On Oct 20, 2009, at 10:42 AM, Michal Hlavinka wrote:
On Monday 19 October 2009 23:55:20 Timo Sirainen wrote:
Can someone find an OS where the attached program doesn't work? It should print "success". So far tested for success: Linux 2.6, Solaris 10, FreeBSD 7.2, OpenBSD 4.2.
on Fedora 11 (Linux 2.6.30.9 64bit) it failed 3 times (of 200 runs) on Fedora rawhide (Linux 2.6.31.4 64bit) it failed 5 times (of 200
runs)
I think the reason why it's failing randomly sometimes is because the
test relies on sleep/usleep to accurately sleep the wanted amount of
time and everything else to take about zero time, and in those failure
cases it probably just doesn't work like that. So I'm not really
worried about those random failures. Increasing the sleep times by 10x
probably would get rid of them in the test.