Re: some notes about svn_client__get_copy_source

On Wed, 30 May 2007, Kamesh Jayachandran wrote:

> Hi David,
>
> > $ svn cp branches/foo branches/baz
> > $ svn cp trunk/bar branches/baz/bar
> >
> >and adds within copies:
> >
> > $ svn cp branches/foo branches/baz
> > $ svn mkdir branches/baz/bar
> >
> >(In neither of these cases should branches/foo be the copy source of
> >branches/baz/bar; in the latter case, it has no copy source.)
>
> No.

"No" what?

> I verified in both the cases above our current 'get_copy_source' of
> 'branches/baz/bar' does not return '/branches/foo/bar'.
>
> For mkdir case, It returns nothing.
> For 'svn cp trunk/bar branches/baz/bar' case it returns 'trunk/bar'

So, we're working as expected?

> >Perhaps we should just expose svn_fs_closest_copy over RA instead of
> >reimplementing its logic in libsvn_client?
>
> Yes.

We'd previously discussed adding a 'svn log --copies-only' to handle
the use case of determining a path's copy source (as an improvement
over the code that is currently on trunk):

  http://subversion.tigris.org/servlets/ReadMsg?listName=dev&msgNo=125541

I've been planning to implement this as time allows. At the time, I
thought it would be easier than exposing closest_copy(). However, I'd
certainly be agreeable if someone wants to go that route instead;
seems like a generally useful API, in addition to being absolutely
perfect for this use case.

- Dan