Hi Ken,
Lukas and I were looking into an intermittent qa/312 failure he is
observing. If the sleep in that test is increased (to 3) its then
back to always passing (312 is a very simple pmie test).
But, seemed odd that pmie takes so long to startup and dump out the
result from that initial sample. ltrace pointed toward some calls
to nanosleep and following that led to unexpected time differences
for scheduled tasks vs the initial notion of "now". It turned out
those delays were due to some places where we do 'now = now + 1sec'
- see attached patch. Do you know the reason behind adding in that
one second difference there? (and is it safe to remove, or reduce,
as the patch does?)
thanks!
--
Nathan
slow-pmie.patch
Description: Text Data
|