Re: copy history thoughts

Hi.

Without much thought, my expectation would be to get the
diff/log/whatever from the ancestor of the entity specified and
not from whatever was earlier at that path/location. In other
words: I wouldn't expect the commands to fail.

I can imagine that both uses makes sense and so the question
gets what kind of notation should refer to what? From what I
understand most (all?) cases refer to the location currently?

To me it sounds like some kind of relative/absolute relationship.
Or dynamic/static, follow/stay. The question is, whether the
specified path is the one of revision 50 or of the current
revision of the WC, which should be translated to revision 50.

That sounds similar to the some cases where an URL is expected,
but a WC path is given. It is translated to the respective URL.
This case is different in that both, the old and the new path
(may) make sense.

Just my 2 cents,

        Benjamin.

PS: Btw, I do not think that the question is related to whether
    rename is atomic or not. A simply "delete, then check in new
    file" operation leads to the same question.

On Wed 2002-10-30 at 21:47:33 +0300, Vladimir Prus wrote:
> I have a question concerning how copy history is traversed.
> For example, I've made some change in revision 72. In revision
> 76, I've moved directory to a different location, and now I'm
> at revision 78, and want to undo the rev72 change, I run:
>
> svn merge -r72:71
>
> and it gives me an error, because the current directory is now in
>
> /trunk/whale
>
> but in revision 72 it was
>
> /whale/trunk
>
> Okay, I can merge by specifying old paths. (Actually, at the
> moment I can't, for the reasons described in "svn diff
> regression bug" thread).
>
> But even if I can, it this a reasonable? Do I mean: merge the
> changes made in revision 72 to the closest ansestors of files
> which are here now 72, or merge changes made to file that were
> here in revision 72? Probably the answer is related to whether
> rename is atomic or not.
[...]