Re: Debian ports [was: Re: Weirdness in "apt-get upgrade"]

[Colin Watson <cjwatson@debian.org>]

On Fri, May 31, 2002 at 04:14:30PM -0700, Brian Nelson wrote:
> Colin Watson <cjwatson@debian.org> writes:
> > It's also worth pointing out that the effort that some people see as
> > being wasted on other ports actually benefits the distribution as a
> > whole in the long run. For example, somebody complained a while back
> > about the number of bugs filed because packages didn't build on hppa;
> > what I suspect he didn't realize is that a large chunk of those bugs
> > were about gcc 3 support in C++ packages, which will be i386's default
> > compiler soon! It starts seeming a lot more worthwhile even to i386-only
> > people at that point.
> 
> Hmm, I've been thinking about this point, and it seems almost like an
> anti-argument to me.  I can't think of a more inefficient way to port
> software to gcc 3.x than to do it first on a platform that very few
> people have access to.  Presumably, woody's release has been delayed for
> months due to problems with hppa while devs tried to find access to an
> hppa machine for testing.  However, if all arches had moved to gcc 3.x
> synchronously, compilation problems across all packages probably would
> have been fixed within a month since everyone has access to i386
> machines for testing.

Moving the whole of Debian to gcc 3 is going to be quite difficult. It
involves a C++ ABI switch: that is, all C++ libraries and programs need
to be recompiled because of the new name mangling scheme. Unless
somebody decides that we don't care about compatibility, this will also
involve changing the SONAME fields of all C++ libraries, and possibly
even synchronizing this between distributions (not sure if this will be
possible).

Thus, the transition is going to be quite complicated as it is, and it's
invaluable that the hppa folks have gone through it before the rest of
us and got the majority of relatively trivial but time-consuming
problems fixed when it isn't so urgent for the distribution as a whole.
I understand several HP employees have been paid to work on it for a
substantial portion of their time, so they've been pretty efficient. We
won't now switch to gcc 3 on i386 and suddenly find that large chunks of
the distribution need quite a bit of hacking in the source code before
they even build again.

-- 
Colin Watson                                  [cjwatson@flatline.org.uk]


-- 
To UNSUBSCRIBE, email to debian-user-request@lists.debian.org 
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org