[PATCH] (reminder) Fixes for revno:N:path

[Matthieu Moy]

Aaron Bentley <aaron.bentley at utoronto.ca> writes:

>>> Specifying a branch can also specify a file to merge.  I doubt this is
>>> the right thing to do.
>> 
>> True. Well, I already mentionned that I disliked this behavior. 
>
> You're not alone there, although my beef is that you can only specify
> one file this way.

Well, globally, I'd say that this way to specify a branch and a file
at the same time is something that seems great and user-friendly at
first, but it makes kind of hidden behavior for the user, and does not
solve all the use-cases.

Indeed, I find it relatively natural for local files, but writting
http://bazaar-vcs.org/bzr/bzr.dev/bzrlib/foo.py, I _think_
./bzrlib/foo.py in branch http://bazaar-vcs.org/bzr/bzr.dev/, then my
brain mixes it in a single identifier, and then bzr has to do the
transformation in the reverse direction. So, it's neither natural nore
convenient to me.

>> You
>> have no way to say file foo/bar in branch path/to/branch if
>> path/to/branch/foo is itself a branch. 
>
> Oh, this is because branches aren't trees, right?  So you can stick one
> branch inside another in ways that wouldn't make sense for a tree.  I
> can see your point.

Yes, but not only. path/to/branch/foo can be an old file which has
been deleted/renamed, and replaced by a branch. And with nested trees,
I do expect people to replace a directory by a sub-tree.

> Great, +1.

Thanks,

-- 
Matthieu