New C++ regression: g++.dg/template/copy1.C

[Mark Mitchell]

Andrew Pinski wrote:

>>This seems to be a new failing test; this one caused sometime in the
>>last couple of days.  We're crashing in the gimplifier.
>>
>>Does anyone know what patch caused this to fail?
>>
>See <http://gcc.gnu.org/ml/gcc-patches/2004-06/msg01864.html>.
>  
>

OK, thanks.  Hopefully RTH is fixing that.

>>I'm serious considering asking the SC for permission to into a Stage 3
>>mode for a little while here on the mainline to try to get things
>>under control...
>>    
>>
>
>If that is the case I have been waiting for a patch to reviewed 
>which fixes a regression which I caused for about a week now and
>it causes some C++ testcases to fail at -O3 and nobody has commented
>on it yet at all.
>
>Patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-06/msg01237.html>.
>  
>
Yes, that's OK -- except that I think you should pass the 
TYPE_MAIN_VARIANT to the hook; since the comparision without the hook 
ingores cv-qualifiers, so should the comparison with the hook.

RTH mentioned this issue to me briefly at the summit; apparently, we are 
doing something weird with pointer-to-member function types in the C++ 
front end.  It's not clear that it's easy to fix that without 
canonicalizing function types...

-- 
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com