[this thread started in a discussion about rackmount A1200s]
Aha, I have a couple of colleagues who are involved with NetBSD, and they sometimes try to lure me over to their side.
NetBSD is pretty stable and well supported. I can honestly say that it's gotten
faster on the same hardware over time.
My gentoo/m68k effort is quite a personal one. I needed a system that could help me build the software I wanted with the features I wanted. Binary distributions simply dont give me that flexibility, Gentoo with its USE-flags and package masking does - I have to build everything myself, but that's ok, Aranym is of great help. I also find the Gentoo developer team quite a helpfull bunch, many of them are also embedded developers and adding m68k as a platform in portage happened after just some chatting over IRC.
I build everything from source, too, on NetBSD. My Amiga 4000 is currently doing a bulk package build of pkgsrc for m68k. I don't use those binaries myself even though I make them, but they're made available for other NetBSD users.
I'd love to hear more about how you've gotten Aranym working with proper MMU support. I've tried a few times, and maybe it's time to try again.
I do follow Debian/m68k too, both mailing list and IRC, so that I have an idea on what's going on. I think the biggest problem with Debian and m68k is that Debian is too big and m68k too small, both as a community and in sheer compilation power. I think it would be better to treat m68k as an embedded platform and move over to Emdebian instead of going for a full fledged distro as squeeze. In my view, m68k systems have much more in common with the embedded platforms than they do with todays PCs for example.
I still have to wrap my head around the whole "archive" thing. In NetBSD, the whole OS exists in one source tree (well, X is another, but they go together easily) and pkgsrc is a separate tree with ten thousand or so software packages. Building the entire OS (which works on many POSIX OSes and regardless of architecture) is as simple as CVS'ing the NetBSD source tree, then running these in the source tree:
./build.sh -m amiga tools
./build.sh -m amiga distribution
./build.sh -m amiga kernel=GENERIC
Building the cross-compiling toolchain on a $200 USD quad core 2.3 GHz AMD system takes about six minutes (I used "-j 4", of course) and building the main distribution takes about an hour. You can also use the target "release" to build all of the tools (AmigaDOS tools to load the NetBSD install disk image to a swap partition and to boot NetBSD from AmigaDOS, for instance) and everything else, including the installation and generic kernels.
On a real Amiga (m68060), it takes about two to three days to do an entire distribution.
It's easy to understand, but I'm not sure if there's a real distinction between what constitutes the base OS and what makes up the rest of the Debian package system. If there were an easy distinction, I would think that once the toolchain bootstrapping issues were handled, it'd just be a matter of at most a week to compile the base OS so that other people could be on the same page. This is what I'm trying to understand now.
My big hope for 2010 is that we will finally get updated glibc (or eglibc) with NPTL. I'm still on old glibc-2.3.6, I know Debian somehow managed to get as far as 2.5, but I never got that compiled here.
The last exchange on
debian-68k@lists.debian.org indicated that the newest glibc is already available and that TLS/NPTL are already available upstream.
Next I have to understand what comes from upstream directly from GNU and what comes from GNU but is patched as part of the Debian sources. I still don't understand the source hierarchy. I'm obviously quite the beginner when it comes to m68k GNU/Linux.
