Re: [PATCH][PPC32] Makefile cleanups and gcc-3.4+binutils-2.14 c

[Tom Rini <trini@kernel.crashing.org>]

On Thu, Jul 29, 2004 at 10:05:49AM +0200, Giuliano Pochini wrote:

> On 28-Jul-2004 Tom Rini wrote:
> 
> > I've taken the binutils-2.14+gcc-3.4 bit out (and none of the other
> > cleanups) as it seems like we get 1-2 reports a week from this bad tools
> > combination:
> 
> I had no time to do a lot of testing, but it seems that binutils 2.15 +
> gcc 3.3.3 is a bad one too. I didn't try to compile the kernel (which
> may also break), but at least I couldn't compile gcc 3.4.1 with the
> above combination. It seems that as doesn't get the -mxxx parameter
> required to compile altivec stuff. Hacking the Makefile to make it
> pass -Wa,-m7455 helped a little, but it eventually failed in another
> weird way. I hadn't time to investigate further, sorry.

Stock gcc-3.3.3 or from the hammer branch?  There is, I think, a second
problem that was left out.  The problem with gcc-3.4 + binutils-2.14 is
that -many gets passed, which zeros out previous flags.  -many is fine
in binutils-2.15 (and 2.13 and 2.12 and 2.12.1 it seems), but 2.15 does
require -maltivec to be passed in order to handle altivec instructions.
Getting this right was part of the cleanup that conflicted with the
mpc52xx changes (Andrew: trying to take care of getting this into Linus'
tree now).

-- 
Tom Rini
http://gate.crashing.org/~trini/