How to communicate with upstream sources intelligently

[John Whitley]

John Arbash Meinel wrote:
> bzr branch TRUNK-1.0 TRUNK-1.1

For what it's worth, I just skip this step entirely and use a single  
branch as the source tracking branch.

E.g. here's the setup I created for tracking Instiki SVN:

instiki/            # a repo created with 'bzr init-repo --trees'
instiki/trunk       # the SVN tracking branch
instiki/bugNNN      # a branch made from trunk for work on bug number  
NNN
instiki/foo-feature # other branches of trunk for feature work
instiki/main        # a rollup-branch of trunk + all other  
'completed' work

In my case, I went to the trouble of doing the initial population of  
instiki/trunk using svn2bzr, then subsequently setting it up as an  
SVN checkout.  That way I have fine-grained history, and it's still  
relatively easy to track ongoing changes to SVN.

The workflow is similar, except that periodic merges happen from  
trunk to the other branches as needed (i.e. until they're accepted  
into SVN).

Much the same technique will work if you're only tracking released  
versions via tarballs.  E.g. create a trunk branch from the 1.0  
release.  Then create a new revision from the 1.1 release, that  
represents the collected changes, and so forth.

-- John