.5.0

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-19 15:04:39 UTC ---
Confirmed.

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |c

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-19 15:08:12 UTC ---
Some NON_LVALUE_EXPR survives.  Probably a C constexpr fallout.

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |jsm28 at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-04-20 08:08:27 UTC ---
int
main ()
{
  int v = 1;
  (void) (1 == 2 ? (void) 0 : (v = 0));
  return v;
}

I believe normally c_fully_fold_internal drops NON_LVALUE_EXPRs in
STRIP_TYPE_NOPS at the end, but in this case the void type COND_EXPR is folded
into int type MODIFY_EXPR (and even a fold_convert wouldn't fix that up,
for MODIFY_EXPR it is just fold_ignored_result and still stays at the original
type) and thus the void type NON_LVALUE_EXPR is kept.
I wonder whether we should be creating NON_LVALUE_EXPRs with void type at all
(aren't void values never lvalues?), or when handling NON_LVALUE_EXPR in
c_fully_fold_internal we should drop them if they are void type.

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[joseph at codesourcery dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

--- Comment #4 from joseph at codesourcery dot com <joseph at codesourcery dot com> 2011-04-21 15:36:26 UTC ---
The C front end should not depend on fold creating NON_LVALUE_EXPRs 
anywhere, but I don't know if C++ has such dependencies.  Where does the 
NON_LVALUE_EXPR come from?  In any case, I don't think NON_LVALUE_EXPR 
should be used for type conversions; a NON_LVALUE_EXPR with void type 
should only ever be used if the contained expression has void type (with 
the same qualifiers), and under those conditions STRIP_TYPE_NOPS should 
remove it.

c_fully_fold_internal needs to preserve the type of the expression, so 
can't just remove a NON_LVALUE_EXPR that changes the type without adding 
some other conversion to the new type - but as noted above, if a 
NON_LVALUE_EXPR is changing the type I think the creation of that 
NON_LVALUE_EXPR is what's wrong.

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-04-22 11:42:59 UTC ---
The NON_LVALUE_EXPR in question is created by the C FE, in build_c_cast:
4712  /* Don't let a cast be an lvalue.  */
4713  if (value == expr)
4714    value = non_lvalue_loc (loc, value);

And it has void_type_node like its argument, which is void_type_node COND_EXPR.
The second argument of the COND_EXPR is void_type_node also, but the third is
a MODIFY_EXPR and for MODIFY_EXPR fold_convert_loc just refuses to convert it:
2030    case VOID_TYPE:
2031      tem = fold_ignored_result (arg);
2032      if (TREE_CODE (tem) == MODIFY_EXPR)
2033        goto fold_convert_exit;
2034      return fold_build1_loc (loc, NOP_EXPR, type, tem);

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[joseph at codesourcery dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

--- Comment #6 from joseph at codesourcery dot com <joseph at codesourcery dot com> 2011-04-22 16:53:53 UTC ---
On Fri, 22 Apr 2011, jakub at gcc dot gnu.org wrote:

> And it has void_type_node like its argument, which is void_type_node COND_EXPR.
> The second argument of the COND_EXPR is void_type_node also, but the third is
> a MODIFY_EXPR and for MODIFY_EXPR fold_convert_loc just refuses to convert it:
> 2030    case VOID_TYPE:
> 2031      tem = fold_ignored_result (arg);
> 2032      if (TREE_CODE (tem) == MODIFY_EXPR)
> 2033        goto fold_convert_exit;
> 2034      return fold_build1_loc (loc, NOP_EXPR, type, tem);

I'd say that's a bug in fold_convert_loc; it should provide the interface 
to the rest of the compiler that the return value always has the requested 
type (or is error_mark_node).

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-04-22 19:20:38 UTC ---
Seems that was introduced by:
http://gcc.gnu.org/ml/gcc-patches/2007-01/msg00279.html
but we don't have GIMPLE_MODIFY_STMT any longer and MODIFY_EXPR does have a
type, so I guess we can just revert it.  Will test that.

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-04-23 10:18:06 UTC ---
Author: jakub
Date: Sat Apr 23 10:18:03 2011
New Revision: 172895

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=172895
Log:
    PR c/48685
    * fold-const.c (fold_convert_loc): Add NOP_EXPR when casting
    to VOID_TYPE even around MODIFY_EXPR.

    * gcc.dg/pr48685.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/pr48685.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/fold-const.c
    trunk/gcc/testsuite/ChangeLog

[Bug c/48685] [4.5/4.6/4.7 regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-04-23 10:18:56 UTC ---
Author: jakub
Date: Sat Apr 23 10:18:54 2011
New Revision: 172896

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=172896
Log:
    PR c/48685
    * fold-const.c (fold_convert_loc): Add NOP_EXPR when casting
    to VOID_TYPE even around MODIFY_EXPR.

    * gcc.dg/pr48685.c: New test.

Added:
    branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/pr48685.c
Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/fold-const.c
    branches/gcc-4_6-branch/gcc/testsuite/ChangeLog

[Bug c/48685] [4.5 Regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |jakub at gcc dot gnu.org
                   |gnu.org                     |
            Summary|[4.5/4.6/4.7 regression]    |[4.5 Regression] ICE in
                   |ICE in gimplify_expr, at    |gimplify_expr, at
                   |gimplify.c:7034             |gimplify.c:7034

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-04-23 10:20:08 UTC ---
Fixed for 4.6+ so far.

[Bug c/48685] [4.5 Regression] ICE in gimplify_expr, at gimplify.c:7034

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.5.3                       |4.5.4

--- Comment #11 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-28 14:51:56 UTC ---
GCC 4.5.3 is being released, adjusting target milestone.

[Bug c/48685] [4.5 Regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-05-03 16:33:15 UTC ---
Author: jakub
Date: Tue May  3 16:33:09 2011
New Revision: 173324

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=173324
Log:
    Backport from mainline
    2011-04-23  Jakub Jelinek  <jakub@redhat.com>

    PR c/48685
    * fold-const.c (fold_convert_loc): Add NOP_EXPR when casting
    to VOID_TYPE even around MODIFY_EXPR.

    * gcc.dg/pr48685.c: New test.

Added:
    branches/gcc-4_5-branch/gcc/testsuite/gcc.dg/pr48685.c
Modified:
    branches/gcc-4_5-branch/gcc/ChangeLog
    branches/gcc-4_5-branch/gcc/fold-const.c
    branches/gcc-4_5-branch/gcc/testsuite/ChangeLog

[Bug c/48685] [4.5 Regression] ICE in gimplify_expr, at gimplify.c:7034

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48685

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

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

--- Comment #13 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-05-03 17:17:08 UTC ---
Fixed.