Re: Bug#176627: a fallacy
On Sat, Jan 18, 2003 at 04:08:46PM -0800, Ron wrote:
> But even the OP agreed that not every piece of software is necessarily
> portable in which case I also agree it's up to someone who wants it on
> the port to do the porting -- the Debian maintainer is not obliged to
> port i386 assembly to some other platform just because 'duty calls for
> equality' any more than upstream is, but I don't think anyone disagreed
> with that. Just that we shouldn't get in the way of that being possible
> if someone has an interest in making it so.
Yeah, absolutely. I made a blanket assertion that "Architecture: i386"
was Evil, Bad, and Wrong back near the start of this thread, but
i386-specific assembly is a good reason for an exception there.
I agree we have a duty as developers not to get in the way of porting
efforts. To me, that means integrating reasonable patches that are
supplied, and supplying review comments on patches that are unreasonable
for whatever reason. It doesn't necessarily mean shouldering the
responsibility of producing architecture-specific assembly for every
architecture we support. I would see no reason not to consider patches
of that nature if they were submitted, but due to the extra obscurity
and inherent difficulty of such code I would very likely consult
upstream about them before unilaterally applying them.
As ever, corner cases end up being judgement calls ...
With regard to upstream support, it's worth noting that upstream may not
even support their package on Linux at all, but we frequently port
things anyway. The Debian maintainer then takes on support
responsibility for the package.
> I would presume wrt to my misunderstanding about testing, that its also
> possible if a port becomes 'permanently' broken for some package, with
> no porter prepared to help fix it, to remove that port from the
> requirement for subsequent versions to enter testing on the remaining
> ports. In that event should I make a request to ftp-admin, remove it
> from the arch list, or something else?
As far as I can remember, the right thing to do here is to ask ftpmaster
to remove the binary package(s) on the offending architecture. I've had
reasonable requests like that honoured in the past. To my knowledge, the
architecture list actually makes no difference to the testing scripts
here.
Cheers,
--
Colin Watson [cjwatson@flatline.org.uk]
Reply to: