Re: 21 GCC HEAD regressions, 2 new, with your patch on 2004-07-07T22:45:18Z.

["H. J. Lu" <hjl at lucon dot org>]

It looks like we may need a undo_make_decl_one_only if we have to
change our mind like that.


H.J.
On Thu, Jul 08, 2004 at 10:40:13PM -0700, Lu, Hongjiu wrote:
> The problem is we changed our mind later after making it one only. I
> don't know if we can turn a global symbol to local early on. We need to
> undo what we did before. However, there is no way to do it properly. It
> looks like every defined MAKE_DECL_ONE_ONLY is
> 
> DECL_WEAK (DECL) = 1
> 
> If we rename MAKE_DECL_ONE_ONLY to something like
> MAKE_DECL_ONE_ONLY_WEAK, we can do
> 
> #ifdef MAKE_DECL_ONE_ONLY_WEAK
> 	DECL_WEAK (DECL) = 1;
> #endif
> 
> instead of
> 
> #ifdef MAKE_DECL_ONE_ONLY
> 	MAKE_DECL_ONE_ONLY (DECL)
> #endif
> 
> Then we can undo the demage. If we need a different MAKE_DECL_ONE than
> DECL_WEAK, we can have MAKE_DECL_ONE_FOORBAR.
> 
> 
> H.J. Lu
> Intel Corporation
> 
> 
> >-----Original Message-----
> >From: Andrew Pinski [mailto:pinskia@physics.uc.edu]
> >Sent: Thursday, July 08, 2004 9:38 PM
> >To: Lu, Hongjiu
> >Cc: pinskia@physics.uc.edu; gcc@gcc.gnu.org; apinski@apple.com
> >Subject: Re: 21 GCC HEAD regressions, 2 new, with your patch on 2004-
> >07-07T22:45:18Z.
> >
> >> How do I reproduce it? It works for me on Linux/x86.
> >
> >By a cross compiler to powerpc-apple-darwin.
> >
> >Here is the anyalsis of the problem, we already had called
> >MAKE_DECL_ONE_ONLY
> >by the time we get to emit_tinfo_decl so your patch is wrong.
> >
> >Figure out where we set DECL_ONE_ONLY instead of setting it late in the
> >game.
> >
> >Thanks,
> >Andrew Pinski