RE: svn commit: r39546 - trunk/subversion/libsvn_wc

> -----Original Message-----
> From: Bert Huijben [mailto:bert-S0G5Crg1r3RmR6Xm/[email protected]]
> Sent: donderdag 24 september 2009 13:10
> To: dev-lmwclWVctOZK/[email protected]; svn-lmwclWVctOZK/[email protected]
> Subject: RE: svn commit: r39546 - trunk/subversion/libsvn_wc

> > -                             _("No such entry: '%s'"), name);
> > +  SVN_ERR(svn_wc__get_entry(&entry, db, local_abspath, FALSE, kind,
> > tweak_stub,
> > +                            scratch_pool, scratch_pool));
> 
> If we are updating a directory here the old code sometimes reads the
> stub, and sometimes the directory. The new code always reads the entry
> from the directory.
> 
> This is probably the change that fixed the merge test, because the url
> is always(?) available in the directory itself, but can be inherited in
> the stub.
> 
> >
> >    if (new_url != NULL
> >        && (! entry->url || strcmp(new_url, entry->url)))
> >      {
> > -      entry->url = apr_pstrdup(state_pool, new_url);
> > +      modify_flags |= SVN_WC__ENTRY_MODIFY_URL;
> > +      tmp_entry.url = new_url;
> >      }
> 
> And this code then detects: Hey, that is not a url change.
> 
> So it might also be the case that we forget to update the url in the
> parent stub.

Ok, forget this mail.. I misread this patch because wc__get_entry has the arguments in the opposite order than tweak_entries.

	Bert

> 
> 	Bert

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=495&dsMessageId=2399243