Just updated the test to workaround a seemingly unrelated platform
specific issue. (that only manifests on older kernels)
http://cr.openjdk.java.net/~robm/8065238/webrev.02/
-Rob
On 03/12/14 16:21, Rob McKenna wrote:
Hi folks,
Looking to fix a regression caused by 8042857. Basically the behaviour
in 8042857 is incorrect. This fix reverts to the previous behaviour
and attempts to beef up the tests a little around Ldap timeouts.
http://cr.openjdk.java.net/~robm/8065238/webrev.01/
The test itself looks quite complex but isn't really. There are two
executor pools. (scheduled and fixed) The fixed pool is for running
tests concurrently. The scheduled pool is for killing tests that test
ldap connects / reads where no timeout is set. (according to the spec
these should wait forever)
For these long running timeout tests, we schedule a thread to
interrupt the test after waiting for 20s. There are 3 of these long
running tests in total, hence the decision to run the tests in parallel.
I'm not averse to breaking this out into separate tests and a library
for the helper classes if people think it makes more sense to leave
the concurrency up to the test framework.
-Rob