revisit disabled tests
We should revisit disabled tests to see if the issues are resolved. If they are, we need to re-enable the tests. For these ones that do not have any comments, we should check them again.
For example: https://github.com/eclipse/openj9/blob/ad7c3f09279a54f8d4d253af03da2a2a31f86f8e/test/functional/cmdLineTests/modularityddrtests/playlist.xml#L26 https://github.com/eclipse/openj9/blob/e2c38649abb06128ee30c06691b6dcc9e19db55b/test/functional/VM_Test/playlist.xml#L148 https://github.com/eclipse/openj9/blob/ab7f03ab6faa069307a3ddedf09ac26f79b9702c/test/functional/cmdLineTests/pltest/playlist.xml#L175
@KapilPowar could you take a look at this issue? Thanks
@pshipton
While testing VM test suite for JDK8 we are seeing test case failures. https://github.com/eclipse/openj9/blob/e2c38649abb06128ee30c06691b6dcc9e19db55b/test/functional/VM_Test/playlist.xml#L148
Here is the grinder link
Test cases enabled -- recreateClassTest (1/20 failed ) CFdumptest (1/8 failed)
There was 1 failure:
17:13:10 1) testInnerClasses(com.ibm.j9.recreateclass.tests.RecreateClassCompareTest)junit.framework.AssertionFailedError: Incorrect number of mismatches found expected:<1> but was:<5>
17:13:10 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.compareClassFiles(RecreateClassCompareTest.java:63)
17:13:10 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.runCfdumpAndCompareTools(RecreateClassCompareTest.java:74)
17:13:10 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.runTest(RecreateClassCompareTest.java:86)
17:13:10 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.runTest(RecreateClassCompareTest.java:79)
17:13:10 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.testInnerClasses(RecreateClassCompareTest.java:154)
17:13:10 at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
17:13:10 at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
17:13:10 at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
17:13:10
17:13:10 FAILURES!!!
17:13:10 Tests run: 20, Failures: 1, Errors: 0
--------------------------------------------------------------------
CFdumptest (1/8 failed)
There was 1 failure:
17:13:11 1) testInvalidClassFile(com.ibm.j9.cfdump.tests.lineardump.LinearDumpTests)junit.framework.AssertionFailedError: Expected: <null> but was: bad magic number
17:13:11 at com.ibm.j9.cfdump.tests.lineardump.LinearDumpTests.testInvalidClassFile(LinearDumpTests.java:123)
17:13:11 at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
17:13:11 at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
17:13:11 at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
17:13:11
17:13:11 FAILURES!!!
17:13:11 Tests run: 8, Failures: 1, Errors: 0
Please confirm if it is product issue.
How do I re-run these tests? The grinder link is expired. When I try to run them every test fails with Error: cfdump not found on java.library.path!. I see that cfdump is part of the test image.
https://openj9-jenkins.osuosl.org/view/Test/job/Grinder/2649
In playlist.xml file please try to add cfdump in the lib path - using $(ADD_JVM_LIB_DIR_TO_LIBPATH) \ in <command>
for the test cases.
e.g.
<command>$(ADD_JVM_LIB_DIR_TO_LIBPATH) \ $(JAVA_COMMAND) $(JVM_OPTIONS) \
Here are latest Grinder Links aarch64_linux --Java 8 x86-64_linux -- Java 8 x86-64_windows -- Java 8
Failure details from recreateClassTest
13:15:46 Start testInnerClasses
13:15:46 Running command: /home/jenkins/workspace/Grinder/openjdkbinary/openjdk-test-image/openj9/cfdump -t:/home/jenkins/workspace/Grinder/aqa-tests/TKG/output_16917741452730/recreateClassTest_0/InnerClassesTest /home/jenkins/workspace/Grinder/aqa-tests/TKG/output_16917741452730/recreateClassTest_0/InnerClassesTest.class
13:15:46 cfdump: Wrote 1212 bytes to output file /home/jenkins/workspace/Grinder/aqa-tests/TKG/output_16917741452730/recreateClassTest_0/InnerClassesTest.j9class
13:15:46 Inner class access flags differ: 0X8, 0X0
13:15:46 Aug 11, 2023 5:15:46 PM com.ibm.j9.test.tools.classfilecompare.ClassFileCompare reportDifference
13:15:46 WARNING: Inner class access flags differ: 0X8, 0X0
13:15:46 Inner class inner name differ: LocalClass, null
13:15:46 Aug 11, 2023 5:15:46 PM com.ibm.j9.test.tools.classfilecompare.ClassFileCompare reportDifference
13:15:46 WARNING: Inner class inner name differ: LocalClass, null
13:15:46 Inner class access flags differ: 0X8, 0X0
13:15:46 Aug 11, 2023 5:15:46 PM com.ibm.j9.test.tools.classfilecompare.ClassFileCompare reportDifference
13:15:46 WARNING: Inner class access flags differ: 0X8, 0X0
13:15:46 Inner class access flags differ: 0X9, 0X0
13:15:46 Aug 11, 2023 5:15:46 PM com.ibm.j9.test.tools.classfilecompare.ClassFileCompare reportDifference
13:15:46 WARNING: Inner class access flags differ: 0X9, 0X0
13:15:46 Inner class inner name differ: InnerClass, null
13:15:46 Aug 11, 2023 5:15:46 PM com.ibm.j9.test.tools.classfilecompare.ClassFileCompare reportDifference
13:15:46 WARNING: Inner class inner name differ: InnerClass, null
13:15:46 Mismatches spotted: 5
@TobiAjila can someone pls take a look at this recreateClassTest failure in the previous comment, it appears we may need a fix.
My changes to run the tests and add debug are in https://github.com/pshipton/openj9/tree/tests
@hangshao0 Can you please take a look at this
Re-enabling CFdumptest via https://github.com/eclipse-openj9/openj9/pull/17939
I will look at this.
I've manually run cfdump on the inner class. Looking at the generated j9class and original class file:
cfdump -d 'InnerClassesTest$InnerClass.class'
Magic: CAFEBABE
Version: 52.0
Constant Pool Size: 17
Access Flags: 0x21 (
access: public super )
ThisClass: 2 -> InnerClassesTest$InnerClass
SuperClass: 3 -> java/lang/Object
Interfaces (0):
Fields (0):
Methods (1):
Name: 4 -> <init>
Signature: 5 -> ()V
Access Flags: 0x1 (
access: public )
Attributes (1):
Code:
Max stack: 1
Max Locals: 1
Code Length: 5
Caught Exceptions (0):
Attributes (1):
LineNumberTable:
pc: 0 line number: 2
Attributes (2):
SourceFile:
Source File: 9 -> InnerClassesTest.java
InnerClasses:
Inner Class: 2 -> InnerClassesTest$InnerClass
Outer Class: 11 -> InnerClassesTest
Inner Name: 13 -> InnerClass
Inner Class Access Flags: 0x9 (
access: public )
cfdump -d 'InnerClassesTest$InnerClass.j9class'
Magic: CAFEBABE
Version: 52.0
Constant Pool Size: 17
Access Flags: 0x21 (
access: public super )
ThisClass: 5 -> InnerClassesTest$InnerClass
SuperClass: 2 -> java/lang/Object
Interfaces (0):
Fields (0):
Methods (1):
Name: 15 -> <init>
Signature: 3 -> ()V
Access Flags: 0x1 (
access: public )
Attributes (1):
Code:
Max stack: 1
Max Locals: 1
Code Length: 5
Caught Exceptions (0):
Attributes (1):
LineNumberTable:
pc: 0 line number: 2
Attributes (2):
InnerClasses:
Inner Class: 5 -> InnerClassesTest$InnerClass
Outer Class: 12 -> InnerClassesTest
Inner Name: 8 -> InnerClass
Inner Class Access Flags: 0x9 (
access: public )
SourceFile:
Source File: 13 -> InnerClassesTest.java
I see the access flags and inner class names are the same. The differences might be from the outer class that cause the test to fail.
cfdump -d InnerClassesTest.class
Magic: CAFEBABE
Version: 52.0
Constant Pool Size: 24
Access Flags: 0x21 (
access: public super )
ThisClass: 3 -> InnerClassesTest
SuperClass: 4 -> java/lang/Object
Interfaces (0):
Fields (0):
Methods (2):
Name: 9 -> <init>
Signature: 10 -> ()V
Access Flags: 0x1 (
access: public )
Attributes (1):
Code:
Max stack: 1
Max Locals: 1
Code Length: 5
Caught Exceptions (0):
Attributes (1):
LineNumberTable:
pc: 0 line number: 1
Name: 13 -> getLocalClass
Signature: 14 -> ()Ljava/lang/Class;
Access Flags: 0x9 (
access: public static )
Attributes (2):
Code:
Max stack: 1
Max Locals: 0
Code Length: 3
Caught Exceptions (0):
Attributes (1):
LineNumberTable:
pc: 0 line number: 8
Signature:
Signature: 16 -> ()Ljava/lang/Class<*>;
Attributes (2):
SourceFile:
Source File: 18 -> InnerClassesTest.java
InnerClasses:
Inner Class: 5 -> InnerClassesTest$InnerClass
Outer Class: 3 -> InnerClassesTest
Inner Name: 6 -> InnerClass
Inner Class Access Flags: 0x9 (
access: public )
Inner Class: 2 -> InnerClassesTest$1LocalClass
Outer Class: (not a member)
Inner Name: 8 -> LocalClass
Inner Class Access Flags: 0x0 (
access: default )
cfdump -d InnerClassesTest.j9class
Magic: CAFEBABE
Version: 52.0
Constant Pool Size: 22
Access Flags: 0x21 (
access: public super )
ThisClass: 7 -> InnerClassesTest
SuperClass: 3 -> java/lang/Object
Interfaces (0):
Fields (0):
Methods (2):
Name: 21 -> <init>
Signature: 14 -> ()V
Access Flags: 0x1 (
access: public )
Attributes (1):
Code:
Max stack: 1
Max Locals: 1
Code Length: 5
Caught Exceptions (0):
Attributes (1):
LineNumberTable:
pc: 0 line number: 1
Name: 12 -> getLocalClass
Signature: 13 -> ()Ljava/lang/Class;
Access Flags: 0x9 (
access: public static )
Attributes (2):
Code:
Max stack: 1
Max Locals: 0
Code Length: 3
Caught Exceptions (0):
Attributes (1):
LineNumberTable:
pc: 0 line number: 8
Signature:
Signature: 4 -> ()Ljava/lang/Class<*>;
Attributes (2):
InnerClasses:
Inner Class: 20 -> InnerClassesTest$InnerClass
Outer Class: 7 -> InnerClassesTest
Inner Name: (anonymous) <------------------------------------- this is different from the original class
Inner Class Access Flags: 0x0 ( <------------------------------------- this is different from the original class
access: default )
Inner Class: 1 -> InnerClassesTest$1LocalClass
Outer Class: (not a member)
Inner Name: (anonymous)
Inner Class Access Flags: 0x0 (
access: default )
SourceFile:
Source File: 5 -> InnerClassesTest.java
The difference is from the inner class attribute of the outer class. The inner class simple name and access flags are not preserved in the romClass of the outer class: https://github.com/eclipse-openj9/openj9/blob/949a4df06ef763ed4f3a83faf8d9534f911f4d07/runtime/bcutil/ClassFileWriter.cpp#L1054-L1056
If we decided to not record these information in the romClass in #13199, the test needs to be updated to accept such differences in .class and .j9class files.
@ChengJin01 ^^
CFdumptest is only enabled for JDK8. I think it is able to run against all versions.
https://github.com/pshipton/openj9/blob/master/test/functional/VM_Test/playlist.xml#L167
@KapilPowar could you run CFdumptest on JDK11 and JDK17 when you have time? Thanks
I tried other versions but there are problems that would need to be resolved.
Grinder test runs for CFdumptest
JDK11 x86-64_linux
JDK11 x86-64_mac
JDK11 x86-64_windows
JDK17 x86-64_linux
JDK17 x86-64_mac
JDK17 x86-64_windows
Thanks @KapilPowar . Could also run on other missing platforms in Grinder? Thanks
Grinder test runs for CFdumptest
JDK17
s390x_linux --UNSTABLE
aarch64_mac --Success
ppc64_aix --UNSTABLE
JDK11 aarch64_mac -- Success s390x_linux--UNSTABLE ppc64_aix --UNSTABLE
@llxia @pshipton FYI
Grinder test run for recreateClassTest still results in same error -- JDK8 x86-64_linux
There was 1 failure:
17:01:36 1) testInnerClasses(com.ibm.j9.recreateclass.tests.RecreateClassCompareTest)junit.framework.AssertionFailedError: Incorrect number of mismatches found expected:<1> but was:<5>
17:01:36 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.compareClassFiles(RecreateClassCompareTest.java:63)
17:01:36 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.runCfdumpAndCompareTools(RecreateClassCompareTest.java:74)
17:01:36 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.runTest(RecreateClassCompareTest.java:86)
17:01:36 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.runTest(RecreateClassCompareTest.java:79)
17:01:36 at com.ibm.j9.recreateclass.tests.RecreateClassCompareTest.testInnerClasses(RecreateClassCompareTest.java:154)
17:01:36 at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
17:01:36 at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
17:01:36 at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
17:01:36
17:01:36 FAILURES!!!
17:01:36 Tests run: 20, Failures: 1, Errors: 0
17:01:36
17:01:36 -----------------------------------
17:01:36 recreateClassTest_0_FAILED
17:01:36 -----------------------------------
I assume this can be moved forward.
Re https://github.com/eclipse-openj9/openj9/issues/10917#issuecomment-1725335609, we're waiting for @ChengJin01 to have time to take a look, based on https://github.com/eclipse-openj9/openj9/issues/10917#issuecomment-1681233241
I agree with @hangshao0 's investigation & conclusions given we don' record the access flags in the romClass by default.
So the next step is to update the test itself to ignore those differences. @TobiAjila
@llxia Grinder test run for cmdLineTester_forceLazySymbolResolution JDK17 x86-64_linux
Grinder test run for cmdLineTester_jvmtitests_hcr_OSRG_nongold JDK17 x86-64_linux JDK17 x86-64_mac JDK17 x86-64_windows
@KapilPowar if the Grinders pass, please create PR to reenable the tests. Thanks
@KapilPowar if the Grinders pass, please create PR to reenable the tests. Thanks
PR created https://github.com/eclipse-openj9/openj9/pull/18181
@llxia
Grinder test run for TestFlushReflectionCache & jit_jitt_openj9_none_SCC
x86-64_linux
x86-64_mac
x86-64_windows
PR created #18206
@llxia
Grinder test run for cmdLineTester_dumpromclasstests
aix_ppc-64 JDK8
aix_ppc-64 JDK11
aix_ppc-64 JDK17
PR created #18215
@llxia Do we still need this issue?
I believe this issue can be closed.