This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: Powerpc Linux build fails
- From: Alan Modra <amodra at bigpond dot net dot au>
- To: Geoff Keating <geoffk at desire dot geoffk dot org>
- Cc: ksp at securelogix dot com, dje at watson dot ibm dot com,binutils at sources dot redhat dot com, bug-binutils at gnu dot org
- Date: Wed, 19 May 2004 14:40:40 +0930
- Subject: Re: Powerpc Linux build fails
- References: <3A3FC75F7C72D711A7DC009027AC9C4B1788D9@jupiter> <20040519033008.GC12690@bubble.modra.org> <200405190427.i4J4RYiU027950@desire.geoffk.org>
On Tue, May 18, 2004 at 09:27:34PM -0700, Geoff Keating wrote:
> I am not sure about this new -mstrict. I am sure someone is expecting
> -mpower3 to really mean 'power3 only' and will get a nasty surprise
> when they use a power4 instruction by mistake and their program
> crashes.
Probably true. However, the feedback I'm getting at the moment is that
it's a nasty surprise that the Linux kernel doesn't compile.. The same
goes for current powerpc64 glibc with VMX extensions.
> It is perfectly acceptable to say "GAS version X will only work with
> GCC version > Y". People using old GCC can always use old binutils
> with it. (They can also hack their specs file to pass -many, if
> that's what they really want.)
>
> The fully-correct way to do this is to have GCC generate a directive
> after option parsing, like '.machine 7400' or so, and have GAS
> interpret it, otherwise you end up in specs hell. (Doesn't ppc gas
> already have a directive like that? I seem to remember seeing such a
> patch fly by...)
Yes, .machine was implemented 2003-11-22.
--
Alan Modra
IBM OzLabs - Linux Technology Centre