[MERGE] Speed up diff spec handling.

[Aaron Bentley]

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

John Arbash Meinel wrote:
> Actually, I explicitly did *not* want to fetch.

I want it not to fetch too, but I want to use it even more than I want
it to not fetch.  And I can't use it in its current state.

> There are other specs
> which can reference a revision_id that is not present, and is not
> accessible. (-r-1:../other_branch  as an example).

That is not an example if "not accessible", as I mean it.  The revno
spec provides a branch location, and that branch location can be used to
access the revision.

The branch spec does not provide a branch location.  This means that
while you can do as_revision_id, there's no reliable way to retrieve the
associated tree.

> I would rather leave branch: as not fetching, rather than continue to
> have all the lock_read/write bugs of having a revision spec mutate its
> branch and repository.

The "branch:" spec needs to be in one of two modes:
a) provides a branch location, does not fetch
b) fetches, does not provide a branch location

It is currently in neither mode, because sometimes it fetches, and
sometimes it does not, but it never provides a branch location.

This means that its RevisionSpec_branch.as_revision_id() is not a
substitute for RevisionSpec_branch.in_history(some_branch).rev_id, which
contradicts the rationale you gave for implementing it.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFIERSC0F+nu1YWqI0RAj/PAJ9tmYksNYLkPJDNuHGeZfX8zvdjcACdF/oV
7AQ0ct7YZN7UtGR1D71cfVo=
=hBQi
-----END PGP SIGNATURE-----