Please Follow Procedure, Use CVS comments

[Michael Schwendt]

On Sat, 12 Feb 2005 17:57:40 +0100, Thorsten Leemhuis wrote:

> Am Samstag, den 12.02.2005, 17:14 +0100 schrieb Michael Schwendt:
> > On Fri, 11 Feb 2005 17:29:02 -1000, Warren Togami wrote:
> > 
> > > gemi and Everyone,
> > > 
> > > Why was unison and yap requested for build?  There has been no request 
> > > on fedora-extras-list for package review, and no APPROVED message on 
> > > fedora-extras-commits showing that somebody reviewed it.
> > > 
> > > (I'm currently using prolog in one of my classes, so I'm interested in 
> > > your pl package.  I am reviewing that one now.)
> > > 
> > > https://www.redhat.com/archives/fedora-extras-list/2005-February/msg00208.html
> > > Please follow the Interim New Package Process.
> > > 
> > > This is also a reminder to everyone, please use CVS commit comments when 
> > > you commit any change.
> > 
> > I'd also like to point out that a quick succession of "commit, request
> > build in wiki, built and published" makes extras-commits list useless.
> > Once built and published, any comments would be too late. Before
> > requesting a build, better give your package changes another look and
> > test-build them in a clean environment, too. It's not like version
> > upgrades and major package changes require urgent builds. Waiting a day
> > and giving list observers the chance to comment on changes would be better
> > in the beginning.
> 
> Okay, shame on me, I did this also. 

I see a big difference between build fixes/security fixes and version
upgrades, which sometimes apply big package changes (lots of activity in
the spec and build requirements). We're kind of running CVS in
rawhide-style, so I don't care whether a build fix release is followed by
a version upgrade the other day.

I'm mildly annoyed when fedora.us QA process was criticised, but a jump to
a QA-less process seems to result in chaos. Without QA, and without a
release manager's added pair of eyes, the packagers must do a lot more QA
themselves. Also taking care that individual package upgrades don't break
repository integrity (gemi's ocaml related packages, for example, depend
on eachother and should only be published if all of them build and work
together). I would also avoid using Seth Vidal as a trial-and-error build
slave for i386. x86_64 is an exception, but you and Aurelien have offered
to help with test-builds there, too.

> But to make this whole situation a bit easier: Is there any reason why
> the Interim New Package Process is not yet in the wiki (or did I
> overlook it?)? If no one has objections I'm going to import it in the
> next days.

It's not as if there are any contributors with CVS access left who haven't
seen Warren's message and the example approvals of new packages. FWIW, I
consider this list the preferred channel for such announcements and
related discussion of it. Modifications applied in the Wiki are overlooked
easily and cause more confusion. We need a primary channel on which to
address contributors. When something's carved in stone, it makes sense
to publish it in some prominent place.

Who knows? Maybe FUDCon will change this interim process already? The
"testing" and "development" repositories should not be forgotten either.