Remote Merge

[Matthieu Moy]

John A Meinel <john at arbash-meinel.com> writes:

> Neither do I. I think it was suggested by Matthieu Moy that it should be
> a relatively simple operation.

I was thinking of pushing to a remote tree with no uncommited changes.
You're right, supporting pushing to a modified directory is much
harder.

The problem I didn't realize is that it's hard to ensure that a remote
tree has no local changes.

One solution (not very satisfying IMO, but ...) would be to suppose
that there have been no uncommited changes, and override them in case
there were.

> I personally don't have a problem having "bzr push" only update the
> .bzr/ data, since any "bzr pull" from that tree will regenerate the
> working directory from local contents.

On the other hand, this partly breaks the idea "working tree =
repository", since bzr would generate a tree which is not up to date
with its own repository. So, if some operations touch the .bzr without
updating the local tree, then I agree with you that there should be a
branch format without the working tree.

BTW, one very nice use for the "working tree = repository" it the
management of a website: the URL for the website is the same as the
one of the branch. A "push" able to update the working tree publishes
both the pages and the history.

-- 
Matthieu