[Bug libstdc++/52215] New: [4.7 Regression] FAIL: ext/profile/mutex_extensions_neg.cc (test for excess errors)

[Bug libstdc++/52215] New: [4.7 Regression] FAIL: ext/profile/mutex_extensions_neg.cc (test for excess errors)

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

             Bug #: 52215
           Summary: [4.7 Regression] FAIL:
                    ext/profile/mutex_extensions_neg.cc (test for excess
                    errors)
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dominiq@lps.ens.fr
                CC: bkoz@redhat.com, jwakely.gcc@gmail.com


Between revisions 184107 (OK) and 184111 (see
http://gcc.gnu.org/ml/gcc-regression/2012-02/msg00033.html and
http://gcc.gnu.org/ml/gcc-regression/2012-02/msg00035.html , likely r184110),
the test ext/profile/mutex_extensions_neg.cc has started to fail with -m32. The
error is

FAIL: ext/profile/mutex_extensions_neg.cc (test for excess errors)
Excess errors:
/opt/gcc/build_w/x86_64-apple-darwin10.8.0/i386/libstdc++-v3/include/profile/impl/profiler_state.h:52:28:
error: conversion from 'enum __gnu_profile::__state_type*' to 'enum
__gnu_profile::__state_type' [-fpermissive]

[Bug libstdc++/52215] [4.7 Regression] FAIL: ext/profile/mutex_extensions_neg.cc (test for excess errors)

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-02-12 13:57:32 UTC ---
It makes no sense unless the signature of __atomic_compare_exchange_n is
different when -m32 is used!

Looking into it ...

[Bug c++/52215] [4.7 Regression] FAIL: ext/profile/mutex_extensions_neg.cc (test for excess errors)

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-02-12
                 CC|                            |amacleod at redhat dot com
          Component|libstdc++                   |c++
     Ever Confirmed|0                           |1

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-02-12 14:13:38 UTC ---
Confirmed, this works with -m64 but fails with -m32

enum E { EE };

int main()
{
  E e = EE;
  __atomic_compare_exchange_n(&e, &e, e, true, __ATOMIC_ACQ_REL,
__ATOMIC_RELAXED);
}


cas.cc:6:82: error: conversion from ‘enum E*’ to ‘enum E’ [-fpermissive]

[Bug c++/52215] __atomic_compare_exchange_n for enumeration type changes signature with -m32

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |rejects-valid
                 CC|jwakely.gcc at gmail dot    |redi at gcc dot gnu.org
                   |com                         |
            Summary|[4.7 Regression] FAIL:      |__atomic_compare_exchange_n
                   |ext/profile/mutex_extension |for enumeration type
                   |s_neg.cc (test for excess   |changes signature with -m32
                   |errors)                     |

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-02-12 14:23:54 UTC ---
It's the second parameter which has the wrong type, this works with -m32:

  __atomic_compare_exchange_n(&e, e, e, true, __ATOMIC_ACQ_REL,
      __ATOMIC_RELAXED);

[Bug c++/52215] __atomic_compare_exchange_n for enumeration type changes signature with -m32

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hp at gcc dot gnu.org

--- Comment #4 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-02-12 20:48:25 UTC ---
Presumably universal for ilp32 machines, at least for cris-elf too.

[Bug c++/52215] [4.7 Regression] __atomic_compare_exchange_n for enumeration type changes signature with -m32

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rth at gcc dot gnu.org
   Target Milestone|---                         |4.7.0
            Summary|__atomic_compare_exchange_n |[4.7 Regression]
                   |for enumeration type        |__atomic_compare_exchange_n
                   |changes signature with -m32 |for enumeration type
                   |                            |changes signature with -m32

--- Comment #5 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-02-13 11:28:04 UTC ---
DEF_SYNC_BUILTIN (BUILT_IN_ATOMIC_COMPARE_EXCHANGE_N,
                  "__atomic_compare_exchange_n",
                  BT_FN_VOID_VAR, ATTR_NOTHROW_LEAF_LIST)

thus it is a variadic function.  But sync_resolve_params () does

      /* Only convert parameters if the size is appropriate with new format
         sync routines.  */
      if (orig_format
          || tree_int_cst_equal (TYPE_SIZE (ptype), TYPE_SIZE (arg_type)))
        {
          /* Ideally for the first conversion we'd use convert_for_assignment
             so that we get warnings for anything that doesn't match the
pointer
             type.  This isn't portable across the C and C++ front ends atm. 
*/
          val = VEC_index (tree, params, parmnum);
          val = convert (ptype, val);
          val = convert (arg_type, val);
          VEC_replace (tree, params, parmnum, val);

where we probably should convert to an integer type of the same size as
arg_type first, before converting to a pointer?

[Bug c++/52215] [4.7 Regression] __atomic_compare_exchange_n for enumeration type changes signature with -m32

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

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                     |

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-02-13 13:59:41 UTC ---
Created attachment 26650
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26650
gcc47-pr52215.patch

Untested fix.  I think it is wrong to decide this based on TYPE_SIZE,
we want to IMHO convert just the I[1248] and I16 arg_types, but not the PTR,
BOOL or INT ones (i.e. arguments in which have the overloaded types, not the
arguments that hold pointers to them or have invariant types).

[Bug c++/52215] [4.7 Regression] __atomic_compare_exchange_n for enumeration type changes signature with -m32

[amacleod at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

--- Comment #7 from Andrew Macleod <amacleod at redhat dot com> 2012-02-13 15:27:04 UTC ---
I think that is better.. its is just the I{1,2,4,8,16} we care about.

[Bug c++/52215] [4.7 Regression] __atomic_compare_exchange_n for enumeration type changes signature with -m32

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-02-13 20:13:02 UTC ---
Author: jakub
Date: Mon Feb 13 20:12:54 2012
New Revision: 184167

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=184167
Log:
    PR c++/52215
    * c-common.c (sync_resolve_params): Don't decide whether to convert
    or not based on TYPE_SIZE comparison, convert whenever arg_type
    is unsigned INTEGER_TYPE.

    * g++.dg/ext/atomic-1.C: New test.

Added:
    trunk/gcc/testsuite/g++.dg/ext/atomic-1.C
Modified:
    trunk/gcc/c-family/ChangeLog
    trunk/gcc/c-family/c-common.c
    trunk/gcc/testsuite/ChangeLog

[Bug c++/52215] [4.7 Regression] __atomic_compare_exchange_n for enumeration type changes signature with -m32

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52215

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

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

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-02-13 20:15:53 UTC ---
Fixed.