When will CVS be replaced by modern version control system?

[Karel Zak]

On Sun, Nov 11, 2007 at 04:51:02PM +0100, Nils Philippsen wrote:
> On Sun, 2007-11-11 at 09:16 -0500, Jesse Keating wrote:
> > On Sun, 11 Nov 2007 15:07:12 +0100
> > Nils Philippsen <nphilipp at redhat.com> wrote:
> > 
> > > "commit to local repository, tag,
> > > push local changes to build repository, build".
> > 
> > You can't just tag locally.  You need to ensure that the same tag
> > hasn't been applied to the central repo by somebody else while you were
> > "away".
> 
> And how's that different from CVS? Of course, "make tag" should take
> things "upstream" to the "central" repository and yell if the tag
> already exists. Been there, done that, you might want to take a look at
> the Makefiles of my system-config-* packages(*). If a tag existed
> already, either the tagging itself or the subsequent push fails.
> 
> (*): E.g. http://hg.fedoraproject.org/hg/hosted/system-config-users --
> note that these have been written for slightly older Mercurial versions
> where using the same tag was possible without --force'ing things. They
> could be a bit simpler with current versions.
> 
> We could also resort to letting "koji build" do the tagging to keep
> things more controlled -- i.e. "koji build <some changeset id>" would
> checkout the package, attempt to "reserve"(**) the tag, build and tag on
> success which would obviate the need for "make force-tag". Of course,
> people should be prohibited from pushing changes to e.g. .hgtags for
> such a scheme.

 Note, git supports GPG signed tags -- it would be nice to use

    koji build <tag> <gpgkeyid>

 as a way how build a trusted rpm package.

    Karel

-- 
 Karel Zak  <kzak at redhat.com>