[Rpm-metadata] CVS access problem

Paul Nasrat pnasrat at redhat.com
Sun Jul 23 14:41:05 UTC 2006


On Sat, 2006-07-22 at 15:32 +0200, Hans-Peter Jansen wrote:
> Am Freitag, 21. Juli 2006 22:01 schrieb Hans-Peter Jansen:
> > Am Freitag, 21. Juli 2006 19:37 schrieb Paul Nasrat:
> > >
> > > Yeah, I've made a whole bunch of changes before I got to it - can
> > > you possibly rediff against HEAD please.
> >
> > Done, attached. Took a bit longer, since one of your changes made
> > createrepo fail to locate any files in the normal case (without
> > --split option at least), because you reference the wrong var
> > [directory instead of dirname] in extension_visitor(). While at it, I
> > made the extension test a bit more eye friendly (IMHO). You may want
> > to apply this fix immediately.
> >
> > > Note I've started adding unittests, not too many right now -
> > > additions welcome.  At the moment just
> >
> > Cool, my "unit test" is currently that: installing the beast and
> > check, if the whole thing explodes ;-).
> 
> Hrmpf, I thought, the simple fix would be enough, but it's not that 
> easy.. Your latest changes made the whole thing explode!
> Biggest issue:
> The href value in location tag in primary.xml.gz is an absolute
> pathname now, with the net result, that subsequent yum runs cannot 
> correctly locate the rpm files.

I'm aware of this and am working to fix it - pointing out problems with
a unit test would be useful ;)

> If the cachedir argument is given as an relative path, it's moved up one 
> directory, compared to former behavior (0.4.5 put them on the same 
> level as "repodata"). I'm not sure, I like that move - pardon - I'm 
> sure, I don't like that move ;-)

Yeah we should use cmds['outputdir'] for everything.

Paul




More information about the Rpm-metadata mailing list