RFR: 8146416: java.lang.OutOfMemoryError triggers: assert(current_bci == 0) failed: bci isn't zero for do_not_unlock_if_synchronized
Jamsheed C m
jamsheed.c.m at oracle.com
Thu May 26 05:58:41 UTC 2016
(previous one was draft mail got accidentally sent)
Fix for https://bugs.openjdk.java.net/browse/JDK-6898462 was incomplete
Realloc failure deopt runtime exception was not handled properly! for
other exception based deopt, it may get a piggy backed feeling, but as
pending exception was not reset, pop logic will not work for first
frame.(InterpreterRuntime::exception_handler_for_exception exit check
for pending exception and forward again).
for normal deopts, Realloc failure runtime exception is not handled at all.
On 5/26/2016 10:34 AM, Jamsheed C m wrote:
> other responses inlined
> On 5/25/2016 11:57 AM, Vladimir Kozlov wrote:
>> Hi Jamsheed,
>> When sending RFR, please, explain the problem and the fix.
>> What about other platforms (including closed)?
Closed part was sent yesterday itself, AARCH64, and PPC was not done, as
i didnt have testing framework. i will intimate them, once review is done.
>> Don't use jump_cc because StubRoutines::forward_exception_entry()
>> address could be far. Other places use short conditional branch to
>> (or over) code which does jump to this function.
> jump_cc do handle far address with a warning, i will make the required
>> I don't think you can use call in SPARC code - you may mess up
>> frames. Use tail jump as in other cases.
> i can use tail jump.
>> On 5/24/16 10:51 PM, Jamsheed C m wrote:
>>> Request for review,
>>> bug id :https://bugs.openjdk.java.net/browse/JDK-8146416
>>> webrev: http://cr.openjdk.java.net/~jcm/8146416/
>>> testing: rbt pit, jprt;
>>> Best Regards,
More information about the hotspot-compiler-dev