On Thu, Jan 13, 2011 at 10:56:09AM +0000, Philip Martin wrote: > Stefan Sperling <s...@elego.de> writes: > > The scenario we're talking about is where people run a 1.7 server > > and use a 1.6 or earlier svnsync binary. > > > > I suppose in this case your hook script would help? > > No, I don't think it does. Daniel's change introduced new protocol > elements to get the new atomic behaviour, a old client using the old > protocol still gets the old non-atomic behaviour.
Right. I'll adjust the release notes then. Thanks. > We could change the behaviour associated with the old protocol, and then > we could backport that relatively small change to 1.6, and that would > allow the script to work. That would be the regression/improvement I > mentioned. I don't think that's necessary. Backporting this fix would give people one less reason to upgrade to 1.7 :) Anyone suffering from the race condition in 1.6 can use the known workaround (run svnsync jobs on one computer and synchronise the processes via file locks). Stefan
