Re: git copies of cvs modules available

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Fri, Nov 06, 2009 at 12:54:28AM -0200, Silas Silva wrote:
> On Wed, Nov 04, 2009 at 11:43:51PM +0100, Rhialto wrote:
> > Many of the files I looked at all had some clauses removed from the
> > license in the cvs checkout, while they were still present in the git
> > checkout. According to cvsweb, it was the most recent checkin that
> > removed the clauses.
> 
> I'm getting the same error at hunt in src and others in xsrc. That means
> we can't trust the git repository yet?

Something must be terribly wrong with the cvs-to-git process.  For example,
it's almost impossible for me to grasp how it could parse out all the CVS
versions and their branch relationships and yet fail to do something as
simple as get the right keyword substitutions into the last version of the
file -- after all, it has to know which version number it's working on, so
why can't it do the keyword substition, if it's going to extract the file
contents directly from the RCS file?  If it's not going to work on the RCS
file in raw form, but rather uses RCS or CVS to do that for it, it's easier
to see how it'd screw up the branch relationships, but much harder to see
how it'd get things like keyword substutions wrong.

I was pretty impressed with git after playing with it a little and after
reading up on how it worked.  I am *much* less so now that I've seen what
a mess it's made of the NetBSD repository and the kind of silly excuses
people are making for the speed and resource consumption issues that users
are reporting.

I am starting to think the FreeBSD team were pretty perceptive to go with
svn, as, perhaps, the least-worst available option...

Thor