Looks fine. It definitely makes sense to return properly after notifying
OOM.
-JB-
On 30.1.2014 12:11, Markus Gronlund wrote:
Greetings,
Kindly asking for reviews for this very small fix:
Bug: https://bugs.openjdk.java.net/browse/JDK-8032518
Webrev: http://cr.openjdk.java.net/~mgronlun/8032518/webrev01/
Background:
Still a bit puzzled about the manifestations of the crashes when inspecting the
.mdmp files, which seems to be dereferencing a (debug) ResourceObj
allocation[0] cookie address (~allocation address) at point of crashing. In
addition, if the issue is an effect of not handling OOM correctly, I would
expect to see a _pending_exception off the problematic thread, but there seems
to be none. Also unknown why this seems to occur more on Windows x64 than any
other platform.
Testing:
I have iterated the testcase nsk/stress/jck60/jck60014 locally - without
suggested fixes I get about 10 crashes in about 300 runs. With fixes I am yet
to see any crashes, currently ~600 iterations.
I suggest to putback this first (since it should be fixed anyhow), to see the
effect, before any more time is spent on tracing this down.
Thanks
Markus
