On Tue, 24 Jan 2006, Peter N. Lundblad wrote:
> On Tue, 24 Jan 2006, Julian Foad wrote:
>
> > lundblad@tigris.org wrote:
> > > http://subversion.tigris.org/issues/show_bug.cgi?id=2188
> > >
> > > ------- Additional comments from lundblad@tigris.org Mon Jan 23 02:27:32 -0800 2006 -------
> > > Taking.
> >
> > Do you have a plan for what to do about this issue? The last discussion I know
> > of was the thread in which I proposed the partial plan in this message:
> > <http://svn.haxx.se/dev/archive-2005-11/0596.shtml>. I still think that's a
> > good plan as far as it goes.
> >
> Well, I had been readign through the threads that were linked to from this
> issue, but I didn't see this mail. This was a good analyzes.
>
> My current plan is to remove the make_child semantics from copy/move,
> making it fail if the destination exists. This is what you propose, except
> for the overwrite flag. I'm also not doing the create parents stuff.
> That's another issue.
>
> Then create a compat wrapper that implements the create_child semantics on
> top of the new API and change the CL client to do the same. That's it for
> now. The benefit is to make the destination URL deterministic.
You might also want to keep in mind -- or fix -- the related issue of
"error due to rename when changing WC resource name case (only) on
case insensitive file system". I'm attaching a rough patch I started
on to add a regression test for this.
The file/directory name case issue may be more closely tied to "true
renames" (#898) or "renames by id" (#895) than #2188, but in any case,
it doesn't appear to have its own issue. Shall I create one?
--
Daniel Rall
- text/plain attachment: patch
- application/pgp-signature attachment: stored
Received on Wed Jan 25 01:07:03 2006