.register pseudo-op (Was: PR 10656 and PR 10657)

[Jeff Sturm]

On 7 May 2003, Anthony Green wrote:
> On Tue, 2003-05-06 at 23:10, Mark Mitchell wrote:
> > We need a fix for these problems ASAP; otherwise, I'll be forced to
> > disable Java by default on Solaris boxes in GCC 3.3, which would be a
> > step backwards.
>
> 10656 is only a regression for unpatched versions of Solaris 2.7.  The
> bug report is about using the .register pseudo-op in v8.S.  Sun has a
> patch for Solaris to add .register support to their assembler, but
> install.texi actually advises against using it (search for Sun patch
> 107058-01).

Thanks for the info.  I have no access to Solaris 7, so I couldn't confirm
the report.

But as the author of the code in question, I'm looking at it now... I'm
willing to try a fix if someone can test on 7.

> To be honest, I don't understand why .register is needed.  Is it safe to
> omit if configure tells us that the assembler doesn't support it?
> Anyone?

Some assemblers require it when %g2 is accessed as a scratch register.
Wrapping the directive with #ifdef HAVE_AS_REGISTER_PSEUDO_OP should be
sufficient.

Jeff