New tracker bugs for the use of ExcludeArchs in packages

[Ralf Corsepius]

On Mon, 2006-01-30 at 11:46 +0100, Thorsten Leemhuis wrote:
> Am Montag, den 30.01.2006, 05:49 +0100 schrieb Ralf Corsepius: 
> > On Sun, 2006-01-29 at 20:03 +0100, Thorsten Leemhuis wrote:
> > > Am Sonntag, den 29.01.2006, 12:31 +0100 schrieb Thorsten Leemhuis:
> > > 
> > > > Just FYI, I created several new tracker bugs:
> > > > 
> > > > 179258 - FE-ExcludeArch-x86
> > > > 179259 - FE-ExcludeArch-x64
> > > > 179260 - FE-ExcludeArch-ppc
> > > 
> > > Okay guys, could someone post a proposal how to handle the whole
> > > ExcludeArch/ExclusiveArch tracking in the future so FESCo can look at it
> > > and change the Package Review Guidelines accordingly? I really would
> > > prefer defined rules that are used in practice over civil
> > > disobedience ;-)
> > Packages that don't build for certain archs due to build problems simply
> > are bugged.
> > 
> > IMO, the appropriate means to handle such cases would be filing
> > individual PRs. I.e. filing them under "package:xyz  arch:foo" should be
> > sufficient.
> > 
> > If you really want something centralized, add a Bugzilla keywords,
> > bugzilla queries could use, but am having doubts on if this would be
> > useful at all.
> 
> Why Bugzilla keywords?
It's an alternative, easier to use if chasing specific issues.

>  We use the tracker bugs for many other things
> already and people in Core and Extras probably are used to it.
IMO, "tracker bugs" are the means of choice for dependencies, e.g. a bug
in one package blocking others from upgrading.

>  And where
> is the difference between adding "FE-ExcludeArch-ppc" in the "Blocks
> Bug:" field to adding it after "Keywords:"? I can't see any notable.
ExcludeA* are a bug's feature/attribute (Actually a band-aid to work
around a deficit inside of a package), not a dependency.

An arch-specific bug in GCC, preventing a package from being upgraded
would be a dependency. 

> > Packages for which "Exclusive/ExcludeArch" is a feature, aren't bugged,
> > therefore I don't see any need to file a PR on them at all.
> > 
> > If you want a list/table of "non-general packages",
> 
> Yes, I think we should have one somewhere. We had one in the wiki in the
> past but it it was dropped because people preferred to have it in
> bugzilla. Now other people don't want it in bugzilla :-|
Well, could have been me, who said that. IMO, bugzilla is a bug tracking
system, and should be applied as such. What Fedora actually needs is a
"package status/tracking system", presenting a summary and status in a
nicely written, simple web package (Something similar to the Debian
devel packages).

> >  a script could
> > extract this info from *.src.rpms (E.g. the buildsys could do this, when
> > shifting a package from "needssign" to "release"). Whether to feed
> > bugzilla with this info is arguable, automatically feeding a Wiki might
> > be more appropriate.
> 
> Well, a quick grep trough the devel checkout showed 46 packages that
> currently have ExcludeArch or ExclusiveArch. For some of them the bugs
> are already filed. Some of those are probably in the category
> "ExcludeArch because the packager was not able to fix it" and have no
> bugs yet. 


> I think round about 20 of these 46 are "ExcludeArch because a package is
> for certain archs only" and have no bug yet -- reporting bugs for them
> is a job that can be done in round about 30-60 minutes (heck, this whole
> discussion took longer already). Writing the script takes longer afaics.
> Both need a bit care later. 
That's what I thought. Somebody, who has a fully checked out CVS or (IMO
preferable) a complete local copy of the FE SRPMS, could easily write a
script and feed a wiki on a, say weekly or daily, basis (No I can't do
this, I don't have the bandwidth required for keeping my SRPMS current).

Ralf