NEWRELIC-3836 remove obsolete ibm workaround - support Semeru/OpenJ9 versions

[tbradellis]

Issue explained in these comments:
We never add the agent-bridge-datastore.jar in the ibmWorkaround path. However, even with that, there is still a bug in the IBM Workaround path where we fail to use the correct branch of logic for versions 8 and above.

         // Check for the IBM workaround
            boolean ibmWorkaround = IBMUtils.getIbmWorkaroundDefault();
            if (System.getProperty("ibm_iv25688_workaround") != null) {
                ibmWorkaround = Boolean.parseBoolean(System.getProperty("ibm_iv25688_workaround"));
            }

            ClassLoader classLoader;
            if (ibmWorkaround) {
                // For the IBM workaround lets just use the System ClassLoader
                //if you come down this path, we never end up
                //performing addBridgeJarToClassPath(inst, AGENT_BRIDGE_DATASTORE_JAR_NAME); hence NoClassDefFound for Record.Sql
                classLoader = ClassLoader.getSystemClassLoader();
            } else {
                ClassLoader agentClassLoaderParent = getPlatformClassLoaderOrNull();

                // Create a new URLClassLoader instance for the agent to use instead of relying on the System ClassLoader
                URL[] codeSource;
                if (isJavaSqlLoadedOnPlatformClassLoader(javaVersion)) {
                    //Semeru and other OpenJ9s for JDK 9+ are not getting to this path, and they should be.
                    //the logic in getIbmWorkaroundDefault() is faulty and causes anything above JDK 7 to not return false
                    URL url = BootstrapLoader.getDatastoreJarURL();
                    codeSource = new URL[] { getAgentJarUrl(), url };
                } else {
                    //If you get here, then the BootstrapLoader will have already
                    // performed addBridgeJarToClassPath(inst, AGENT_BRIDGE_DATASTORE_JAR_NAME); so, no problem (i.e. older JDKs did this).  
                    //IBM JDK 8 versions should have ended up here too, but did not because of the bug in  getIbmWorkaroundDefault() is faulty 
                  
                    codeSource = new URL[] { getAgentJarUrl() };
                }

                classLoader = new JVMAgentClassLoader(codeSource, agentClassLoaderParent);

                redefineJavaBaseModule(inst, classLoader);
                addReadUnnamedModuleToHttpModule(inst, agentClassLoaderParent);
            }
            
            

You can see in the if condition here that we only really check 1.7, so IBM JVMs 8 and above will always be true. 8 is above srNum 4. Also, 9+ uses platform classloader, so this is all just uneccesary. Had this check been correct, this issue probably wouldn't have come up. Still, best to rip this stuff out because it just is pointless complexity in the BootstrapAgent.

    public static boolean getIbmWorkaroundDefault() {
        try {
            if (isIbmJVM()) {
                String jvmVersion = System.getProperty("java.specification.version", "");
                int srNum = getIbmSRNumber();
                //We only set to false if 1.7 and above srNum
                //that means anything greater than jdk 7 will go through the true branch.
                if ("1.7".equals(jvmVersion) && srNum >= 4) {
                    // IBM Ticket says 7.SR3 is fixed, but no jvm can be found to test against
                    return false;
                }
                return true;
            }
            return false;
        } catch (Exception e) {
            return true;
        }
    }

This entire logic is obsolete anyway because the agent no longer supports Java 6 and 7.

Tests:
We need to add OpenJ9 and/or Semeru runtimes to our AITs

[tbradellis]

relates to:
#841

[tbradellis]

Related to ait work needed for these runtimes:
actions/setup-java#279
actions/setup-java#289

[tbradellis]

IBMj9 throws this as Division by zero and our tests look for the String / by zero.

[tbradellis]

The Classloader class used in the J9 implementation does not have the field scl.

[tbradellis]

All of these tests failed, but the failures were J9+Cassandra compatability. From what I can tell, anyone attempting to use Cassandra with OpenJ9 is already having issues.

[tbradellis]

lasp_strip_exceptions.py test needs to be reworked specifically for OpenJ9. The return value for ArithmeticException is different than openJDK. / by zero vs divide by zero

This will need to be done when the AIT work for IBM Semeru/OpenJ9 is implemented.

[tbradellis]

Our current gradle version is blocking us from testing with Semeru. Long story, but we had upgrading our gradle version on the road map anyway.
This will get us to 7.5.1. Gradle will work well with Semeru builds starting in 7.4
#997

[tbradellis]

errors.py:

--Warning: Unexpected Scoped Metrics--
Did not expect <Error ts:1662248600895
 path:Unknown
 message:divide by zero
 exception_type:java.lang.ArithmeticException
 params:{'userAttributes': {'str': 'dude'}, 'intrinsics': {'error.expected': True}, 'agentAttributes': {'request.uri': ''}, 'stack_trace': ['\tcom.nr.test.servlet.NewRelicApiServlet$1.run(NewRelicApiServlet.java:26)', '\tjava.lang.Thread.run(Thread.java:827)']}>
--/Warning: Unexpected Scoped Metrics--

This is similar to the SpanErrorsTest failure:
#993 (comment)

It looks like OpenJ9 uses the string divide by zero vs / by zero which is the string our tests compare against.

        self.assertIterAttributes(
            self.test_env.collector.get_errors(),
            {
                'path'           : 'Unknown',
                'message'        : '/ by zero',
                'exception_type' : 'java.lang.ArithmeticException'
            })

Various in inf_tracing_test.py

runtest.sh: running test tests/java/functionality/basic_features/inf_tracing_test.py with args 
/home/runner/work/newrelic-java-agent/newrelic-java-agent/agent-integration-tests/src/newrelic/agenttest/testenv.py:295: SyntaxWarning: "is not" with a literal. Did you mean "!="?
  if len(self.exceptions) is not 0:
/home/runner/work/newrelic-java-agent/newrelic-java-agent/agent-integration-tests/src/newrelic/agenttest/testenv.py:323: SyntaxWarning: "is not" with a literal. Did you mean "!="?
  if os.environ.get('TEST_FAIL_EXCEPTIONS', False) and len(self.unique_exceptions) is not 0:

Others as well:
https://github.com/newrelic/newrelic-java-agent/runs/8172652668?check_suite_focus=true

Unit tests for Java 8 and 11 OpenJ9 do not finish. It is unclear why.
https://github.com/newrelic/newrelic-java-agent/runs/8172302024?check_suite_focus=true