condy arrayindexoutofboundsexception fix

[vimil]

this is to fix arrayindexoutofbounds issue when running jmockit in eclipse 2020.03 with code coverage

[jbennett2091]

If a (real-world) test is required, please have a look at the example provided with #688

[jbennett2091]

On a whim, I pulled this branch, built it locally, and used the resulting artifact in my project. It did not work for me, but that could very well be because I did something wrong.

Here's the build error (in my code) that I get when using JDK 11.0.8 + Jacoco 0.8.6 + JMockit 1.49-mod:

[ERROR] Failed to execute goal org.apache.maven.plugins:maven-compiler-plugin:3.8.1:testCompile (default-testCompile) on p
roject omni-core: Compilation failure: Compilation failure:
[ERROR] /C:/Users/jb14684/git/omni-project-alt4/omni/omni-core-project/omni-core/src/test/java/com/ibi/omni/util/CoreUtilsTest.java:[46,38] method newInstanceUsingCompatibleConstructor in class mockit.internal.reflection.ConstructorReflection cannot be applied to given types;
[ERROR]   required: java.lang.Class<?>,java.lang.String
[ERROR]   found: java.lang.Class<com.ibi.omni.util.CoreUtils>
[ERROR]   reason: actual and formal argument lists differ in length

Here's the test code being compiled:

	@Test
	public void testCtor()
	{
		ConstructorReflection.newInstanceUsingCompatibleConstructor(CoreUtils.class);
	}

Here's the class under test:

	private CoreUtils()
	{
		// static use only
	}

These combos work:

• JDK8 + Jacoco 0.8.6 + JMockit 1.49 (i.e. use JDK8)
• JDK11 + Jacoco 0.8.3 + JMockit 1.49 (i.e. use old-Jacoco)

[vimil]

On a whim, I pulled this branch, built it locally, and used the resulting artifact in my project. It did not work for me, but that could very well be because I did something wrong.

Here's the build error (in my code) that I get when using JDK 11.0.8 + Jacoco 0.8.6 + JMockit 1.49-mod:

[ERROR] Failed to execute goal org.apache.maven.plugins:maven-compiler-plugin:3.8.1:testCompile (default-testCompile) on p
roject omni-core: Compilation failure: Compilation failure:
[ERROR] /C:/Users/jb14684/git/omni-project-alt4/omni/omni-core-project/omni-core/src/test/java/com/ibi/omni/util/CoreUtilsTest.java:[46,38] method newInstanceUsingCompatibleConstructor in class mockit.internal.reflection.ConstructorReflection cannot be applied to given types;
[ERROR]   required: java.lang.Class<?>,java.lang.String
[ERROR]   found: java.lang.Class<com.ibi.omni.util.CoreUtils>
[ERROR]   reason: actual and formal argument lists differ in length

Here's the test code being compiled:

	@Test
	public void testCtor()
	{
		ConstructorReflection.newInstanceUsingCompatibleConstructor(CoreUtils.class);
	}

Here's the class under test:

	private CoreUtils()
	{
		// static use only
	}

These combos work:

• JDK8 + Jacoco 0.8.6 + JMockit 1.49 (i.e. use JDK8)
• JDK11 + Jacoco 0.8.3 + JMockit 1.49 (i.e. use old-Jacoco)

ConstructorReflection.newInstanceUsingCompatibleConstructor(CoreUtils.class);

the method ConstructorReflection.newInstanceUsingCompatibleConstructor takes two parameters. A class and a String.

[jbennett2091]

OK, mea culpa, and that's what I get for using a class in the 'internal' package. You are correct, and switching to the 'newInstanceUsingDefaultConstructor' method (rather than newInstanceUsingCompatibleConstructor) solved that unrelated issue. In my defense, "it used to work" 👍

---

My new answer, I forked this branch, and tried it in my own application (affected by the aforementioned issue). This code works for me. Take that for what it's worth.

I'd love to see this pull request get released. Even if it is just a 'milestone' or 'beta', it resolves blocking issues for a number of folks, including myself. JMockit, as a library, seems to have gone from releases every 3-months, to no release in the last 9-months, and this fix languishing (without real comment) for 6-months. I don't mean to sound critical, I am a big proponent of this framework. Just hoping to see it move forward.

[Saljack]

I can confirm that this MR fix our problem with JMockit and Jacoco too. So with this MR I can use a newer version of Jacoco (I tested it with 0.8.6). I hope that it will be merged soon. I start to be little nervous because I love this project.

[rliesenfeld]

Next year, once I am no longer working at home; no time for JMockit now.

[reinhapa]

@rliesenfeld looking forward for this fix to land in the next release

[nissane]

@rliesenfeld Can this get merged in? Love this project, but don't see much work on it recently. Is JMockit going to be maintained? I really hope so!

[legacycode]

@rliesenfeld This fix works for me. Can you please merge it and create a new release for mvnrepository.com?

[dellgreen]

Works for my projects also. Recently did an investigation at work for our internal projects. The graph below may or may not be useful to others:

gradle	java	jmockit	jacoco	result	notes	recommend
7.2	11	1.49	0.8.3	pass
7.3	11	1.49	0.8.3	pass		*
7.2	11	1.49a	0.8.3	pass
7.2	11	1.49a	0.8.6	pass
7.2	11	1.49a	0.8.7	pass
7.3	11	1.49a	0.8.7	pass		*
7.2	14	1.49a	0.8.3	fail	jacoco agent exceptions
7.2	14	1.49a	0.8.6	pass
7.2	14	1.49a	0.8.7	pass
7.3	14	1.49a	0.8.7	pass		*
7.2	16	1.49a	0.8.3	fail	jacoco agent exceptions
7.2	16	1.49a	0.8.6	pass
7.2	16	1.49a	0.8.7	pass
7.3	16	1.49a	0.8.7	pass		*
7.2	17	1.49a	0.8.3	fail	jacoco agent exceptions
7.2	17	1.49a	0.8.6	fail	jacoco agent exceptions
7.2	17	1.49a	0.8.7	pass
7.3	17	1.49a	0.8.7	pass		*
7.2	18ea	1.49a	0.8.7	fail	gradle/groovy exceptions
7.3	18ea	1.49a	0.8.7	fail	jacoco agent exceptions (will be fixed in 0.8.8)
7.3	18ea	1.49a	0.8.8	TBC

[DhavalUpadhyayaTR]

Any idea when it is going to be merged and a new release be available?

[sumit1sen]

Does anyone know when this will be merged and released? I'd prefer to use an official release instead of creating my own private jmockit jar.

[rliesenfeld]

Hopefully I will have some time for this, in 2022. That said, I see it as really the fault of JaCoCo. The change they made that generates the Condy bytecode, was supposedly a performance optimization. However, no benchmarking was ever done (that I know of). IMO, the JaCoCo team should reconsider if generating such unusual bytecode is a good idea.

[altair2010]

Work for me too after a JDK bump

JDK: 16 Gradle : 7.4.1

[abtabhi]

works for me too, because of Jacoco to SonarQube integration its becoming hard to skip any of these components :(

[karnick]

It is surprising, why no more releases after Dec 2019 (v1.49)?

[peterkempy]

@rliesenfeld We all understand that you have competing priorities, and we appreciate the time you have devoted to JMockit.

As regards this fix from @vimil , if this PR were merged then it would unblock a lot of devs - many more than are listed above.

(Plus maybe https://github.com/jmockit/jmockit1/pull/736 if required too ?)

[weifang6601]

Hope this can get fixed, I use jmockit for many years and I think it is the best testing tool for java.

[zbflcy]

really hope this can get fixed,

[raytapan]

Is it possible to merge this PR and make a new release ? I have been blocked for a while now.