cmdLineTester_reflectCache seesm to not work on jdk25

[judovana]

    [javac]   where CAP#1 is a fresh type-variable:
    [javac]     CAP#1 extends Object from capture of ?
    [javac] /home/tester/aqa-tests/functional/cmdLineTests/reflectCache/src_15/test/reflectCache/Test_ReflectCache.java:107: error: cannot find symbol
    [javac] 			rootObject2 = unsafe.getObject(testConstructorB, constructorRootOffset);
    [javac] 			                    ^
    [javac]   symbol:   method getObject(Constructor<CAP#1>,long)
    [javac]   location: variable unsafe of type Unsafe
    [javac]   where CAP#1 is a fresh type-variable:
    [javac]     CAP#1 extends Object from capture of ?
    [javac] 8 errors

Considering the fate of unsafe.. Maybe to disable them?

I guess the fix willg o to https://github.com/eclipse-openj9/openj9/blob/master/test/functional/cmdLineTests/reflectCache/ ?

eg:

it diff
diff --git a/test/functional/cmdLineTests/reflectCache/playlist.xml b/test/functional/cmdLineTests/reflectCache/playlist.xml
index 805a907122..9050909df8 100644
--- a/test/functional/cmdLineTests/reflectCache/playlist.xml
+++ b/test/functional/cmdLineTests/reflectCache/playlist.xml
@@ -24,6 +24,12 @@
        <include>../variables.mk</include>
        <test>
                <testCaseName>cmdLineTester_reflectCache</testCaseName>
+               <disables>
+                       <disable>
+                               <comment>thisIssue</comment>
+                               <version>25+</version>
+                       </disable>
+               </disables>
                <variations>
                        <variation>NoOptions</variation>
                </variations>

[pshipton]

Are you trying to run this test on Hotspot? It's supposed to be specific to openj9/ibm.
https://github.com/eclipse-openj9/openj9/blob/master/test/functional/cmdLineTests/reflectCache/playlist.xml#L42

[pshipton]

The test is passing on openj9 for jdk25.
https://openj9-jenkins.osuosl.org/job/Test_openjdk25_j9_extended.functional_x86-64_linux_Nightly_testList_1/77/console

[judovana]

hi! Yes, I had noted the <impl>openj9</impl> <impl>ibm</impl> lines, and was wondering why it keeps running on hotspot. The aqavit should disable it, shouldnt it?

[judovana]

I'm not doing anything nasty.

export TEST_JDK_HOME=/usr/lib/jvm/java-25-openjdk/
export BUILD_LIST=functional
make compile # dies here
# make _functional

I thought the <impl> tag would exclude here. @pshipton anyway the isue is not for openj9/tests . Thanx!

[judovana]

@sophia-guo Any ideas?

[smlambert]

What is the System.getProperty('java.vm.name') and System.getProperty('java.vendor') of the JDK under test?
If those are undetermined, its possible we do not know that JDK_IMPL=hotspot

You can also try explicitly export JDK_IMPL=hotspot to see if the test will be excluded (which would be the expected behaviour if JDK_IMPL is known to be hotspot).

[judovana]

Hi!!

|  Welcome to JShell -- Version 25.0.1
...
jshell> System.getProperty("java.vm.name")
$1 ==> "OpenJDK 64-Bit Server VM"
jshell>  System.getProperty("java.vendor")
$2 ==> "Red Hat, Inc."

in both local case (fedora portables/rpms), and in remote case (rhel portables/rpms). The export JDK_IMPL=hotspot did not helped. make compile still fails.

[smlambert]

@judovana - can you also try doing a make clean and then export DYNAMIC_COMPILE=true before retrying your steps ?

As I presume this is a case where we are trying to compile code that will not compile unless it is openj9, and the playlist really governs the execution phase, not the compile phase.

[judovana]

export DYNAMIC_COMPILE=true` before retrying your steps ?

export TEST_JDK_HOME=/usr/lib/jvm/java-25-openjdk/
export BUILD_LIST=functional
# export JDK_IMPL=hotspot  # not needed anymore
export DYNAMIC_COMPILE=true
# make compile  # fails .. correctly? Error: cannot find the following tests: compile (note: group target such as sanity is not accepted inside testList)
make _functional  # now works as it should
....
FAILED test targets:
	ssl-tests_0 - Test results: passed: 4; failed: 1; did not meet platform requirements: 2 
		Failed test cases: 
TOTAL: 484   EXECUTED: 121   PASSED: 120   FAILED: 1   DISABLED: 22   SKIPPED: 341

Thanx for resolution. Is the export DYNAMIC_COMPILE=true recommended way then?

[smlambert]

Thanx for resolution. Is the export DYNAMIC_COMPILE=true recommended way then?

Yes, what that does is completely skip trying to compile stuff that is openj9 specific, when JDK_IMPL=hotspot
I should have caught on faster to the problem.

[judovana]

Thanx! Is there some documentation I can improve?

[smlambert]

Someone else is working on a documentation update that includes this topic (via #6772)