Use arraycmplen opcode

[Spencer-Comin]

arraycmp opcode with arrayCmpLen flag set is replaced with the opcode arraycmplen in eclipse/omr#6983.

[Spencer-Comin]

This has passed internal jenkins sanity tests for JDK11 and JDK8 on Z, x86, Power, and AArch64

[jdmpapin]

Please see my follow-up response to my earlier review comment - mentioning it here just for visibility

[jdmpapin]

Is this in draft state just in order to wait for eclipse/omr#6983 to be merged first? In case it is, the two PRs will actually need a coordinated merge

[jdmpapin]

Jenkins test sanity.functional,sanity.openjdk all jdk8,jdk11,jdk17,jdk20 depends eclipse/omr#6983

[jdmpapin]

Please look into the 32-bit Windows failures, which look related to these changes. They're crashing while compiling methods that can be expected to involve arraycmplen:

• jit/test/loopReduction/longMemCmp.test
• com/ibm/jit/JITHelpers.intrinsicIndexOfStringUTF16

Note that Java_java_lang_invoke_ThunkTuple_initialInvokeExactThunk is not actually related to these crashes. It shows up in the stack trace because of a lack of symbols in the code that's actually running.

Probably codegen will need to be adjusted for x86-32.

---

Other failures so far:

The Mac failure is #11930.

The AIX builds failed with "Deadlock condition if locked," which seems to be #8625.

The z Linux failure is a timeout in a CRIU test (cmdLineTester_criu_nonPortableRestore_Xdump_events_4):

Testing: Restore trace options test with -Xtrace before checkpoint - 2
...
***[TEST INFO 2023/06/20 01:06:47] ProcessKiller detected a timeout after 300000 milliseconds!***
...
***[TEST INFO 2023/06/20 01:07:47] executing kill -ABRT 1242972***
***[TEST INFO 2023/06/20 01:07:47] kill -ABRT signal sent***
Time spent executing: 360334 milliseconds
Test result: FAILED

I don't see that this particular issue is known, but I can scarcely imagine that it's related to these changes.

[Spencer-Comin]

Reverting this PR (and associated OMR PR eclipse/omr#6983) to draft while I resolve the failing tests

[Spencer-Comin]

I did some tweaking to this PR and eclipse/omr#6983 to resolve failures in x86-32, it should be good to go now

[jdmpapin]

ddrgen failed on Windows. This has been observed recently in the OMR acceptance build - see #18015

Jenkins test sanity.functional,sanity.openjdk win jdk21 depends eclipse/omr#6983

[jdmpapin]

The z Linux JDK11 sanity.openjdk failure is in java/nio/Buffer/Chew.java

[2023-08-28T19:20:25.270Z] java.lang.OutOfMemoryError: Direct buffer memory
[2023-08-28T19:20:25.270Z] 	at java.base/java.nio.Bits.reserveMemory(Bits.java:175)
[2023-08-28T19:20:25.270Z] 	at java.base/java.nio.DirectByteBuffer.<init>(DirectByteBuffer.java:123)
[2023-08-28T19:20:25.270Z] 	at java.base/java.nio.ByteBuffer.allocateDirect(ByteBuffer.java:317)

This is #17723

Jenkins test sanity.openjdk zlinux jdk11 depends eclipse/omr#6983

[jdmpapin]

The x86 Mac JDK17 sanity.functional failure is SharedCPEntryInvokerTests_2 crashing here:

getJNIMethodID+0x6c (0x0000000004F841BC [libj9vm29.dylib+0x401bc])
prepareToFixMemberNamesObjectIteratorCallback+0x158 (0x0000000005C186B8 [libj9jvmti29.dylib+0x336b8])
j9mm_iterate_region_objects+0xdd (0x000000000552E60D [libj9gc29.dylib+0xe560d])
j9mm_iterate_all_objects+0xbb (0x000000000552F40B [libj9gc29.dylib+0xe640b])
prepareToFixMemberNames+0x46 (0x0000000005C18546 [libj9jvmti29.dylib+0x33546])
redefineClassesCommon+0x31b (0x0000000005BECFCB [libj9jvmti29.dylib+0x7fcb])
jvmtiRedefineClasses+0xc2 (0x0000000005BECC42 [libj9jvmti29.dylib+0x7c42])
Java_com_ibm_jvmti_tests_redefineClasses_rc001_redefineClass+0xb1 (0x0000000005D4E941 [libjvmtitest.dylib+0x13941])

I'll look into this one myself

[Spencer-Comin]

The AIX failures are an infrastructure issue; they're all failing with "No space left on device".

The Test_openjdk8_j9_sanity.functional_x86-64_windows_Personal failure also looks like an infra issue: "Cannot delete workspace".

Besides the infra issues and the failure in Test_openjdk17_j9_sanity.functional_x86-64_mac_Personal that @jdmpapin is investigating, all other failures are due to known issues.

[jdmpapin]

The crash on Mac is unrelated. I've opened #18076 for it

[jdmpapin]

Jenkins test sanity.functional,sanity.openjdk all jdk8,jdk11,jdk17,jdk21 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk21_j9_sanity.openjdk_aarch64_linux_Personal failed in java/foreign/loaderLookup/TestLoaderLookup.java:

[2023-09-05T17:27:12.594Z] java.lang.reflect.InvocationTargetException
[2023-09-05T17:27:12.594Z] 	at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:115)
[2023-09-05T17:27:12.594Z] 	at java.base/java.lang.reflect.Method.invoke(Method.java:580)
[2023-09-05T17:27:12.594Z] 	at TestLoaderLookup.main(TestLoaderLookup.java:57)
[2023-09-05T17:27:12.594Z] 	at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:103)
[2023-09-05T17:27:12.594Z] 	at java.base/java.lang.reflect.Method.invoke(Method.java:580)
[2023-09-05T17:27:12.594Z] 	at com.sun.javatest.regtest.agent.MainWrapper$MainTask.run(MainWrapper.java:138)
[2023-09-05T17:27:12.594Z] 	at java.base/java.lang.Thread.run(Thread.java:1595)
[2023-09-05T17:27:12.594Z] Caused by: java.lang.NullPointerException: Cannot invoke "java.util.Optional.empty()" because "mapper" is null
[2023-09-05T17:27:12.594Z] 	at java.base/java.util.Optional.map(Optional.java:258)
[2023-09-05T17:27:12.594Z] 	at java.base/jdk.internal.foreign.abi.AbstractLinker.stripNames(AbstractLinker.java:279)
[2023-09-05T17:27:12.594Z] 	at java.base/jdk.internal.foreign.abi.AbstractLinker.downcallHandle0(AbstractLinker.java:102)
[2023-09-05T17:27:12.594Z] 	at java.base/jdk.internal.foreign.abi.AbstractLinker.downcallHandle(AbstractLinker.java:88)
[2023-09-05T17:27:12.594Z] 	at invoker.Invoker.invoke(Invoker.java:31)
[2023-09-05T17:27:12.594Z] 	at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:103)
[2023-09-05T17:27:12.594Z] 	... 6 more

This is #18074. The issue says it's a segmentation fault, but at least this failure linked from the issue manifests as an InvocationTargetException caused by NPE

Test_openjdk21_j9_sanity.openjdk_s390x_linux_Personal crashed in OOMEInAQS (#16659)

Test_openjdk21_j9_sanity.openjdk_x86-64_windows_Personal timed out in java/lang/Thread/virtual/stress/Skynet.java#default (#16728)

Jenkins test sanity.openjdk all jdk21 depends eclipse/omr#6983

[jdmpapin]

Jenkins test sanity.openjdk alinux,zlinux,win jdk21 depends eclipse/omr#6983

[jdmpapin]

Jenkins test sanity.openjdk alinux64,zlinux,win jdk21 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk11_j9_sanity.functional_x86-64_windows_Personal failed with an off-by-1 value in testThreadMXBeanProxy:

[2023-09-05T20:17:45.313Z] FAILED: testThreadMXBeanProxy
[2023-09-05T20:17:45.313Z] java.lang.AssertionError: expected:<25> but was:<24>

Looks like #17866

Test_openjdk11_j9_sanity.functional_x86-64_linux_Personal didn't launch properly, and there are no testList_0/1 links in the log, but there is this:

Cannot contact ub16x64j91: java.lang.InterruptedException

Jenkins test sanity.functional xlinux,win jdk11 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk11_j9_sanity.openjdk_x86-64_windows_Personal failed with RuntimeException: PortUnreachableException not thrown in sun/security/krb5/auto/Unreachable.java, which is #17743

Jenkins test sanity.openjdk win jdk11 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk21_j9_sanity.openjdk_s390x_linux_Personal segfaulted in OOMEInAQS again (#16659)

Test_openjdk21_j9_sanity.openjdk_ppc64_aix_Personal crashed and timed out in Skynet.java#default (#16728), which should now be excluded

Jenkins test sanity.openjdk zlinux,aix jdk21 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk11_j9_sanity.openjdk_x86-64_linux_Personal: java/security/SecureRandom/NoSync.java timed out while Thread-34 is reseeding something and other threads are waiting for it (#17576)

Jenkins test sanity.openjdk xlinux jdk11 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk21_j9_sanity.openjdk_x86-64_mac_Personal failed to run testList_0 due to an infra problem:

Cannot contact mac12-x86-1: java.lang.InterruptedException

Jenkins test sanity.openjdk xmac jdk21 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk17_j9_sanity.openjdk_aarch64_mac_Personal: testList_1 failed with Cannot contact mac13-aarch64-8: java.lang.InterruptedException, but it looks like all tests passed

[jdmpapin]

Test_openjdk8_j9_sanity.openjdk_x86-64_linux_Personal: testList_1 failed with what looks to be an infra problem:

19:23:29  Exception: java.io.IOException: Failed to copy file:/var/lib/jenkins/jobs/test.getDependency/builds/846/archive/jython-standalone.jar to /home/jenkins/workspace/Test_openjdk8_j9_sanity.openjdk_x86-64_linux_Personal_testList_1/aqa-tests/TKG/lib/jython-standalone.jar
...
19:23:29  Cannot run copyArtifacts from test.getDependency. Skipping copyArtifacts...
[Pipeline] fileExists
...
19:23:29  Exception: org.jenkinsci.plugins.workflow.support.steps.AgentOfflineException: Unable to create live FilePath for cent6-x64-5; cent6-x64-5 was marked offline: Timed out for last 5 attempts

Jenkins test sanity.openjdk xlinux jdk8 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk11_j9_sanity.functional_x86-64_linux_Personal: Cannot contact ub16x64j91

Jenkins test sanity.functional xlinux jdk11 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk17_j9_sanity.openjdk_x86-64_linux_Personal: testList_2:

AgentOfflineException: Unable to create live FilePath for cent7-x64-1; cent7-x64-1 was marked offline: Timed out for last 5 attempts

Jenkins test sanity.openjdk xlinux jdk17 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk21_j9_sanity.openjdk_ppc64le_linux_Personal: Skynet.java#default failed

Test_openjdk21_j9_sanity.openjdk_x86-64_linux_Personal: testList_2:

AgentOfflineException: Unable to create live FilePath for ub16x64j91; ub16x64j91 was marked offline: Timed out for last 5 attempts

Jenkins test sanity.openjdk xlinux,plinux jdk21 depends eclipse/omr#6983

[jdmpapin]

These two failed with Cannot contact cent7-x64-$n: java.lang.InterruptedException in the build step (Build_JDK{8,17}_x86-64_linux_Personal)

Jenkins test sanity.openjdk xlinux jdk8,jdk17 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk17_j9_sanity.functional_x86-64_mac_Personal: testList_0: Cannot contact mac10-x86-1: java.lang.InterruptedException

Jenkins test sanity.functional xmac jdk17 depends eclipse/omr#6983

[jdmpapin]

Test_openjdk8_j9_sanity.openjdk_x86-64_linux_Personal: timed out in java/util/concurrent/BlockingQueue/MultipleProducersSingleConsumerLoops.java (#10158)

The following jobs were aborted for some reason. Also, oddly, they still show as pending in the list of checks for this PR ("3 pending," "Pending — Build Started"). I might have looked at them sooner if there had been an ❌

• Test_openjdk11_j9_sanity.functional_x86-64_linux_Personal
• Test_openjdk11_j9_sanity.openjdk_x86-64_linux_Personal
• Test_openjdk21_j9_sanity.openjdk_x86-64_mac_Personal

[jdmpapin]

Jenkins test sanity.functional,sanity.openjdk xlinux jdk11 depends eclipse/omr#6983

[jdmpapin]

Jenkins test sanity.openjdk xlinux jdk8 depends eclipse/omr#6983

[jdmpapin]

Jenkins test sanity.openjdk xmac jdk21 depends eclipse/omr#6983

[Spencer-Comin]

Test_openjdk11_j9_sanity.functional_x86-64_linux_Personal failure is unrelated to arraycmplen. It's a CRIU failure, failing with "can't bind server address: Address already in use"

[jdmpapin]

The failure Spencer mentioned is #14706. There are no other failures in that job

The only other job that appears to have failed is Test_openjdk17_j9_sanity.openjdk_aarch64_mac_Personal, in which all tests actually passed as mentioned in #17382 (comment)