JDK 9 RFR of JDK-8143097: Test java/net/ipv6tests/UdpTest.java fails

[Chris Hegarty]

On 2 Nov 2016, at 11:06, Amy Lu <amy.lu at oracle.com> wrote:
> 
> On 11/2/16 6:17 PM, Chris Hegarty wrote:
>> Amy,
>> 
>>> On 2 Nov 2016, at 09:43, Amy Lu <amy.lu at oracle.com> wrote:
>>> 
>>> Please reviewthe patch for java/net/ipv6tests/UdpTest.java
>>> 
>>> bug: https://bugs.openjdk.java.net/browse/JDK-8143097
>>> webrev: http://cr.openjdk.java.net/~amlu/8143097/webrev.00/
>> I think what you have will probably be ok, given the 50% tolerance, but should’t
>> the end time be 10_000 + 2_000 ( the delay + the socket timeout )?
> 
> Thanks for your review, Chris.
> 
> Let me try explain more.
> 
> The end time here is for s1 socket timeout:
>    s1.setSoTimeout(10000);
> No matter how long the delay is in runAfter (which is for s.send(p)), s1.receive works when < 10000 (the given timeout); otherwise it will throw SocketTimeoutException.

Ah ok. So we do not expect receive to throw, it should actually receive the data.
In which case the 10,000 is really just an upper bound. Ok got it. Your changes
are fine as are.

-Chris.

> Thanks,
> Amy
> 
>> 
>> -Chris.
>> 
>>> This test fails intermittently in a test scenario for checking DatagramSocket (with SO_TIMEOUT enabled) 'receive'works even after a delay (but within the specified timeout):
>>> 
>>> 120     static void test2 () throws Exception {
>>> 121         s1 = new DatagramSocket ();
>>> ......
>>> 151         s1.setSoTimeout(10000);
>>> 152         runAfter (2000, new Runnable () { <<<< --- run after the given time (2000) has elapsed
>>> 153             public void run () {
>>> ......
>>> 156                     s.send (p);
>>> ......
>>> 158             }
>>> 159         });
>>> 160         t1 = System.currentTimeMillis();
>>> 161         s1.receive (new DatagramPacket (new byte [128], 128)); <<<< ---receive should works here
>>> 162         checkTime (System.currentTimeMillis() - t1, 4000);
>>> 
>>> The final checkTime method is for checking the time got (System.currentTimeMillis() - t1) is equal to (with 50% tolerance) the time expected (4000). This assumption is not correct. Test should check that the time got (System.currentTimeMillis() - t1) is between 2000 (the given delay) and 10000 (the given timeout).
>> 
>