Re: [patch] vc-find-root with invert

[David Kastrup <dak@gnu.org>]

Justin Bogner <mail@justinbogner.com> writes:

> David Kastrup wrote:
>> Justin Bogner <mail@justinbogner.com> writes:
>>> However, for the other case, we need traverse upwards until we don't
>>> find the directory, returning the last one that does.
>>
>> Why?  CVS or SVN do not do this either.  Subdirectories with .svn in
>> them are self-contained work directories with associated repository
>> location.
>>
>>> As far as getting rid of invert, we could do that, and the function
>>> would then return something more reasonable than it does now, but it
>>> wouldn't actually find the root unless you happened to try the root
>>> first.
>>
>> There is no dedicated root in CVS or SVN.  You never need to look at
>> .. in order to do local operations.  You can move your directory out to
>> a different location under a different "root" and things will work just
>> the same from there.
>>
>> Which is pretty much the principal reason for every directory having its
>> own CVS or .svn subdirectories.
>>
>
> Is this still true if you need to do nonlocal operations?

With "local" I meant "local to the current tree location" whether in the
work directory or repository, not "local to this computer".

> If so then this is indeed a misfeature.

The only sort of justification I could imagine is when you are trying to
create a patch and/or work with a file set.  Then you want to have some
sort of common root.  However, just walking upwards is not guaranteed to
give you such a root: it is perfectly feasible to place some directory
under CVS for backup purposes, but have a non-registered subdirectory
_also_ under CVS with a completely different repository.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum