POM bug: separate jvm.arg.allowSecurityManager from jvm.arg.addOpens
Before, this was in a single variable, and for JDK 18+, the security
manager setting also falsely overwrote the '--add-opens' command. This
was the root cause for a few dozen LTW tests to fail on JDK 18, if they
were not run in full LTW mode, i.e. in a separate JVM.
After this fix, it should be possible to revert the corresponding
commits, at least their non-cosmetic parts.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Add some smoke test cases for '--add-reads' and '--add-modules'
in order to show that the compiler options basically work. The test
cases are by no means exhaustive and simply verify that it works at all.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Add test case + experimental fix for AJC option '--add-exports'
I am expecting the test case to pass, but other tests to fail. This
temporary commit is meant to create feedback from GitHub CI test runs.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Name annotation vs native style aspects uniformly in github_128 tests
Trying to find the difference between byte code generated by AJC for
functionally identical annotation vs native style aspect, I move the
aspects into subdirectories in order to be able to name them
identically. This way, when decompiling them with javap or Fernflower it
is easier to diff them later.
Why the decompilation? Because for the thread pool testing scenario
native syntax passes while annotation syntax fails. I.e., we need to
find the difference. That can be done by reading source code, if you
know where to look, or by starting with reverse engineering in order to
first understand more and look at the code later.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Expand "asynchronous proceed for nested around-advice" to 4 scenarios:
1. @AspectJ syntax, threads created inside advice code
2. @AspectJ syntax, submit runnable to thread pool
3. native syntax, threads created inside advice code
4. native syntax, submit runnable to thread pool
Scenarios 1, 3 and 4 are passing, while 2 is failing.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
AtAjLTWTests: 4 test cases switched to 'useFullLTW' mode
After the Java 18 JDT Core upgrade, some LTW tests are failing and
have to be switched from in-process to full LTW mode due to them now
obviously calling some code paths which need '--add-opens'.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Regression test for #125: Bugs198Tests.testGitHub_125
This test fails when run against AspectJ 1.9.8 with JDT Core 1.9.8.RC3.
It passes when using the latest JDT Core 1.9.9-SNAPSHOT. It sets system
property 'org.aspectj.weaver.openarchives=20', provoking open classpath
JAR file exhaustion when compiling a simple class with AJC, i.e. JARs
are being forcibly closed and automatically re-opened, as soon as they
are needed. Before the JDT Core bugfix, this test causes:
java.lang.NullPointerException
at ....compiler.batch.ClasspathJmod.getModulesDeclaringPackage
With the bugfix incorporated into AspectJ Tools, the problem is gone.
Note: New test dependency 'io.github.bmuskalla:scoped-system-properties'
helps to test compilation with the temporarily changed global system
property in isolation, saving the environment in a thread-local
variable and later cleanly restoring the original values again. If we
ever switch to parallel test execution, this would otherwise influence
other tests and potentially cause weird side effects. Better safe than
sorry.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Move some tests from Ajc198TestsJava to Bugs198Tests
They were accidentally stored in thw wrong test class, not just because
they are bug regression tests, not 1.9.8 features, but also because they
are meant to work on Java 5+, not on Java 17+ like the ones in
Ajc198TestsJava.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
This fixes:
- negating annotation style if() pointcuts doesn't work
- annotation style if() pointcut not able to use a binding
that is not exposed
Fixes #120,#122
Fix annotation style support for if(true) and if(false)
The documentation specifies annotation style pointcuts
can use if(false) or if(true) and not require a boolean
return value and body for the @Pointcut annotated
method but it doesn't work without this change to validation
that recognizes the situation.
Fixes #115
Trailing whitespaces are useless. Most of code-styles forbids them. Most of editors always trim them on save.
I propose to clean up project from trailing whitespaces in all java files at once.
The test worked on my local workstation with German locale, but not on
GitHub with English locale.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Update ECJ version, activate Java 17 preview features tests
After JDT Core (ECJ) was updated to the final Java 17 feature set, the
tests now pass as expected.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
- Fix one fault sanity test configuration
- Deactivate Java 16 preview tests (no longer supported by Java 17
compiler)
- Test sealed classes as final on Java 17 (no longer preview)
- Add tests for JEP 406, pattern matching for switch (preview). At
present, the beta 17 branch of JDT Core does not handle the tested
features and expected compile errors correctly yet, so I had to
temporarily deactivate test execution, only printing TODO messages.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Refactor AnnoBinding test class used by Ajc1612Tests some more
Improve improved general logging, error messages, variable naming and
indentation, making the code of class AnnoBinding a bit more readable.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
By using System.nanoTime() instead of currentTimeMillis(), the flakiness
even with the original 10,000 rounds is significantly lower than before.
Making my IDE repeat the test until failure, it took on average 150 runs
to make it fail. So, the more accurate timing helps. With 100,000
rounds, it was even more stable, but eventually I could make it fail.
With 1,000,000 rounds however, even running the test 500x could not make
it fail. So for all practical purposes, I think the test is reasonably
stable now.
Closes #83.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
By increasing from 10,000 to 1,000,000 rounds, the times compared for
performance become considerably longer (but still in the tens or
hundreds or milliseconds), decreasing the probability of the test
failing due to CPU load or some other random effect.
Closes #83.
Signed-off-by: Alexander Kriegisch <Alexander@Kriegisch.name>
Alexander Kriegisch
Andy Clement
Andrey Turbanov