Re: debian ppc64
On Fri, Dec 03, 2004 at 08:17:10PM -0200, Rafael Espíndola wrote:
> On Fri, 3 Dec 2004 21:02:47 +0100, Sven Luther <email@example.com> wrote:
> > On Thu, Dec 02, 2004 at 04:46:59PM -0200, Rafael Ávila de Espíndola wrote:
> > > Hi, I have done some work on porting debian to the powerpc64. The original
> > > post:
> > > http://lists.debian.org/debian-powerpc/2004/10/msg00193.html
> > We have also been working silently on it ...
> > > I also made some bug reports to add support for the powerpc architecture. One
> > > missing is the one against dpkg
> > > (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=263743). Who is the
> > > "powerpc64 port team"? What is the correct name of the architecture?
> > IT is not, we need to thinkg about this, and it will depend on wheter we go
> > the pure64 way, or the biarch way. In any case i would like to have a working
> > biarch compiler and stuff before i build kernels. Well, i already have a
> > biarch compiler, but it is a hacked one, and not fit to be put into debian.
> Correct me if I am wrong but I think the amd64 started as a pure64 and
> is now adding multiarch support. A pure ppc64 may be a god starting
That is hardly the point, we need a biarch compiler, so we can produce 64bit
code on 32bit machines and vice-versa. This is independent of what will
finally be chosen for the userland. The amd64 debian situation is so muddied
in ugly politics that i would best not use it as reference for this.
> > > I see that you need a kernel. Maybe this helps:
> > > http://lists.debian.org/debian-powerpc/2004/07/msg00449.html
> > No, i don't think so, a ppc64 kernel which cannot be built inside debian is
> > not really all that usefull, and there is still the question of the glibc.
> > Work is underway on this, but there is still a way to go.
> Would it be worth if it could build in a pure64? If not, what do you
> think is a task worth doing?
well, we already have a plan to hack together a mini-glibc package, which
would provide only the ppc64 glibc headers, which i hope are enough to build
the biarch compiler. Once that is done, we can build ppc64 kernels from it,
and also start working on the userland, but a new glibc can't and won't enter
sarge, while there may still be some (minimal) hope for the biarch compiler.
Stay tuned for more on this.