[Svnmerge] svnmerge & foreign repositories

[Eric Engstrom]

Dustin J. Mitchell wrote:
> OK, I've posted a wrap-up of the unmerged patches at
>   http://subversion.tigris.org/issues/show_bug.cgi?id=2869

I applied the patch (attachment 971, locid-final.patch) from the issue
above against the trunk (revision 35663) and have been using for a few
days now.

All seems fine to me, though in the interest of full disclosure, I've
only used it in the URL mode, not UUID.  I will be using it extensively
in the 'traditional' PATH mode in the next few days/weeks.

Also, I reviewed the changes made by the patch, though not being
intimately involved in the code base, I can only say that I didn't see
anything clearly wrong.

I will ask a perhaps silly question - why can't svnmerge.py determine
that it should use a different location type (i.e. URL) during init when
the source/remote repository root does not match the repository root of
the working copy?  IMO, it's no worse than the assumption that is made
now (for a 'PATH' location type) that the source is relative to the same
repository root as the working copy.

Clearly, one would have to leave the option to prefer using a repository
UUID.

If there's not a reason why we _should NOT_ do make the script
automatically choose a location-type of URL, then I'll happily work
through the changes required to make that assumption.  Otherwise, the
script should at minimum emit a warning when the URL provided does not
have the same repository root as the working copy - and perhaps fail to
perform the property modifications.

Regardless, I'm very happy to abandon my own script hacks to do similar
things and use this patched version of svnmerge.py for now.  Thanks so
much for your efforts and interest in keeping this idea alive for so long.

Eric