Comment 6 for bug 861742

05:30 < lifeless> I will file upstream when I wake again
05:30 < lifeless> tell me, our log rotate stuff, does it matter if it takes a little longer (a couple of thread context switches..)
05:30 < lifeless> the rotate is 'mv foo; signal appserver process' right ?
05:31 < elmo> checking
05:32 < elmo> (but I'd imagine so)
05:32 < elmo> lifeless: yeah, it is - so it doesn't matter at all
05:32 < lifeless> great
05:32 < lifeless> I'll move the reopening out of the signal handler
05:33 < lifeless> set a non-threadsafe flag when signaled
05:33 < lifeless> and approximately the very next log message will read it and close+open
05:33 < lifeless> (and set the flag back to false)
05:34 < lifeless> this will avoid using threading code from the signal handler, introducing a small potential race if you rotate twice the second time within the rotation code for the first
05:35 < lifeless> said race will result in the second rotate being ignored (because it happened while the first rotate was doing its thing). Shrug.