Compatibility fix for Branch.{get,set}_parent()

[Wouter van Heyst]

On Thu, Jun 15, 2006 at 02:18:00PM -0500, John Arbash Meinel wrote:
> As part of the Encoding work, Wouter and I fixed Branch.set_parent() so
> that it takes a full URL, but only saves the relative portion if possible.
> So the contents of .bzr/branch/parent are always a URL, and relative if
> possible.
> It means we do weird things with old formats, where they included a
> fully absolute local path.
> So I added a workaround that would detect this, and change it into a
> full url before returning it.
> I kept running into this on my branches. I think it is a reasonable fix
> and comes with a test case. Can I get a +1?
> 
> John
> =:->

> === modified file 'bzrlib/branch.py'
> --- bzrlib/branch.py	2006-06-15 17:26:25 +0000
> +++ bzrlib/branch.py	2006-06-15 19:16:55 +0000
> @@ -1188,10 +1188,14 @@
>          assert self.base[-1] == '/'
>          for l in _locs:
>              try:
> -                return urlutils.join(self.base[:-1], 
> -                            self.control_files.get(l).read().strip('\n'))
> +                parent = self.control_files.get(l).read().strip('\n')
>              except NoSuchFile:
> -                pass
> +                continue
> +            # This is an old-format absolute path to a local branch
> +            # turn it into a url
> +            if parent.startswith('/'):
> +                parent = urlutils.local_path_to_url(parent.decode('utf8')
> +            return urlutils.join(self.base[:-1], parent)
>          return None

Will this dealing with old formats always be needed? I have a hard time
estimating how necessary a win32 codepath is. Considering no one has
complained yet, a +0.5 from me.

Wouter van Heyst