Re: PATCH - Fix an ICE compiling for ppc64

[Fariborz Jahanian <fjahanian at apple dot com>]

On Jun 3, 2005, at 1:57 PM, Eric Christopher wrote:

> On Fri, 2005-06-03 at 13:53 -0700, Fariborz Jahanian wrote:
> > On Jun 3, 2005, at 1:41 PM, David Edelsohn wrote:
> >
> > > 	Where is a testcase containing bad.c.  It is difficult to know if
> > > the patch is the correct fix without being about to analyse the
> > > failure.
> >
> > Test is in the patch.  But here it is as a separate file.
>
> +/* { dg-do compile { target *-*-darwin* } } */
> +/* { dg-options "-m64 -O1 -static" } */
>
> So, why is this only on darwin and only with those compile options?
> Theoretically you should be testing with various different options via
> dejagnu (-O1, -O2, etc being part of torture.exp anyhow) and -m64 can 
> be
> multilibbed. And since this is only a compile testcase why -static?

Pattern which causes the ICE only happens with the above options. Any 
other option will not expose this bug. Somehow only with -static reload 
decides to generate move to memory and thus the ICE. Without -static, 
it doesn't go to memory and thus no ICE.

- fariborz

> -eric