Dave and I spent some time on this yesterday. He has the notes so can give a better summary than I can, but our feeling is that something bad is going on with replication and this patch at best masks it. However the problem only seems to arise within etap tests where database creation and replication happens in Erlang - do everything via the HTTP API and something happens between operations that allows it to work OK.
When Dave has time to post more detailed findings someone may look at them and spot the problem quickly. Otherwise I'm not sure of the best way forward: the community could take as long as it takes to get to the bottom of the problem, or we could say that the issue probably relates only to tests, remove the ones that break and put out a release while working on a solution in the background. Neither option is great. Nick On 7 March 2014 08:48, Dirkjan Ochtman <dirk...@ochtman.nl> wrote: > On Fri, Mar 7, 2014 at 9:43 AM, Alexander Shorin <kxe...@gmail.com> wrote: > > I'll check it on FreeBSD to make sure for that and need someone to > > confirm that problem is gone on OS X - just to avoid rc3 if we miss > > something. > > Great, thanks! >
