Pierangelo Masarati wrote:
----- Howard Chuhyc@symas.com ha scritto:
I see that on server6, but server5 is correct. Also there appears to be an erroneous rename in server6.
Actually, I have been unable to reproduce this issue. To summarize: so far, I only got a failure of test019 with re24& db-4.7; no failures for re24& db-4.6, so I believe the db version could be unrelated, as the issue does not seem to be repeatable here. I'll keep trying and investigating. In fact, the failure I saw, related to a non-propagated delete, seems also to occur in other tests I'm carving for HEAD, but does not show up in test050 as I modified by adding an add and a delete for each participant in the MMR pool.
I've definitely identified a bug in rename handling, working on a fix. The detection of deleteOldRDN is wrong, it always results in deleteOldRDN being set even if the old RDN wasn't actually deleted. There are some other problems due to modRDN falling thru to modify as well...