OpenJDK SADebugDTest.java test : handling of "Address already in use"

[Yasumasa Suenaga]

I've sent PR for this problem:
   https://github.com/openjdk/jdk/pull/6941

> `testResult` and `portInUse` are important variables in this logic,
> however they are overwritten in lambda expression - they won't affect
> to the caller.
> (I wonder why the test can compile without any error...)

They are class member, so they work fine.
However it is a little confusing, so I refactored it in this PR.


Thanks,

Yasumasa


2021-12-31 18:49 に Yasumasa Suenaga さんは書きました:
> Hi Matthias,
> 
> I think SADebugDTest.java wouldn't work that we expect.
> 
> ```
> 114                 Process debugd = startProcess("debugd", pb, null,
> 115                         l -> {
> 116                             if (!useRmiPort && l.contains(GOLDEN)) 
> {
> 117                                 testResult = true;
> 118                             } else if (useRmiPort &&
> l.contains(RMI_CONNECTOR_IS_BOUND + finalRmiPort)) {
> 119                                 testResult = true;
> 120                             } else if 
> (l.contains(ADDRESS_ALREADY_IN_USE)) {
> 121                                 portInUse = true;
> 122                             }
> 123                             return (l.contains(GOLDEN) || 
> portInUse);
> 124                         }, 20, TimeUnit.SECONDS);
> ```
> 
> `testResult` and `portInUse` are important variables in this logic,
> however they are overwritten in lambda expression - they won't affect
> to the caller.
> (I wonder why the test can compile without any error...)
> 
> Anyway, we should fix it. I will fix and will send PR.
> 
> 
> Thanks,
> 
> Yasumasa
> 
> 
> On 2021/12/30 17:29, Baesken, Matthias wrote:
>> Hello, I have a question regarding the test  SADebugDTest.java .
>> 
>> The test
>> 
>> https://github.com/openjdk/jdk/blob/master/test/hotspot/jtreg/serviceability/sa/sadebugd/SADebugDTest.java 
>> <https://github.com/openjdk/jdk/blob/master/test/hotspot/jtreg/serviceability/sa/sadebugd/SADebugDTest.java>
>> 
>> seems to handle already cases where “Address already in use” is found 
>> in the output (see line 120 ).
>> 
>> However in our nightly test,  the SADebugDTest.java   fails with this 
>> output :
>> 
>> LingeredApp stdout: [];
>> 
>> LingeredApp stderr: []
>> 
>> LingeredApp exitValue = 0
>> 
>> ----------System.err:(17/1318)----------
>> 
>> [debugd] Attaching to process ID 372 and starting RMI services, please 
>> wait...
>> 
>> [debugd] Error attaching to process or starting server: 
>> sun.jvm.hotspot.debugger.DebuggerException: 
>> java.rmi.server.ExportException: Port already in use: 39399; nested 
>> exception is:
>> 
>> *[debugd] java.net.BindException: Address already in use*
>> 
>> [debugd] at 
>> jdk.hotspot.agent/sun.jvm.hotspot.HotSpotAgent.setupDebugger(HotSpotAgent.java:384)
>> 
>> java.lang.RuntimeException: Expected message "RMI connector is bound 
>> to port 39399" is not found in the output.
>> 
>>    at SADebugDTest.testWithPid(SADebugDTest.java:132)
>> 
>>    at SADebugDTest.runTests(SADebugDTest.java:68)
>> 
>>    at SADebugDTest.main(SADebugDTest.java:60)
>> 
>>    at 
>> java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104)
>> 
>>    at java.base/java.lang.reflect.Method.invoke(Method.java:577)
>> 
>>    at 
>> com.sun.javatest.regtest.agent.MainWrapper$MainThread.run(MainWrapper.java:127)
>> 
>>    at java.base/java.lang.Thread.run(Thread.java:833)
>> 
>> JavaTest Message: Test threw exception: java.lang.RuntimeException: 
>> Expected message "RMI connector is bound to port 39399" is not found 
>> in the output.
>> 
>> JavaTest Message: shutting down test
>> 
>> Looks like  the *ADDRESS_ALREADY_IN_USE *string is checked below;  
>> however later in the coding  portInUse  is not checked but  the 
>> RuntimeException  “Expected message . . .”  is thrown :
>> 
>>                  Process debugd = startProcess("debugd", pb, null,
>> 
>>                          l -> {
>> 
>>                              if (!useRmiPort && l.contains(GOLDEN)) {
>> 
>>                                  testResult = true;
>> 
>>                              } else if (useRmiPort && 
>> l.contains(RMI_CONNECTOR_IS_BOUND + finalRmiPort)) {
>> 
>>                                  testResult = true;
>> 
>> *                            } else if 
>> (l.contains(ADDRESS_ALREADY_IN_USE)) {*
>> 
>> *                                portInUse = true;*
>> 
>>                              }
>> 
>>                              return (l.contains(GOLDEN) || portInUse);
>> 
>>                          }, 20, TimeUnit.SECONDS);
>> 
>>                  // If we are here, this means we have received the 
>> golden line and the test has passed
>> 
>>                  // The debugd remains running, we have to kill it
>> 
>>                  debugd.destroy();
>> 
>>                  debugd.waitFor();
>> 
>> *                if (!testResult) {*
>> 
>> *                    throw new RuntimeException("Expected message \"" 
>> +*
>> 
>> *                            RMI_CONNECTOR_IS_BOUND + rmiPort + "\" is 
>> not found in the output.");*
>> 
>> *                }*
>> 
>>              } while (portInUse); // Repeat the test if the port is 
>> already in use
>> 
>> Is this really intended ?  from the comments in the test I had the 
>> impression that in case of   “port in use”   / “address already in 
>> use”   the test should be repeated ?
>> 
>> Otherwise if this is intended, should we flag the test  with  @key 
>>  intermittent  ?
>> 
>> Thanks and best regards, Matthias
>>