RFR [8022584] Memory leak in some NetworkInterface methods

[Ivan Gerasimov]

Thank you Michael!

I'm working on the test.

Chris, if it's not too late, I would like to include a regtest into the 
change.
It will be ready in a few minutes and I'll send an updated webrev.

Thanks,
Ivan

On 08.08.2013 17:51, Michael McMahon wrote:
> The patch looks good to me. I guess a regression test isn't feasible.
> So, the bug will be tagged noreg-hard
>
> Michael
>
> On 08/08/13 14:39, Ivan Gerasimov wrote:
>> Thanks, David
>>
>> I've updated the webrev 
>> http://cr.openjdk.java.net/~igerasim/8022584/2/webrev/.
>>
>> On 08.08.2013 9:01, David Holmes wrote:
>>> Ivan,
>>>
>>> On 8/08/2013 2:05 PM, Ivan Gerasimov wrote:
>>>> David, Alan,
>>>>
>>>> I added checking for NULL  results and throwing OOMException if 
>>>> necessary.
>>>
>>> You don't need to throw it yourself:
>>>
>>>   JNU_ThrowOutOfMemoryError(env, NULL);
>>>
>>> Assuming a correct VM implementation if NULL is returned then an 
>>> OOME should already be pending and will be thrown as soon as your 
>>> native code returns to Java.
>>
>> It seemed to me that, JNU_ThrowOutOfMemoryError only throws an 
>> exception in a case one is not pending.
>> But I don't mind to remove it, relaying on the correct implementation 
>> of GetStringUTFChars.
>>
>> Sincerely yours,
>> Ivan
>>
>>> Thanks,
>>> David
>>>
>>>> I've also added some spaces along the code to improve indentation.
>>>>
>>>> Would you please review the updated webrev?
>>>> http://washi.ru.oracle.com/~igerasim/webrevs/8022584/1/webrev/
>>>>
>>>> Sincerely yours,
>>>> Ivan
>>>>
>>>>
>>>> On 08.08.2013 5:35, David Holmes wrote:
>>>>> On 8/08/2013 11:20 AM, Ivan Gerasimov wrote:
>>>>>> Thanks Alan!
>>>>>>
>>>>>> I've checked that and it turns out that GetStringUTFChars cannot 
>>>>>> return
>>>>>> NULL.
>>>>>> For allocation of the result string it calls AllocateHeap() with the
>>>>>> default EXIT_OOM fail strategy.
>>>>>> Thus, in case of being unable to allocate memory it simply stops VM.
>>>>>
>>>>> Ouch! That is a hotspot bug. GetStringUTFChars is supposed to throw
>>>>> OutOfMemoryError if it has memory issues, not abort the VM!
>>>>>
>>>>> I recommend that you check for NULL and/or a pending exception.
>>>>>
>>>>> David
>>>>>
>>>>>> Sincerely yours,
>>>>>> Ivan
>>>>>>
>>>>>> On 08.08.2013 5:05, Alan Bateman wrote:
>>>>>>> (cc'ing net-dev).
>>>>>>>
>>>>>>> The change looks okay to me. One suggestion (while you are 
>>>>>>> there) is
>>>>>>> to check the return from  GetStringUTFChars so that the name 
>>>>>>> returns
>>>>>>> when it fails with NULL.
>>>>>>>
>>>>>>> -Alan.
>>>>>>>
>>>>>>> On 07/08/2013 17:46, Ivan Gerasimov wrote:
>>>>>>>> Hello everybody!
>>>>>>>>
>>>>>>>> Some methods of NetworkInterface class were reported to leak 
>>>>>>>> memory.
>>>>>>>> http://bugs.sun.com/view_bug.do?bug_id=JDK-8022584 (not yet
>>>>>>>> available.)
>>>>>>>>
>>>>>>>> Examples are isUp() and isLoopback().
>>>>>>>>
>>>>>>>> Affected versions of jdk are 6, 7 and 8
>>>>>>>>
>>>>>>>> Would you please review a simple fix that removes the unnecessary
>>>>>>>> allocation?
>>>>>>>> http://cr.openjdk.java.net/~igerasim/8022584/0/webrev/
>>>>>>>>
>>>>>>>> Sincerely yours,
>>>>>>>> Ivan
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>
>
>
>