> Chris Noe wrote:
> > Well today I went digging around, picked up the latest binutils from
> > cygnus CVS and discovered that it has fixed every bug uncovered by my
> > patch. The 16-bit i386 back-end from 2.9.1 was notoriously bad and from
> > the (quick) looks of it 2.9.5 is a hell of a lot more sane.
>
> IMHO gas 16-bit use will be inconvenient for a while... my RH 6.0 box
> uses gas 2.9.1, so older systems will be even worse.
Then again, what better way to get it in use than to break it loudly? We
can't sit here forever and just hope to accomodate those who want to sit
on their old binutils. We should be able to say 'the versions prior to
x.x.x.x are buggy, don't use them; if you do, then it's your problem'
because as I remember it, even the semi-old binutils circa 2.9.1.0.17 were
fine with the patch. And that's pretty close to the 2.9.1.0.15 recommended
in the Documentation/Changes for 2.2.
The latest and greatest 2.9.5 would be a pleasure to have as the minimum
requirement, simply for the abundance of fixed 16-bit stuff, but I realize
that that is not the best way to do it. When the time comes, and everyone
is using 2.9.5, I can remove the hardcoded instruction and be at peace.
But that isn't really a big deal. It's just something ugly that I want to
remove someday.
I guess the thing just needs a bit of testing to *see* if it will work
with the older binutils. That's something I'd like to get a feel for: how
many people are going to be bit by a bug and need to upgrade. If that
number is too high, I'd be happy to drop this idea and start on a patch to
change it all to NASM. But if that number isn't so bad, then we can apply
the patch, remove the dependency on as86/ld86, and be happy.
http://www.northlink.com/~stiker/newboot-2.3.12-2.diff.gz, for those who
feel so moved :)
Chris Noe
(stiker@northlink.com)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/