[rfc][icedtea-web] RH976833 ClassLoader deadlock

[Andrew Azores]

On 09/17/2013 04:19 AM, Jiri Vanek wrote:
> On 09/16/2013 07:31 PM, Andrew Azores wrote:
>> On 09/16/2013 11:17 AM, Jiri Vanek wrote:
>>> On 09/16/2013 05:09 PM, Andrew Azores wrote:
>>>> On 09/16/2013 10:05 AM, Jiri Vanek wrote:
>>>>> (snip)
>>>>>
>>>>> I think you can post another verison (maybe final...) of patch
>>>>>
>>>>>
>>>>
>>>> Attached. It's the same as last time but with some comments added.
>>>>
>>>> Thanks,
>>>>
>>>
>>>
>>> Ugh. I would say even more javadoc needed. You should include 
>>> similar statements in javadoc which you are suing to persuade Omair 
>>> and Me that your synchronisation is correct.
>>>
>>> The only reason for this is that sometime in future one may 
>>> accidentaly revoke your patch in favour of more simple (as before) 
>>> synchronisation. And so easily return back to RH976833
>>>
>>> Sorry for troubeling
>>> J.
>>
>> Is this enough?
>>
>> Thanks,
>>
>
> I'm happy with it as it is. Please confirm push with Omair.
>
> J.

Anything in particular you think the unit tests should be doing, Omair? 
I still haven't been able to recreate the exact deadlock conditions in 
an offline reproducer. I do think that a regression test against this is 
important but I'm a bit stumped on how to produce one.

Thanks,

-- 
Andrew A