Comment 3 for bug 688101

I've got a fix in the pipe.

The root issue is that we generate two conflicts involving the same file-id in some circumstances where a single one should be enough. The symptom here is therefore a fallout.

The circumstances are a file is deleted in one branch and renamed *and* modified in the other. In this case we generate a PathConflict for the deleted/renamed part and a ContentsConflict for the deleted/modified part. The later should be enough.

The fix I have filters out the path conflicts when there is a corresponding contents conflict (since it should be enough for the resolution needs in all cases). I have to write tests to verify I'm not missing some edge cases here. I also have to check whether the user get sufficient feedback to realize the file has also been renamed (which may occur before, after or at the same time as the modifications) but that may not be relevant here.