> -----Original Message----- > From: Daniel Shahaf [mailto:d...@daniel.shahaf.name] > Sent: donderdag 8 november 2012 08:07 > To: dev@subversion.apache.org > Subject: Re: svn commit: r1406870 - in /subversion/trunk/subversion: > libsvn_wc/update_editor.c libsvn_wc/wc_db_wcroot.c > tests/cmdline/special_tests.py > > Branko Čibej wrote on Thu, Nov 08, 2012 at 06:13:22 +0100: > > On 08.11.2012 05:26, Daniel Shahaf wrote: > > > Branko Čibej wrote on Thu, Nov 08, 2012 at 04:13:52 +0100: > > >> I believe that the correct approach would be to always treat a changed > > >> node kind (that's either the appearance/disappearance of the > svn:symlink > > >> property, or a change of the initial keyword in the special-file > > >> contents) as if it were a replacement, for the purpose of conflict > > >> detection and resolution, even though the node didn't actually get > > >> replaced. > > > Should we allow nodes to change their special-ness (namely: whether > > > they have svn:special set, and if yes what's the initial keyword) > > > without a replace? > > > > > > i.e., sure, current clients can do that --- "svn ps svn:special yes > > > COMMITTERS" --- so we'll have to handle that in libsvn_wc. But maybe > we > > > shouldn't allow any more instances of that. > > > > Good point. It might be a good idea to simply forbid setting or deleting > > svn:special explicitly, and also refusing to commit the file if its > > contents were modified in a way that changes the special type. > > > > Effectively that means you could only create special files through > > indirect means, e.g., by "svn add"ing a symlink. > > > > That wouldn't remove the work needed to fix the tree conflict bug, but > > it would make the svn:special semantics clearer. I personally don't > > think we have to worry about backward compatibility at that level; I'd > > rather treat the fact that you can directly manipulate svn:special as a bug. > > > > I would agree that being able to add/rm svn:special on a file node _that > already exists in the repository_ is a bug. But being able to do that > on a locally-added file is a feature --- it's what allows Windows users > to create versioned symlinks:
+1 on not allowing to change the symlinkness of existing files. And +1 on still allowing it on *new* files. In the update editor we could change the behavior of incoming symlink changes to be handled as a delete + add, introducing a tree conflict (and a local replacement of the incoming node to keep the original state) if there are local changes. I think I should be able to introduce this update editor behavior easily, but it is hard for me to test the on-disk symlink behavior. So I might break a few buildbot runs while working on this. Bert