Re: git copies of cvs modules available

["S.P.Zeidler" <spz%serpens.de@localhost>]

Thus wrote Thomas Adam (thomas.adam22%gmail.com@localhost):

> On Mon, Jan 11, 2010 at 07:08:23PM +0100, S.P.Zeidler wrote:
> > 
> > Work for me is not the point (unless it's pointless work :).
> > The current git repo is situated on the ftp server, not on any cvs
> > server. The frequent updates of the git repo are sufficient extra load
> > already for that machine.
> 
> I might have missed something in previous emails, but I am still struggling
> to understand where all this extra load is.

A git repository of the state and history of src or pkgsrc of last August
is useless (mostly). Therefore, the git repository needs to be updated fairly
frequently. This update process adds load.

> > To create a perfect conversion, one would need to take cvsroot, parse the
> > commitlogs, add another information stream for repo moves, and replay the
> > evolution of the respective trees commit by commit, for every branch there
> > is. All existing tools just look at the cvs module itself, and take
> > guesses otherwise, and fail (depending on target repo by creating wrong
> > information or by falling over dead) because they lack information that is
> > just not in the cvs module, but vital to recreate historical checkouts.
> 
> Yes.  The git-cvsimport stuff just uses cvsps which is about as good as
> you're going to get, AFAIK.

Everything uses cvsps, and that is just not good enough.

regards,
        spz
-- 
spz%serpens.de@localhost (S.P.Zeidler)