Re: Suggestions for how ARM Ltd could help Debian?
on Thu, Nov 27, 2003 at 06:07:43PM +0000, Wookey wrote:
> Do we have people who could free up more time for this stuff is someone paid
> them to do it?
In principle, yes, I'm sure there are numerous people who would
happily work on Debian/ARM if they were paid to do so. I don't
think this would be unduly divisive: there have been plenty of
instances in the past of people being paid to develop things either
for Linux on ARM or for Debian.
> Faster build/test hardware (e.g. a couple of Iyonix boxes - 600Mhz Xscale
> with Hard drives)? Would that help?
Probably not quite as much as you might think. We don't really
have any shortage of autobuilder CPU time; the four existing
machines are quite well able to keep up under normal circumstances.
Debussy does sometimes get overloaded, but I don't think this is
seriously impeding work on the port.
> Maybe paying for some time spent on getting DI or embedded debian into
> better shape. That would almost certainly speed things along a little.
Right now I think d-i is the biggest problem we have. Not being
able to install sarge would be a blow. So anything that helps with
that has got to be good news.
> ARM have apparently been doing good things with gcc too. Is the ARM gcc
> corner holding it's own? Phil? What would help you most? Things seem to be
> going quite well on that front to me with gcc 3.3 working pretty well these
Yup. I don't think we have any serious problems here at the
moment. ARM are already providing a fair amount of support for
> Another thing I was wondering about - which isn't directly related - is: do
> we need a big-endian version? Some newer ARM systems are defaulting to
> big-endian use, although most of them can run little-endian too. Is it worth
> doing a big-endian arch to support these directly or not?
I haven't seen a great deal of demand for this. Although
big-endian ARM systems certainly do exist, they tend to be aimed
more at deeply embedded networking or control applications where
Debian is unlikely to be the OS of choice anyway. I don't think a
big-endian port would have enough users to justify the disk space
in the archive or the manpower to create it.