[merge] update --revision

[Matthieu Moy]

Robert Collins <robertc at robertcollins.net> writes:

> I think its better to offer --local if you want that.

If "update" has --local, then diff, cat, and any other command taking
--revision should have it.

> Trivial example of why:
>
> bzr update -r 4
>
> local branch has revisions: [A, B, C, D, E]
> master has [A, B, C, F, G]

Agreed, but it's not very different from

$ bzr cat foo -r 4

or 

$ bzr diff -r 4

Since the local branch takes precedence in all other cases, update
should not be the exception. And indeed, this -r 4 is most likely
written by the user by looking at the output of "bzr log", which shows
the local history.

Now, we could say simply that we match against the local branch, and
fail otherwise. But I considered it better to try again after fetching
from the master, but I admit one can argue against this.

One possible solution would be to fail if the revisionspec corresponds
to a revision not in local history, and if there have been local
commits.

-- 
Matthieu