Philip Martin <philip@codematters.co.uk> writes:
> > Is it possible for svn to cd back to the original directory immediately
> > before invoking EDITOR?
>
> No, it was a deliberate decision to cd. The temporary file is created
> in the temp. area of the path to be committed. By cd'ing we don't
> need to put the path into a system() call, which means that the shell
> cannot interpret characters in the path as "special".
Oops, I'm a bit confused here.
Alexis is trying to solve this problem:
When you run "svn ci" in directory P, the editor is not necessarily
invoked in directory P, but rather in the lowest common root of all
the targets being committed, which may or may not be P.
The solution I thought he was proposing, and which I liked, is simply:
If P is also an svn working copy directory, then we can use P's
temp area. In other words, the target paths should always be
constructed relative to the directory the editor is running in, and
that directory should be the one from which the editor was invoked
when possible (i.e., when it's an svn-controlled directory), else
the lowest common root of all the targets when P is not an
svn-controlled directory.
This is still not a perfect solution, but until we have TMPDIR support
from APR, we're not going to get a perfect solution. At least this
one behaves "more reasonably, more often" :-). Right now, we *always*
use the lowest common root (the "anchor") of the committed targets,
even when we *could* do what Alexis wants and use P instead.
Is this at least an accurate description, or am I just missing some
big aspect of the problem?
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Aug 8 18:07:45 2002