Re: Feaping Creature-ism in core Debian Packages
On Thu, Sep 02, 1999 at 01:25:48AM -0700, Joey Hess wrote:
> Dale has merely found out that the first steps of porting debian to a new
> architecture is damned hard. There are perhaps 16 people in the world that
> really know just how hard: about 2 per port. It's something you only do
> once. While it'd be nice if it were easier, I just don't see why it should
> be a priority. Even if we did everything Dale wants, porting debian to a new
> architecture would still be damned hard.
> (I'd really like to hear some of those 20-odd people speak up, BTW.)
Ok, well, as one of the people who helped on the early stages of the
arm port: Joey is right.
The arm port was almost trivial in some ways -- the corel folks had done a
lot of work already. However, I wound up having to deal with documented
functionality in libc not working (so, for example, ssh wouldn't build),
and no debugging tools to speak of (gdb wouldn't build, and strace was
in even worse shape).
In my opinion: the early stages of porting involve porting gcc, and libc,
and the kernel. I've not worked on this.
The next stage of porting is where you build yourself enough of an
environment that you can work natively. This is *throwaway* work.
Think of this as building something analogous to what the debian install
disks would create (except you're building an environment to build and
test packages as well as one in which to install packages) -- you're
not ready to install packages yet, and if you don't have to rebuild
everything several times before you're done you're lucky.
Only after you have a system to work with can you really tackle the issues
of building, testing and debugging packages.