> -----Original Message-----
> From: Markus Schaber [mailto:m.schaber_at_3s-software.com]
> Sent: vrijdag 29 juli 2011 12:42
> To: Stefan Sperling
> Cc: dev_at_subversion.apache.org
> Subject: AW: no error message if committing child of a deleted node
>
> Hi,
>
> Von: Stefan Sperling [mailto:stsp_at_elego.de]
> > On Fri, Jul 29, 2011 at 08:14:58AM +0200, Markus Schaber wrote:
> > > Von: Stefan Sperling [mailto:stsp_at_elego.de]
> > >
> > > > When trying to commit a child of a deleted node, svn commit says
> > > > nothing and just exists with status zero.
> > > >
> > > > This surprised me, because it complains loudly about committing
> > > > children of copies (see http://svn.haxx.se/dev/archive-2011-
> > 07/0773.shtml).
> > > > Shouldn't we raise some sort of error for the delete case, too?
> > >
> > > For me, this is not as surprising. Chidren of deleted nodes are also
> > > deleted. And it perfectly makes sense for the server to get the
> > > deletion of the child without getting the deletion of the parent.
> > >
> > > But getting the addition of a copied child without getting its
> > > (not-yet existing) parent is a completely different thing.
> >
> > Yes to all that, but why shouldn't Subversion at least point out that
> > "Cannot commit bar/foo because it is a child of the deleted directory
> bar;
> > you must commit bar instead"?
> > The structure of the local change is clearly incompatible with what
> the
> > user attempts to commit in this case.
>
> There is no conceptual reason to forbid the commit of a child of a
> deleted directory. The right thing to do here is to send the deletion of
> the child to the server.
This is not that simple as you state. The child might have been switched to
a different url (and some descendants to even other urls).
Would you like that url to be deleted, while a commit of the root of the
delete has never done that?
Bert
Received on 2011-07-29 13:02:01 CEST