[Bug debug/11325] New: [3.4 regression] nested class debug info incorrect

["bkoz at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11325

           Summary: [3.4 regression]  nested class debug info incorrect
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P1
         Component: debug
        AssignedTo: mark at codesourcery dot com
        ReportedBy: bkoz at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org,nathan at codesourcery
                    dot com
 GCC build triplet: x86-unknown-linux
  GCC host triplet: x86-unknown-linux
GCC target triplet: x86-unknown-linux

Sometime between 2003-02-16 and 2003-03-19, the following small code snipped
started to be un-debuggable. This is a nice small snippet abstracted from v3's
locale code.


namespace bug
{
  class outer
  {
  public:
    class inner;

    static inner*  _S_;
    inner*	   _M_;

    outer();

    ~outer();
  };

  class outer::inner
  {
    typedef unsigned int size_t;

  private:
    size_t				_M_references;
    const void**			_M_facets;
    size_t 				_M_facets_size;
    char** 				_M_names;
  };

  outer::outer()
  { _M_ = new outer::inner; }

  outer::~outer()
  { delete _M_; }
}

int main()
{
  bug::outer loc;

  return 0; // (gdb) p *loc._M_
}


What happens in gcc-3.2.x, gcc-3.3.x:

(gdb) p *loc._M_
$2 = {
  _M_references = 0, 
  _M_facets = 0x0, 
  _M_facets_size = 0, 
  _M_names = 0x0
}

What happens in gcc-head and gcc > 20030319, is:
(gdb) p *loc._M_
$3 = <incomplete type>

Why you're getting cc'd: 

Perhaps you know what went wrong. Perhaps not: if so, then please re-assign
this. Nathan in particular seems to have some heavy lifting work around
2003-02-20. If you've got pre-built cc1pluses, it should be easy enough to
narrow down the date to a smaller range than what I've been able to do.

This is kind of a blocker for v3, but more of a sad bummer on an otherwise
depressing day. I can regress to a 2003-02-16 toolchain and work with that until
this is fixed without major problems. (In any case, I spent much of the day
doing this.) Still, it should probably get fixed.

best,
benjamin