ando@sys-net.it wrote:
Full_Name: Pierangelo Masarati Version: HEAD OS: irrelevant URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (82.63.140.131) Submitted by: ando
After updating tests/run.in from CVS (rev 1.59) and reconfiguring, executing test050 with SYNCMODE=ro results in a failure. The failure is not strictly reproducible, but it always occurred to me, either affecting one or another delete that is not propagated to all participants to the MMR pool. I've raised the sleep time, to make sure there's no chance replication did not occur yet.
I wonder if refreshOnly makes any sense for a MMR pool. If it doesn't, it should not be allowed. If it does, deletes should be replicated correctly.
This sounds like a dup of #5843. Note that you cannot disallow MMR for refreshOnly, since servers must be able to refresh when starting up.