[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: What would be the best way to create "working repositories"?

From: Andy Levy <andy.levy_at_gmail.com>
Date: Tue, 29 Jun 2010 09:17:14 -0400

On Tue, Jun 29, 2010 at 08:54, Tom Malia <tommalia_at_ttdsinc.com> wrote:
> I’m looking for an easy way to allow programmers on a large project to
> create something equivalent to a personal “branch” on the main project that
> they can check their “Work In Process” (WIP) in and out of while they
> actively develop a feature or enhancement over the period of several days to
> several months.  Then when they feel it’s ready they need to do the
> functionally equivalent to merging the branch back into trunk.
> It appears that all this would be pretty simple if they literally did just
> create a branch on the main project.  However, for a number of different
> reasons, I’d prefer that they could have a separate repository for their
> WIP.

Developer/feature branches really are the best way to go here,
especially given your need to keep things simple and abstract the
actual Subversion usage away from your users as much as possible.

> The scenario I’m thinking of is one in which there’s a main company that
> controls the project and tightly manages the Subversion Repo for it.  They
> are granting subcontractors rights to check-out parts of the project from
> the main repository but they only want to allow those subcontractors to
> “check in” product that is ready for at least Beta testing.  In the mean
> time, the subcontractors want to be able to check their WIP files in and out
> of a repository over the course of the days or weeks that they are working
> on it.

Why not allow those subcontractors to commit to their private
branches, then have someone central who is the "gatekeeper" and either
merges from the branch to somewhere else, or accepts patches, at the
beta stage?

> One thought I had, that I’d like to know if there’s an easy way to do it or
> if I’m just heading down the completely wrong path.
> I’d like to be able to:
> 1)       “export” sections of the “main” repository
> 2)       Then add the directories that were just exported to a repository
> that would the “working repo”
> 3)       Check out the stuff just added to the “working repo” to a “working
> copy” (WC)
> 4)       Developer does their development in this WC, checking things into
> the “working repo” whenever they wish to make sure the have a snapshot or
> backup of their work
> 5)       When the developer feels they are ready to “merge” their work back
> in to the “main” repo they would do something like:
> a.       Check out the equivalent section of the “main” repo to a separate
> WC
> b.       Delete everything from that WC
> c.       Copy everything from the “WC” that came from their “Working repo”
> to the now “empty” WC from the main repo
> d.       Diff the current files in the “main” WC against the current version
> of the “main” repo and resolve conflicts
> e.       Check the WC from the “main” repo back into main
> f.         Delete all WC folders from their local machine
> g.       Delete the entire “working repo”
> One big flaw I see in this is, I’m not sure how deletes would be tracked,
> though I’m sure there’s plenty of other problems with this that I haven’t
> thought of.  But then I’m guessing I’m not the first person to want to do
> something like this.

The other big flaw is that it prevents changes in the central
repository from propagating down to your subcontractors. If you keep
it all in one repository, you can merge changes from trunk (or other
branches) to each developer's private branch. There's a lot of
overhead here and a huge margin for error & possibly even loss of
data. Portions of step 5 would be covered by svn_load_dirs.pl but
overall I'm having trouble understanding how this solves your
situation well while satisfying your below requirement.

> A key aspect is, the end result HAS to be REALLY easy to teach people to do
> and I need it in a Windows environment.  Currently most of the developers
> are not interested in knowing any more about Subversion than what they know
> right now, which is, it acts “kind of like a network share”.

I'm afraid this may break *any* solution proposed. At some point,
people *do* have to understand the tools their using, especially if
they're using them in a non-standard way.

SVK or git-svn may be able to meet your technical requirements without
too much trouble, but they'd probably break your "people"
Received on 2010-06-29 15:19:14 CEST

This is an archived mail posted to the Subversion Users mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.