[plt-dev] Push emails & private clones

From: John Clements (clements at brinckerhoff.org)
Date: Thu Apr 22 01:06:59 EDT 2010

On Apr 21, 2010, at 6:59 PM, Eli Barzilay wrote:

> I've changed the emails so pushes to a branch get mailed only to
> whoever pushed them.  I know that this is lacking compared to what we
> had in svn, where some people wanted to know about *all* commits,
> including branches.  I can come up with some new email for those
> things, but I think that I have a better overall plan for branches, so
> this change is temporary, until this plan is implemented.
> The plan is to not have any private branches -- unless (or until) we
> run into a branch that is truly a shared branch (in the past, the only
> branch that could qualify for that was the v4 branch).  I have
> implemented server-side cloning, which means that it is easy to get
> your own clone of the plt repository, which will be done at the
> server.  (Since it's on the same FS, git creates such clones using
> hard links for the data files, which means that it's a very cheap
> operation.)  I will also implement a way for people to customize their
> own private repositories, which means that after you create your
> private clone, you would be able to configure it to send notifications
> anywhere you want.

I want to make sure I understand what's being proposed.

One way to work on some new feature is to branch, work on the new feature, and then merge with master (or is it origin/master... the latter is the "remote tag", yes?). These branches are pushed to the main repository, but do not appear as part of the 'master' branch until explicitly merged.

A *different* way to work on some new feature is to create a separate clone of the repository, do the work there, and then merge this whole repository with the main one. If I understand correctly, this latter strategy is equivalent to working on a local machine and never doing a "push", except that the "local machine" repository can live on the server.

Is this a correct picture?

If so: what's the problem with branches? Is it that all of the code in the branches goes out to all of the plt developers, whereas in the second strategy it happens only when it gets merged back in with the trunk?

FWIW: Git transition ambient stress level: orange.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4669 bytes
Desc: not available
URL: <http://lists.racket-lang.org/dev/archive/attachments/20100421/c08486bd/attachment.p7s>

Posted on the dev mailing list.