Updates:
Status: Fixed
Comment #1 on issue 230 by [email protected]: timer accuracy v one shot
timer v expiry time (issue_163.t)
http://code.google.com/p/memcached/issues/detail?id=230
Wow, a thorough report!
I'm sorry it was on a silly test like that. Some of the tests are subject
to timer failure; on a busy system or even on a fast system they'll fail
sometimes. I've been bumping the sleep from 2 to 3 when I find them. While
clock_handler will change the time to the actual amount of seconds passed,
it won't fire on a second alignment as you said. This means that
occassionally time bumps forward two seconds on a tick, or whatever.
Dustin has a "time travel" test method that's in uh... the 1.6 tree (I
think?) which completely removes the need for sleeping here at all.
In the meantime, I'll bump that guy to 3. Eventually we should have the
tests repaired anyway.
