On Tue, Apr 21, 2009 at 08:23:59PM +0200, Michael Schwendt wrote:
On Tue, 21 Apr 2009 08:44:49 -0500, Bruno wrote:
> On Tue, Apr 21, 2009 at 12:13:24 +0300,
> Axel Thimm wrote:
> > If I as a packager craft a package for Fedora, how can I check that
> > this package's contents aren't accidentially conflicting with another
> > package in Fedora? Is there some magic python script that would check
> > a package against a whole repo?
> There should also be checks run as a project over the whole set of
> packages. I know of at least two conflicts currently in rawhide (down
> from 3 assuming the new R packages got pushed out this morning) and
> I think there is a another one that if I find again I'll make sure is
> bugzilla'd. I have a lot of stuff installed, so I see more of these
> than most people, but I am guessing there are others that I don't see.
There are more. Search bugzilla for "conflicts". I've had the filing of bz
tickets semi-automated till I lost the work-in-progress ticket tracking
script which I had stored on a non-backuped-machine. At least the
following conflicts appear to be new compared with my last log:
R-Matrix conflicts with R-core
R-Matrix-devel conflict with R-core
astronomy-backgrounds conflicts with *-backgrounds* (in bugzilla)
globus-common-progs conflicts with grid-packaging-tools (in bugzilla)
The script I use to check a set of repositories is old and unfinished:
What is unfinished in this script? Can one use it against a repo like
Bruno writes, or against a package and a repo like I suggested?
There are a couple of useful scripts floating around, maybe they
should get into one upstream source? Is perhaps yum-utils the proper
place to submit them?
Working on such scripts and enhancing them is only fun if the filed
tickets are dealt with and are not ignored.
I think that such a script could even become part of koji/bodhi
automation and alert packagers/builders upon creating a buggy
package. Sometimes even a package can evolve into conflicting with
another if for example new exectuables are added. A koji check would
do wonders to keep this off the repos.
(Of course performance is an issue, so if the scripts takes a while it
can's be chained in, yet, but it would be a start)
Axel.Thimm at ATrpms.net