http://fedoraproject.org/wiki/TomCallaway/SecondaryArchitectures

[David Woodhouse]

On Wed, 2007-07-11 at 20:24 -0400, Christopher Aillon wrote:
> So why make them do anything at all?  Let's just auto file the bug.
> Let the ones who want to care shine out.

Because I was only talking about the lowest common denominator; the
people who are either unwilling or unable to maintain their package
properly. In the cases where the packager _is_ actually doing their job
properly, they _will_ be looking at the build failures; at least for
long enough to decide whether it's an arch-specific issue (which mostly
it won't be, in my experience).

It isn't _really_ expected that anyone would ever ship a failed build
without even looking at the failure; I was just pointing out that they
_could_. If we really want that kind of person responsible for packages
carrying the 'Fedora' brand, nothing in what I've suggested would
actually _prevent_ them from doing just that.

Their packages are probably a lost cause if there are ever any problems
(of any kind), both in primary architectures and in secondary. But for
packages which _are_ actually maintained properly¹ we don't want
partially-failed buggy builds landing in the repository automatically,
and making the secondary trees gratuitously inconsistent; we want that
to happen only after a _conscious_ decision by the maintainer that
that's what should happen. The idea is that it should be a relatively
rare occurrence. If it's frequent, we might as well ditch the whole
secondary architecture thing altogether; they're no better off than they
were when they were entirely separate.

We also want to encourage coherent bugzilla reports, by giving them a
template to fill in instead of just automatically filing it. Again,
those who really can't be bothered wouldn't _have_ to fill it in
properly; they could just submit the template unaltered. But again, we
wouldn't expect that to happen in the common case.

-- 
dwmw2

¹ which is most, if not all of them. While people have got all pointlessly
  Californian about the terminology used to describe these 'lowest common
  denominator' packagers, I don't think anyone _specific_ has ever been
  mentioned, apart from myself (mostly in the context that I really 
  shouldn't be responsible for anything even remotely involved with dbus).

  But you seem worried about making life harder for these people if they
  exist, so I'm pointing out that it _wouldn't_ be harder for them. As
  long as they can refrain from breaking their mouse by drooling on it,
  they don't even need to _look_ at build failures before they ship 
  their buggy package. For better or worse.