On 07/20/2015 08:53 PM, Rafael Schloming wrote:
I'm fine going ahead with Gordon's fix. I don't have a lot of time to dig into the refcounting issue personally right now, but I'd at least leave the bug open until we have made it through a bit more testing. I have an uneasy feeling it (or something closely related) may pop up again if we push harder on testing.
You were right I'm afraid! I've seen further failures of a similar nature even with that fix now committed. They seem less frequent (previously I was seeing a core dump every run, now its more like 2 out of 10 or so).
My inclination is to revert the original fix for PROTON-905 for now, until more detailed testing and investigation can be carried out. The fix prevents a build up of memory on long running connections where multiple sessions (and links) are opened then closed. As reported this only happens with older versions of the broker. In any case it does also have a workaround (albeit a very inconvenient one) of restarting the connection every so often, at which point the memory is freed.
By contrast I don't understand enough about these crashes to suggest any workaround and the effects are more serious even for the latest broker.
