On Mon, 2007-03-12 at 13:52 +0100, Timmerman, Bert wrote:

> I allready miss the $Id$ tag with the version, status and commit date in
> it, which is a quick and visible pointer to me, as to see something has
> changed in a schematic.

Funny.. I find myself detesting files with those tags in them, as they
cause false merge problems which I have to check out!

git is VERY flexible in what you can do with branching, and I think this
is what makes it great for lots of people working on gEDA at once.

> A question that come to mind is: Is there a decent way (as in not
> cumbersome) to revert from git to cvs and not losing the intermediate
> commits made in git ?

I believe there is, but I've not tried the automated approach myself.
I've been staging patches in git, and applying them individually (after
testing) to the noscreen branch in CVS.

> IMO that is what's needed when stable code written by one or more
> students is to be merged into the geda/gaf or pcb cvs repository.

The approach (and opinion) I have taken is that git is great for
working, incrementally checking in small changes. When it goes into CVS,
each commit should be a working change. Perhaps underlying that working
change are more than one git-commits. I don't necessarily want a 1-1
mapping between git and CVS. (Ok - strictly, I do... but I produce a new
git branch based at the end of the CVS branch I'm working on - then
apply those patches to CVS).

Extracting patches between any arbitrary commits in git is _easy_ and
there are scripts to automate their output.

> It's very likely that someone has to do a test to find the caveats, if
> any exist.

There are some, Peter Brett and myself have encountered some in keeping
a tracked git repository of CVS - but it is possible, and I've found it
really useful for what I've been doing.

Regards,

Peter C




_______________________________________________
geda-dev mailing list
[email protected]
http://www.seul.org/cgi-bin/mailman/listinfo/geda-dev

Reply via email to