RFR: JDK-8206330: Revisit com/sun/jdi/RedefineCrossEvent.java

[Gary Adams]

On 10/17/18, 6:46 PM, Alex Menkov wrote:
> Hi Gary,
>
> - RedefineCrossEvent.java:
>
> you can drop "@modules jdk.jdi" as well.
> jdk.jdi is added to all test in the dir (it's specified in 
> test/jdk/com/sun/jdi/TEST.properties)
Done.
>
> about InstanceFilter test - it's strange that RedefineCrossEvent fails 
> while InstanceFilter does not.
> Maybe it's caused by different compiler options?
> InstanceFilter.java contains:
> @run compile -g InstanceFilter.java
>
> and RedefineCrossEvent.java has:
> @run build InstanceFilter
>
> Note that all other classes in RedefineCrossEvent are compiled with 
> "-g" option.
The InstanceFilter test has no problem running stand alone.
The RedefineCrossEvent test runs all the other tests in com/sun/jdi dir.
The problem is the redefine() that is called for each of the classes 
encountered.
See TestScaffold.java.

In jdk9 there was an introduction of jdk.internal classes. In this 
particular
test the jdk.internal.TerminatingThreadLocals class is observed, but the
redefine() processing has no compiled bytes to provide if the class were 
redefined.
The test does not expect to process system classes and already includes
filters to exclude processing of "java.", "com.", and "sun." prefixed 
classes.
>
> --alex
>
> On 10/17/2018 09:46, Gary Adams wrote:
>> The RedefineCrossEvent test has been on been on the ProblemList for a 
>> very long time.
>> In the past this test had some dependency on the Java EE modules, but 
>> they  were
>> deprecated for jdk9 and later removed completely in jdk11. This 
>> changeset to restore
>> it, removes the corba module reference and blocks the redefine 
>> operations
>> for jdk.internal classes that presented an issue when the 
>> RedefineCrossEvent
>> test launches the InstanceFilter test.
>>
>>    Issue: https://bugs.openjdk.java.net/browse/JDK-8206330
>>    Webrev: http://cr.openjdk.java.net/~gadams/8206330/webrev/index.html