https://bugs.openldap.org/show_bug.cgi?id=7974

Howard Chu hyc@openldap.org changed:

What |Removed |Added ---------------------------------------------------------------------------- Resolution|--- |WONTFIX Status|UNCONFIRMED |RESOLVED

--- Comment #8 from Howard Chu hyc@openldap.org --- Fixing this implicitly inside LMDB is the wrong place. The calling app has to make the decision whether they can cope with breaking an atomic read txn into multiple pieces or not. Most applications that take a read txn expect it to stay atomic for the entire duration of their operation. Likewise, the calling apps are the only ones that can decide whether it's safe to kill any of their processes.