Re: [PATCH][C++] Fix PR29433, make C++ use a lot less time/memory

[Richard Guenther <rguenther at suse dot de>]

On Fri, 23 Feb 2007, Richard Guenther wrote:

> On Thu, 22 Feb 2007, Jan Hubicka wrote:
> 
> > > Hi,
> > > 
> > > On Wed, 21 Feb 2007, Mark Mitchell wrote:
> > > 
> > > > Please spell that as TFF_UNQUALIFIED_NAME, and document as "do not print 
> > > > the qualifying scope of the top-level entity".
> > > 
> > > Done.
> > > 
> > > > Since Daniel has signed off on the GDB impact, this patch is OK if it
> > > > passes the testsuites.
> > > 
> > > It does, hence committed.
> > 
> > Hi,
> > is this yours?
> > It indeed saves a lot of memory but it seems to be due to:
> > /sw/gcc-current/include/c++/3.4/bits/stl_tree.h:592: error: declaration
> > of 'typedef struct std::_Rb_tree_node<_Val> std::_Rb_tree<_Key, _Val,
> > _KeyOfValue, _Compare, _Alloc>::_Rb_tree_node'
> > /sw/gcc-current/include/c++/3.4/bits/stl_tree.h:121: error: changes
> > meaning of '_Rb_tree_node' from 'struct std::_Rb_tree_node<_Val>'
> > 
> > These errors didn't appeared previously on the preprocessed file.
> > Perhaps I just need to regenerate, if making compiler more restrictive
> > was intended.
> 
> Yes, that's an error in libstdc++.  I have the following pending patch
> which allows -fpermissive to make this a warning.

In fact it looks like this change was already pre-approved by Mark.
I'll commit it after a bootstrap&regtest.

Richard.

> 
> Richard.
> 
> 2007-02-20  Richard Guenther  <rguenther@suse.de>
> 
> 	* class.c (note_name_declared_in_class): Make declaration
> 	changes meaning a pedwarn.
> 
> 	* g++.dg/warn/changes-meaning.C: New testcase.