Profiling numbers are wrong

[sebastianelsner]

From time to time I get these numbers in the logs:

Thu 07 Feb 20:08.02: Server load profiling:
Clients per second: 282.16, Now: 64 (processed 409600 connections in last 1451.65 seconds).
Prep: -18367622028115080.00, Proc: 5.18, Post: 18367622028115248.00, Total: 172.00 ms

-18367622028115080.00 for exampe seems to be really wrong. Where can this come from? Could this be an ovverflow?

[timurhai]

Hm. It seems like overflow. I`ll check it.

[timurhai]

Hi. Sorry can't reproduce or even imagine double arithmetic overflow. May be i am looking for error in a wrong place. It is time for 2.3.1 release due some important bug fixes. Lets keep this issue, but postpone it.

[sebastianelsner]

Ok, will do an upgrade soon-ish anyways. Might be a CentOS issue as well...

[timurhai]

Hi! So, have anybody noticed invalid profiling numbers last year?

[sebastianelsner]

Yes I see them all the time, but we still have not done any upgrades...