On Wed, Aug 25, 2010 at 05:00:09PM +0100, Julian Foad wrote: > On Tue, 2010-08-24, Julian Foad wrote: > [...] > > What should happen? > > ------------------- > > > > I think the required changes are: > > > > * Commit should unconditionally bail out if there are any conflicts > > inside a node being committed. No more testing the 'keep_local' flag at > > this stage. > > I'm sure that's the behaviour we should have. The only thing I don't > understand is why Neels changed harvest_committables() to ignore > conflicts if 'keep_local' is set in r878027. That commit was about > fixing some problems with keep_local, but it's not clear to me why this > particular change was made. I assumed it was to preserve some 1.6.x > behaviour, but I've just tested with a 1.6.x branch build and it doesn't > allow the commit. > > I'm removing that condition, so the commit will fail if there are > unresolved conflicts, even inside a directory scheduled for delete. > This doesn't cause any test failures except tree_conflicts_tests 17 > (which is specifically about this behaviour) and it brings the commit > code back into line with 1.6.x in which > bail_on_tree_conflicted_children() was called unconditionally. > > I've committed this in 989189. > > If Neels or anyone can confirm this diagnosis, that would be great.
I guess that auto-resolving tree conflicts inside a locally deleted subtree was intended as a user convenience. But requiring an explicit 'svn resolve' invocation to resolve the conflict before commit is in line with the general behaviour of Subversion, so I think your change (reverting Neels changes) is sound. Stefan
