I see that mjg has already answered this to a fair degree; therefore, I'll
mostly only be commenting on things he didn't cover in detail, that I
happen to know.
On Mon, Jan 19, 2004 at 09:00:05AM -0600, John Goerzen wrote:
> Hi,
>
> I am quite interested in the port to the NetBSD kernel. I've checked
> the recent list archives and the website, and couldn't find an answer to
> these questions, so please pardon me if they're FAQs. Here we go:
>
> 1. Exactly how much of the NetBSD port comes from NetBSD and how much
> comes from Debian? Is, for example, the NetBSD base system used in its
> entirety, or are only necessary things like ifconfig pulled in from
> there?
As said, there are two answers. For the 'native' port, we have the
following things either officially or almost-officially packaged:
netbsd-make (NetBSD native make/pmake/bmake variant, needed for building )
netbsd-libc (libc, libm, et al kernel-source-current (aka CVS src/sys )
The first two are known to build in a chroot environment and produce valid
and useable Debian packages. The latter is still being packaged, and may
need a kernel-package style helper package; the main holdup on this is the
issue over introducing that much source code under a 4-clause BSD license
from a source known to care about advertising clauses.
I am in the process of contacting upstream authors, on behalf of Debian and
with the permission of The NetBSD Foundation, and asking them to consider
relicensing to a 3-clause BSD license instead; so far, this effort has
enjoyed a nearly 100% success rate, but as it involves a great deal of
manual effort, both in preparing the patches and contacting the authors,
the going is somewhat slow.
The not-quite-auto-builder box for the lightbearer.com archive is currently
down, pending new (and vastly faster) hardware; the archive has been let
slip a bit, due to this, I'm afraid. The hardware even exists already,
but it currently has some business-critical stuff on it that I can't move
until I either find time to rewrite a chunk of source code, or the upstream
author gets around to completing the same.
> 2. Are people actively working on this port?
Yes. Less steadily than at times, but still actively. I'm also working on
the naming issue with TNF and other folks, which is a necessary long-term
issue to resolve.
> 3. Where could I best contribute to the port? Would an autobuilder be
> useful? (I do have experience running those; I used to run the one for
> our Alpha port.)
It would be, once we can get one even chroot-installed, prefferably
real-installed; while I had very good success at squashing bug problems on
my old box, and getting things basically workable, it ate so much time (due
to bootstrapping a port always being time-intensive) that I mostly only
ever got around to recompiling things if a newer version showed up in a
dependancy of some sort.
> 4. Is there any infrastructure on debian.org machines for this? (Areas
> on sid, etc.) or does one have to use the ucam.org repository?
We have a wishlist ftp-archive bug open to get space, but that almost
certainly won't be resolved until we can show the world (or at least
debian-devel) a compelling demonstration that we have our act together,
and can actually get, for example, an entire base system installed and
functioning, using debian-installer. Until then, we're unlikely to have
enough evidence to convince folks that it's worth taking the hit on the
various infrastructure pieces to set this up.
And, really, I can't say I disagree with them...
Never fear, however. Disk is getting cheaper, and if we run out of network
bandwidth to the current servers, I strongly suspect that we could find a
couple of folks with fast connections to host hardware for a little while.
I don't think a lack of infrastructure is going to be the main stumbling
block.
> 5. Has any thought been given to supporting the non-i386 ports of
> NetBSD?
Yes; as mjg said, there has been work on both sparc and alpha ports. I am
more than willing to hammer on the source packages I maintain, and make
sure that they can cope with this; I suspect they have a few i386-isms
right now (or, rather, they have single-arch-isms that happen to be
i386-isms due to that being the platform I'm working on), but without
someone to actively hammer on them and find what breaks, I'm disinclined
to blowing too much time on trying to fix this (which is to say, if it's
obviously non-portable and costs nothing much, and I see it, I'll fix it;
if it costs a lot, which is the current case of what I think is broken,
I'll fix it as soon as there is any benefit to paying the cost :)
I believe I still have the full archive from debian-bsd.lightbearer.com,
including the chroot tarball, floating around. However, since we ran
into major problems with threading under pth breaking, and -current (aka
pre-2.0) has native threading, I decided that moving to that was the next
requirement, and that's when most of what I was doing stalled...
FWIW, NetBSD -current has a couple of significant patches we submitted, and
I have CVS (or svn) commit access to both the GCC and XFree86 packages,
which were the two biggest pain-in-the-butt situations; both have probably
bitrotted some, but should be easily fixed back to good working order once
we have a core system back up and running.
Anyone wanting to help Lars with the wishlist bugs on Enemies of Carlotta
would probably speed up getting the lightbearer.com archive back, so if
you happen to like python, that might, oddly enough, be very useful. :)
--
Joel Baker <fenton@debian.org> ,''`.
Debian GNU/NetBSD(i386) porter : :' :
`. `'
`-
Attachment:
pgp6uSBmmI_wQ.pgp
Description: PGP signature