rs6000 patch fix enable-checking failure.

[Geoff Keating]

> Date: Tue, 30 Jul 2002 20:50:07 +0100
> From: Graham Stott <graham.stott@btinternet.com>
> User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.9) Gecko/20020513
> X-Accept-Language: en-us, en
> CC: gcc-patches@gcc.gnu.org
> X-OriginalArrivalTime: 30 Jul 2002 19:50:09.0444 (UTC) FILETIME=[4F7C6E40:01C23802]
> 
> Geoff Keating wrote:
> > Graham Stott <graham.stott@btinternet.com> writes:
> > 
> > 
> >>All,
> >>
> >>Operand 3 of a CODE_LABEL is a BB (basic block) format code 'B' which is not a valid
> >>argument to the X0INT macro which is expecting format code '0'.
> >>
> >>The patch replaces X0INT (XEXP (k, 0), 3) with (unsigned ) BLOCK_FOR_INSN (XEXP (k, 0))
> >>
> >>Built powerpc-eabi --enable-checking enabled all languages no regressions.
> > 
> > 
> > It's supposed to be the number of the insn the LABEL_REF refers to.  I
> > think there were some changes in this area since it was written...
> Ok thanks for that pointer I'll rework my patch,
> 
> > 
> > On the PCH branch, I think I finally gave up and no longer allow
> > LABEL_REFs to get into the hash table.
> I took a quick look and it's the same on pch branch, unless you've undated
> the branch in the last couple days.

The code that inserts into the hash table blocks LABEL_REFs, though.

-- 
- Geoffrey Keating <geoffk@geoffk.org> <geoffk@redhat.com>