[Bug c++/57107] New: tree check fail in unlink_stmt_vdef

[Bug c++/57107] New: tree check fail in unlink_stmt_vdef

[dcb314 at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57107

             Bug #: 57107
           Summary: tree check fail in unlink_stmt_vdef
    Classification: Unclassified
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dcb314@hotmail.com


I just tried to compile the package luminance-hdr-2.3.0-6
on gcc-4.9 trunk dated 20130428 on an AMD x86_64 box.

The compiler said

/home/dcb/rpmbuild/BUILD/luminance-hdr-2.3.0/src/Fileformat/pngwriter.cpp:205:1:
internal compiler error: tree check: expected ssa_name, have var_decl in
unlink_stmt_vdef, at tree-ssa-operands.c:1322
 }
 ^
0xc8c92a tree_check_failed(tree_node const*, char const*, int, char const*,
...)
    ../../src/trunk/gcc/tree.c:8972
0xbc0db7 tree_check
    ../../src/trunk/gcc/tree.h:3690
0xbc0db7 unlink_stmt_vdef(gimple_statement_d*)
    ../../src/trunk/gcc/tree-ssa-operands.c:1322
0xae2919 optimize_clobbers
    ../../src/trunk/gcc/tree-eh.c:3298
0xae681b cleanup_empty_eh
    ../../src/trunk/gcc/tree-eh.c:4227
0xae681b cleanup_all_empty_eh
    ../../src/trunk/gcc/tree-eh.c:4355
0xae681b execute_cleanup_eh_1
    ../../src/trunk/gcc/tree-eh.c:4384
0xae681b execute_cleanup_eh
    ../../src/trunk/gcc/tree-eh.c:4409
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.

Preprocessed source code attached. Flag -O2 required.

[Bug c++/57107] tree check fail in unlink_stmt_vdef

[dcb314 at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57107

--- Comment #1 from David Binderman <dcb314 at hotmail dot com> 2013-04-29 08:08:34 UTC ---
Created attachment 29969
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=29969
gzipped C++ source code

[Bug c++/57107] tree check fail in unlink_stmt_vdef

[markus at trippelsdorf dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57107

Markus Trippelsdorf <markus at trippelsdorf dot de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |markus at trippelsdorf dot
                   |                            |de

--- Comment #2 from Markus Trippelsdorf <markus at trippelsdorf dot de> 2013-04-29 09:00:33 UTC ---
Reduced:

class A {
public:
  template <typename _Tp1> struct rebind {
    typedef A other;
  };
  ~A() {}
};
template <typename _Alloc> struct __alloc_traits {
  template <typename _Tp> struct rebind {
    typedef typename _Alloc::template rebind<_Tp>::other other;
  };
};
template <typename _Tp, typename _Alloc> struct B {
  typedef typename __alloc_traits<_Alloc>::template rebind<_Tp>::other
  _Tp_alloc_type;
  struct C : _Tp_alloc_type {
    C(_Tp_alloc_type) : _Tp_alloc_type() {}
  };
  typedef _Alloc allocator_type;
  B(int, const allocator_type &p2) : _M_impl(p2) {}
  C _M_impl;
};
template <typename _Tp, typename _Alloc = A> class D : B<_Tp, _Alloc> {
  typedef B<_Tp, _Alloc> _Base;
public:
  typedef _Tp value_type;
  typedef _Alloc allocator_type;
  D(const value_type &p1, const allocator_type &p2 = allocator_type())
      : _Base(0, p2) {
    _M_fill_initialize(p1);
  }
  void _M_fill_initialize(const value_type &);
};
void _setjmp();
void writeQImageToPng() {
  _setjmp();
  D<unsigned*>(0);
}

[Bug c++/57107] tree check fail in unlink_stmt_vdef

[mpolacek at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57107

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-04-29
                 CC|                            |mpolacek at gcc dot
                   |                            |gnu.org, rguenth at gcc dot
                   |                            |gnu.org
     Ever Confirmed|0                           |1

--- Comment #3 from Marek Polacek <mpolacek at gcc dot gnu.org> 2013-04-29 16:31:26 UTC ---
Confirmed.  Started with http://gcc.gnu.org/r198096

[Bug middle-end/57107] [4.9 Regression] tree check fail in unlink_stmt_vdef

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57107

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
          Component|c++                         |middle-end
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
   Target Milestone|---                         |4.9.0
            Summary|tree check fail in          |[4.9 Regression] tree check
                   |unlink_stmt_vdef            |fail in unlink_stmt_vdef

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-30 08:24:44 UTC ---
I will have a look.

[Bug middle-end/57107] [4.9 Regression] tree check fail in unlink_stmt_vdef

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57107

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

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

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-30 10:53:21 UTC ---
(gdb) call debug_gimple_stmt (stmt)
# .MEM_159(ab) = VDEF <.MEM>
D.63506 ={v} {CLOBBER};

SSA form is not up-to-date.  It's a latent issue with sinking clobbers
during EH cleanup.  We have

<L6>:
  # .MEM_32(ab) = VDEF <.MEM_31(ab)>
  D.2688 ={v} {CLOBBER};
  resx 3

<L7>:
  # .MEM_61 = VDEF <.MEM_32(ab)>
  MEM[(struct new_allocator *)&D.2706] ={v} {CLOBBER};
  # .MEM_62 = VDEF <.MEM_61>
  D.2706 ={v} {CLOBBER};
  # .MEM_34 = VDEF <.MEM_62>
  D.2706 ={v} {CLOBBER};
  resx 2

<L8>:
  # .MEM_35 = VDEF <.MEM_34>
  t ={v} {CLOBBER};
  resx 1

and first sink D.2688 ={v} {CLOBBER}; into the successor (L7), dropping
virtual operands.  Then we continue sinking but call unlink_stmt_vdef
on the moved stmt which no longer has up-to-date SSA form and thus
we ICE.

As we only sink across forwarder edges we can trivially keep the original
virtual operands in place.

I have a patch.

[Bug middle-end/57107] [4.9 Regression] tree check fail in unlink_stmt_vdef

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57107

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

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

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-30 13:54:32 UTC ---

Author: rguenth
Date: Tue Apr 30 13:54:14 2013
New Revision: 198454

URL: http://gcc.gnu.org/viewcvs?rev=198454&root=gcc&view=rev
Log:
2013-04-30  Richard Biener  <rguenther@suse.de>

    PR middle-end/57107
    * tree-eh.c (sink_clobbers): Preserve virtual SSA form.

    * g++.dg/torture/pr57107.C: New testcase.

Added:
    trunk/gcc/testsuite/g++.dg/torture/pr57107.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-eh.c