c++/8534: [mainline regression] When compiling qt contradicting aliasing class created for bitfield constructor

[Jan Hubicka]

The following reply was made to PR c++/8534; it has been noted by GNATS.

From: Jan Hubicka <jh@suse.cz>
To: Wolfgang Bangerth <bangerth@ticam.utexas.edu>
Cc: Jan Hubicka <jh@suse.cz>, gcc-gnats@gcc.gnu.org
Subject: Re: c++/8534: [mainline regression] When compiling qt contradicting aliasing class created for bitfield constructor
Date: Tue, 12 Nov 2002 16:57:59 +0100

 > 
 > > > > on other systems.  Perhaps you have disabled check?
 > > > 
 > > > No, I just looked it up, but I haven't.
 > > > 
 > > > I think I have to leave this to you then. Thanks for your cooperation 
 > > > anyway!
 > > OK, I will to analyze it bit more.
 > > I am quite puzzled why it reproduces on x86_64 only as the code path is
 > > quite generic.  Perhaps the wrong alias set leaks there from some
 > > backend hook, but there are not many (only va_args fiddles aliasing as
 > > far as I can remember)
 > 
 > My machine is presently used up with different tasks, but if I find some 
 > spare cycles this afternoon, I'll retry with a clean build of gcc
 > (--enable-checking is the default on the main branch, right?).
 
 I've checked and it works for i386 compiler, so it looks like iteraction
 between the frondend and backend to cause the problem ;(
 
 Honza
 > 
 > Cheers
 >   Wolfgang
 > 
 > -------------------------------------------------------------------------
 > Wolfgang Bangerth              email:           bangerth@ticam.utexas.edu
 >                                www: http://www.ticam.utexas.edu/~bangerth
 >