RFR (s) 8069005: Hotspot crashes in System.out.println with assert(resolved_method->method_holder()->is_linke, d()) failed: must be linked
Coleen Phillimore
coleen.phillimore at oracle.com
Fri May 1 23:52:39 UTC 2015
David,
Thank you for reviewing the change. I was trying to find a way to also
fix bug https://bugs.openjdk.java.net/browse/JDK-8078653 but that's more
complicated. So I want to use the original change to resolve this issue
with non-agent initialization. Now I need someone else to dare to
review it.
Thanks,
Coleen
On 4/30/15, 1:57 AM, David Holmes wrote:
> Didn't see this until after Reviewing the original patch ...
>
> On 29/04/2015 5:07 AM, Coleen Phillimore wrote:
>>
>> On 4/28/15, 2:20 PM, Aleksey Shipilev wrote:
>>> Hi,
>>>
>>> On 28.04.2015 21:02, Coleen Phillimore wrote:
>>>> Summary: move java.lang.Class initialization sooner in
>>>> bootstrapping the
>>>> jvm
>>>>
>>>> I tested this manually with the case in the bug. I don't know how to
>>>> add such a test without -Xbootclasspath:/a . Also ran
>>>> vm.quick.testlist, vm.defmeth.testlist and hotspot jtreg tests to make
>>>> sure this doesn't break anything.
>>>>
>>>> See bug for full discussion.
>>>>
>>>> open webrev at http://cr.openjdk.java.net/~coleenp/8069005/
>>>> bug link https://bugs.openjdk.java.net/browse/JDK-8069005
>>> This looks similar to:
>>> https://bugs.openjdk.java.net/browse/JDK-8078653
>>>
>>> Does the fix also addresses it?
>>
>> No, it doesn't. This case is that an agent has caused the
>> initialization of java.lang.Thread before the JVM has completed
>> initializations that it needs (like java.lang.String). There's not much
>> we can do but give a better error message.
>>
>> I think I will rework my change so that we get an exception at this
>> point (if possible) and this case has motivated that.
>
> Will wait to see what the looks like, but most (all?) the cases of
> this I have encountered have been caused by trying to throw an
> exception before the VM is capable of doing that - as all the
> necessary classes have not yet been linked.
>
> David
>
>> Thank you!
>> Coleen
>>
>>>
>>> Thanks,
>>> -Aleksey
>>>
>>>
>>
More information about the hotspot-runtime-dev
mailing list