Dealing with PPC in Fedora 9(+)

[David Woodhouse]

On Tue, 2007-12-04 at 22:29 -0500, Jeremy Katz wrote:
> On Tue, 2007-12-04 at 22:58 +0000, David Woodhouse wrote:
> > On Mon, 2007-11-12 at 14:58 -0500, Jesse Keating wrote:
> > > This seems a reasonable compromise all together.  I can be happy with
> > > this for Fedora 9.  Hopefully by the time 9 is let loose, we'll have
> > > had at least one other full fledged secondary arch up and running and
> > > proving that the method can work.
> > 
> > I suspect this is going to work a whole lot better if I have commit
> > access to anaconda, kudzu, rhpl, booty, etc.
> 
> I'm just going to come right out and say that if Fedora as a project
> starts dictating commit access to hosted "upstream" projects, that's a
> quick way to kill the use of Fedora for hosting upstream projects.
> Because that's not the way that commit access for projects should be
> given.  Ever.

Yes, that's very true. It would be very silly.

Josh is right, however -- I wasn't asking the Board to mandate it. I was
just continuing the thread because it started here. Sorry if I was
unclear.

> > At the moment, the round-trip time between me generating a patch to
> > something like kudzu and seeing it in a testable rawhide build is
> > somewhat suboptimal.
> 
> Getting rawhide builds at all right now is a bit of a challenge all its
> own with the openssl/openldap rev :)

This is also true.

> > I don't mean to complain -- I know people are busy and have better
> > things to do than commit my patches and kick off builds so that I can
> > get on with testing rawhide. But people are going to be busy in the
> > run-up to te releases too, when I most want my fixes to be getting into
> > builds promptly.
> 
> Actually, run up to release is easier because it's "bug-fix mode" time
> rather than "integrate huge new chunks of code" time.  Especially when
> some of the changes underway are directly related to the change that
> you're also wanting to do.

Perhaps so. Can I hold you to that? :)

> And < a day of turnaround for the most recent one really isn't bad.

That's true; thanks for that. I think it's the fastest I've ever managed
to get such a patch applied -- it makes me wonder if I should have been
submitting patches with a tempting 'FIXME' in them all along... :)

Unfortunately, it still isn't in a built anaconda package -- and even if
it was, that anaconda package would have been built against
kudzu-1.2.80, which still lacks the changes I made last week.

Again, I should point out that I don't mean to criticise. I'd just like
to find a way to improve the situation without constantly having to
badger people to commit stuff for me and build stuff for me.

> Also, there is a member of what is ostensibly the ppc team (pnasrat) who
> does have commit access to most of the above.

If Paul is happy to be bothered and will respond in good time, that
works for me. Paul, can you get a kudzu into rawhide with my latest
changes please (which I think are all now in kudzu upstream cvs; thanks
Bill)?

> > So it's probably best if I can be a little more self-sufficient in that
> > respect, by having commit access to both upstream and package
> > repositories and being able to do builds (at least for rawhide). Please.
> > 
> > Actually, we've spoken often of "arch teams" having commit access to
> > _all_ packages. Is that feasible?
> 
> There's a very large difference between "committing to packages" and
> "committing to upstream".  

Yes, there is. From my point of view, I would settle just for committing
to packages. I could add patches then -- after I submit them upstream of
course.

Since the packages I was thinking of tend to ship without _any_ patches,
because we _are_ upstream for them and we tend to just ship a new
tarball for each build, I'm not sure that's the best answer -- hence the
request for upstream access too, for those packages. 

> We don't have a great way of doing the "arch
> maintainers can commit to any package", but since we're not talking
> about huge numbers of arch teams, we could probably go with the quick
> answer (just adding people to cvsadmin)

That would be great; yes please.

(Although I don't feel wonderfully happy about accepting such rights for
myself alone when I feel strongly that _all_ Fedora contributors should
be able to commit to all packages. It's not as if we can't revert stuff,
and can't promote an attitude of violence towards people who abuse the
privilege, like we already have.)

-- 
dwmw2