fix for a keyerror in finding least common ancestors

[Aaron Bentley]

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

Michael Hudson wrote:
> This 3-line but 600-odd kilobyte file:
> 
>     http://python.net/crew/mwh/g.txt
> 
> contains:
> 
>  - on line 1, a sanitized-revid repr() of the _graph_ancestors of the
> graph with ghosts of the repository that I can reproduce the problem with
>  - on lines 2 and 2 the revids that should be used as arguments to
> find_unique_lca.

Hi Michael.

I believe I've fixed your problem, but it's an ordering-sensitive issue,
and your sanitized revids change the order, so I can't reproduce it with
the data you provided.

Can you please check out this branch, and let me know if it works for
your example?

http://code.aaronbentley.com/bzr/bzrrepo/bzr.ab

Aaron

(Yes, I hacked on bzr on my vacation.  In a cottage with no internet,
phone, electricity or running water)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGuIWo0F+nu1YWqI0RAhwNAJ9KT36OEQuM4rkmdslC5qapPUQs/wCfSv9M
ac6u2j6v8kEcbPi++O0fbrs=
=dr59
-----END PGP SIGNATURE-----