[Adonthell-devel] Re: git, anyone?

[Chris Frey]

On Thu, Jan 22, 2009 at 07:57:43PM +0100, Kai Sterker wrote:
> That's something that wasn't so clear from the information in the
> Savannah git FAQ. However, it sounds like it would be possible to use
> git locally (and thus learn something new) without forcing any
> potential developer to use it.

That's true.  Multiple developers can use git and then commit changes
to CVS as the backend, and import from CVS again.

I keep my original git-cvsimport command in a script under .git,
and run it periodically to update from CVS.

It's probably worth running some experiments.  I've only done this as
a CVS writer, not with a multi-write CVS repo.


> I personally feel comfortable with CVS or SVN and I would give git a
> try too (and I may well do that locally at first). The thing is,
> people kept shaking their heads about us still using CVS. I know there
> are some disadvantages compared to SVN, but nothing I would consider
> essential. I know nothing about git in that respect (i.e. whether
> transactions are atomic, whether it's easy to restructure the project
> without losing history information, etc.), but I like how it is
> distributed and allows pushing/pulling stuff to/from other
> repositories.

I don't think CVS is a bad SCM, and "upgrading" from CVS to SVN seems like
a waste of time to me.

But once you use git, you realize just how slow CVS and SVN are.  Entire
new possibilities are opened to you, and you start using git in places
you hadn't thought of before.

Git's commits are atomic.  Each commit is a SHA1 sum, which is basically
globally unique.  It allows for a lot of optimization, including when
downloading updates, checkouts, etc.  It also means that data
integrity is part of the system by design.  As long as you know your own
top level SHA1 sum, you can verify that your entire adonthell history
is valid and has not been modified or hacked.  You could lose your entire
disk, clone from someone else, check the sum, and know the entire history is
the same as what you had before.

As for restructuring, git does track renames, but I haven't found a need
for that very much.  When using git, I view the entire project as a chain
of commits, rather than individual files that have a history.  The entire
tree is represented by a SHA1, and I look for difference between commits
rather than changes to a particular file.  As a long time user of CVS,
I don't feel I've lost anything, compared to what I've gained from git.
Indeed, I wouldn't go back.


> The question is, will others continue to complain that we now use
> something "unusual" and unknown to them? So far there has been
> positive feedback from Rian as well, so of the people actually active
> right now, most seem to like the idea.

Git does have a learning curve, and people do complain about that.

- Chris