Hi Alen (and everyone else)!
Thank you for your replies.
I have added an explicit-GC/sleep loop similar to the example fixes. I
have also added type parameters to avoid use of raw types as suggested.
The updated webrev is available below. Please have a look and let me
know if this is acceptable for pushing.
http://cr.openjdk.java.net/~dbuck/7042126/webrev.01/
Cheers,
-Buck
On 2013/01/31 23:05, Alan Bateman wrote:
On 31/01/2013 09:40, David Buck wrote:
Hi!
I was curious to see what others have done in the past and took a look
at about 15 different testcases for memory leaks in the jdk tree and
basically found 3 patterns:
Another one that you'll find is tests that use jmap to look at the
instance count of specific objects to see if they are increasing. I
personally do not like these tests as they are too targeted, also they
involve a second process and historically have been troublesome.
Anyway, my concern with the test as proposed is that it will might fail
intermittently. Last year there were fixes to several issues like this,
here are two that Eric Wang that come to mind:
http://hg.openjdk.java.net/jdk8/tl/jdk/rev/4ad204cc7433
http://hg.openjdk.java.net/jdk8/tl/jdk/rev/97eb7a4b1fdd
Adding a loop + sleep to your test should be fine for now.
I see Martin's suggestion about adding a supporting library to the test
suite for use by tests like this. As it's something that tests in may
areas then it may be worth exploring.
-Alan
