Robert Greig wrote:
On 07/09/2007, Rafael Schloming <[EMAIL PROTECTED]> wrote:
The message ordering problem shows up for me when just publishing three
messages in a row, i.e. not really heavy load, just a quick burst. My
machine may be more prone to it than others though because I test on a
dual CPU system.
Based on my understanding of the problem, I would expect there to be
almost no ordering maintained under continuous load.
Hmm, this is indeed worrying. Our main test platform is a four way
dual core system so I am not sure why this hasn't been red flagged
sooner.
Something to investigate further when we're back in the office next week.
Keep in mind I haven't actually tested the M2 branch directly. However
the problem showed up on the trunk after the M2 -> trunk merge which is
why I strongly suspect it is on the M2 branch also. I posted
instructions in the JIRA on how to tweak the timings so as to reliably
replicate the problem. Without the tweaking it only happens once every 5
or 10 test runs which is a little bit annoying when debugging. I believe
Rob (the other one) said he was going to have a look early next week.
--Rafael
