On Fri, Apr 6, 2012 at 8:53 AM, Greg Stein <gstein_at_gmail.com> wrote:
> On Apr 6, 2012 8:56 AM, "Hyrum K Wright" <hyrum.wright_at_wandisco.com> wrote:
>> On Thu, Apr 5, 2012 at 9:48 PM, Greg Stein <gstein_at_gmail.com> wrote:
>> > On Apr 5, 2012 2:43 PM, <hwright_at_apache.org> wrote:
>> >> Author: hwright
>> >> Date: Thu Apr 5 18:43:20 2012
>> >> New Revision: 1310005
>> >> URL: http://svn.apache.org/viewvc?rev=1310005&view=rev
>> >> Log:
>> >> On the ev2-export branch:
>> >> Use an Ev2-style driver to handle repos->repos copies.
>> >> Part of this commit is rather bogus, namely the bit where we still do a
>> >> delete+add. This *should* be a move, and will be adjusted to such in a
>> >> future commit, but the fact that it doesn't cause any tests to fail as
>> >> currently implemented is somewhat strange (the Ev2 internal checks
>> >> should
>> >> catch this, methinks).
>> > It can't.
>> > $ svn rm file
>> > $ svn copy file_at_REV newpath
>> > Totally fine sequence of operations. The user *could* have moved it, but
>> > that's a different story...
>> Those aren't repos->repos operations. This chuck of code is only for
>> repos->repos copy/move.
> Sure, but I don't see how Ev2 can possibly know that, and throw an error
> "should have used move".
Ah. I got my wires crossed and thought you were referring to the
comment in the code ("we should use 'move' here") and not the comment
in the log message ("Ev2 should detect this problem"). You are right,
Ev2 can't detect this, but the caller should still use the move action
to accomplish it.
uberSVN: Apache Subversion Made Easy
Received on 2012-04-06 16:17:25 CEST