[MERGE] bzr-dir phase 2

[John A Meinel]

Robert Collins wrote:
> If folk are up to it, theres an updated bzr-dir branch at
> http://people.ubuntu.com/~robertc/baz2.0/bzr-dir.
> 
> This adds the bzrdir api, which supports creating a bzrdir, and from
> there repositories, branches, working trees, branch references [thus
> allowing the library to create checkouts], setting a default branch
> format which is not the 'newest' branch format, and separates out the
> previously somewhat conflated 'duplicate with state' and 'create a new
> line of development' operations into (respectively) 'bzrdir.clone' and
> 'bzrdir.sprout'.
> 
> I propose that we consider the new branch formats in this 'unstable' and
> not upgrade any branches to them until:
>  * bzrdir phase 3 [the upgrade overhaul] is done
>  * Martins redone branch locking logic is integrated into the new
> formats.
> 
> From here I'm going to do checkouts in my checkout branch, for a little
> lightweight fun, and then do phase three, which will test the detached
> repository facility (I don't consider bzrdir 'complete' until thats able
> to work), and make upgrade handle new formats more easily. After that
> I'm going to start integration versioned file/bound branches/ other
> ready and outstanding branches.
> 
> I'm not attaching the diff, its uhm, a little large.
> 

I'm starting to go through it, so here goes...

What is the use case for get_branch_transport() based on branch_format?

The parameter for get_workingtree_transport() is labeled 'branch_format'
not 'workingtree_format'.

Should __init__() always be the first entry in a class?

You almost use epydoc format for the doc strings, but it would be:

:param _format: stuff

rather than just
_format: stuff


'BzrDir.open()' says "Open an existing *branch*, rooted at 'base'.

We probably shouldn't be raising NotBranchError from
BzrDir.open_containing(). We should raise NotBzrError, or somesuch.


Why does 'sprout()' always create a standalone branch?

Why does get_repository_transport() create the 'repository' directory?

You have 'BzrDirFormat5' having format 4 branches, and format 6
repositories (always). Shouldn't that be format 5 repositories? (I think
this is just a documentation bug.)


10k lines is a lot of code to review. But your results do speak for
themselves. (I didn't get through everything, but I tried).

John
=:->


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060212/4fbaee47/attachment.pgp