On 15/01/2013 01:31, David Holmes wrote:
On 15/01/2013 7:12 AM, Rob McKenna wrote:
Simple enough fix but to be honest I'm not sure any value will *always*
work for the dead process waitFor(). Our testing infrastructure seems to
glide past whatever we consider to be acceptable tolerances.
http://cr.openjdk.java.net/~robm/8005618/webrev.01/
<http://cr.openjdk.java.net/%7Erobm/8005618/webrev.01/>
Using the latch seems reasonable but the existing wait/sleep times do
not. Why waitFor(10000) if the main thread is going to interrupt you
after a sleep(1000) ???
It's testing that Process.waitFor will be interrupted by
Thread.interrupt so it requires a thread to block in waitFor. Using
sleeps is always going to be problematic as the load on test machines is
unpredictable but I think Rob's proposed change does make this test a
bit more robust.
-Alan.
