Fix a botched log-message

[Stephen J. Turnbull]

Matthieu Moy writes:

 > > But that is not a problem in principle because
 > 
 > Err, have you _really_ used git to say such thing ?

Do you understand the difference between "in principle" and "in
practice"?

 > Git considers commit sha1 as commit identity, and all the history
 > information is based on that. Re-creating a new commit with almost
 > the same content, but different sha1 means you totally break any
 > DAG-aware further commands (merging in particular).

No, you don't *totally* break anything, in fact.  The reason is that
the rebased branch is isomorphic to the old one.  git shows that it is
reasonable to compute these things on the fly.  When you do, you'll
end up at either the same common ancestor, or an isomorphic one (ie,
one which has the same chain of trees).

The problem of inconsistent branch refs when pulling from other repos
needs to be handled, and it's nontrivial.  But that's probably not
impossible.