lp-open ought to check the parent location and bound location

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ian Clatworthy wrote:
> Public bug reported:
>
> lp-open currently checks the public location and push location. It ought
> to also check the parent location and bound location.
>
> In particular, I'd expect this to work:
>
> bzr branch lp:xxx
> cd xxx
> bzr lp-open
>
> but it doesn't.

For me, it often does, because I have public_location set in
locations.conf with appendpath.

A common reason for branching is because you intend to diverge. If I
branch lp:bzr, then start committing work on a new feature, lp-open
should not open lp:bzr, because that doesn't represent the contents of
my branch.

I can see a rationale for the public_location, because you've declared
where your branch is stored. I can see a rationale for push_location,
because that's a location you'll actually push your branch to.

The fact that you've branched from something does not give any certainty
that they are the same thing now.

> Ditto for:
>
> bzr checkout lp:yyy
> cd yyy
> bzr lp-open
>

Yes, this one makes sense, because people don't make checkouts in order
to diverge.

> If both the bound location and parent location are set, I think bound
> location should take precedence. In summary, I think the order ought to
> be:
>
> * public location
> * push location
> * bound location
> * parent location.

I would put bound location first, because bzr strongly discourages
divergence between a bound branch and its bound location. A bound
branch is much less likely to have a push location, and if it has one,
it could be outdated.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAksNUvwACgkQ0F+nu1YWqI3zAQCeLjr7gz3PqcP6DSSqNwbFIIxW
vVoAn3AFACcFG2RIRd3TWEweojZcPXIj
=S0ra
-----END PGP SIGNATURE-----