Workflow - Tracking upstream repository with local patches

[Andrew Bennetts]

Paul Moore wrote:
> On 13/03/2008, Andrew Bennetts <andrew at canonical.com> wrote:
> >  FWIW, I have an import of Python SVN trunk that I've made with bzr-svn at
> >  http://people.ubuntu.com/~andrew/python/trunk/.  I've also got an import of the
> >  py3k branch at http://people.ubuntu.com/~andrew/python/branches/py3k/.
> 
> Thanks, I may grab a copy. However, I'm more interested in getting a
> smooth means of integrating with svn, than specifically with getting a
> copy of the Python trunk. I presume your version worked more because
> bzr-svn (or likely the svn python bindings) is less flaky on Linux
> than on Windows, than because of anything specific you did?

I wish I knew why :)  You're right that I am on Linux.

I do track the stable branch of bzr-svn fairly closely and report problems back
to Jelmer as I run into them.  I've had Python trunk and py3k/p3yk imports
working for me for almost a year now (IIRC I first did it in May 2007).

> Of course, there's something to be said for just keeping it simple:
> 
> svn co
> bzr init
> bzr add
> bzr commit -m "Initial checkout"
> svn up
> bzr commit -m "Update to revision XXXX"
> 
> ... and forget about tracking all the history. It's not entirely clear
> if I'll ever really be that bothered about going through the history
> in detail while offline. Sometimes it's easy to lose sight of the
> woods for the trees :-)

This is worth considering.  After all, you can always keep the SVN repository
archived somewhere just in case you do want to refer to older history.
Personally, I think the best thing about bzr-svn is that it makes it possible to
migrate to Bazaar gradually, one developer at a time.  The importing of history
is great, but making incremental (and reversible) migration possible is even
better.

-Andrew.