Re: [PATCH] Fix miscompilation in output_constant_def

[Jeff Law <law at redhat dot com>]

Maxim Kuvyrkov wrote:
> Maxim Kuvyrkov wrote:
> > <Taking this to gcc-patches@ from gcc@; Was: Invalid code generated 
> > for Coldfire target>
>
> ...
>
> > > --- varasm.c.orig       2008-11-14 18:04:27.693643900 +0100
> > > +++ varasm.c    2008-11-14 17:58:06.522748300 +0100
> > > @@ -3245,7 +3245,7 @@
> > >      }
> > >
> > >    maybe_output_constant_def_contents (desc, defer);
> > > -  return desc->rtl;
> > > +  return copy_rtx (desc->rtl);
> > >  }
> >
> > This patch seems correct to me.  It makes output_constant_def() 
> > always return a copy of the value rather than the value.
> >
> > I will regtest this patch on ColdFire [where the miscompilation 
> > occurs], wound anyone volunteer to test it on x86[_64]?
>
> Tested on coldfire, x86 and x86_64 without regressions (gcc, g++ and 
> libstdc++ testsuites).
Yea, I would expect it to work fine.  What needs to be investigated is 
what are the sharing rules for the constant pool as that will determine 
if the pool should have a copy of the rtx in question or if the rest of 
the compiler has to copy the rtx when making modifications to constant 
pool entries.


Jeff