withincode doesn't work as specified

[jafarre-bi]

I have been experiencing strange problems since I started using withincode together with execution pointcuts.
Trying to find a minimal reproducer, I found out that the simplest example, which appears in the documentation, is enough to show that something is wrong:

before(): withincode(* *(..)) && execution(* *(..)) {}
The advice is woven into absolutely no method.

If I remove the withincode pointcut, it works as expected:
before(): execution(* *(..)) {}
The advice is woven into every single method.

According to the documentation:
«execution(void m()) && withincode(void m()) is the same as execution(void m())»

Surprisingly enough, withincode(new(..)) && execution(new(..)) seems to work correctly, or I haven't detected any issue.

I'm testing with Aspectj versions 1.9.24 and 1.9.21 and both Eclipse Adoptium 21.0.7.6 and GraalVM community 24.0.1.

[jafarre-bi]

If not mistaken, the reason for the observed behaviour is in the static initialisation of the WithincodePointcut class.
In line 37, all the enclosing kinds are being filtered out from the matchedShadowKinds, including both method and constructor execution.
Right after, in line 42, the constructor execution bit is reintroduced to support inlining of field initialisers.
This perfectly reflects the observed behaviour that the withincode pointcuts match the constructor-execution join points, but not the method-execution join points.

While I think the current implementation should perform better, because fewer join points are selected, without loss of functionality, it doesn't comply with the documentation:
«The withincode pointcuts picks out each join point where the code executing is defined in the declaration of a particular method or constructor. This includes the method or constructor execution join point as well as any join points associated with the statements and expressions of the method or constructor. It also includes any join points that are associated with code in a method or constructor's local or anonymous types.»

I suggest fixing either the code or the documentation.

[kriegaex]

Thanks for raising this issue and for investigating about the possible root cause. I have not tested or reproduced it to validate it yet.

---

Please note my July 2024 message to the aspectj-users and aspectj-announce mailing lists regarding my situation as an AspectJ maintainer.

[jafarre-bi]

Added a draft pull request as a suggestion. Still untested.

[jafarre-bi]

Just to leave this issue updated, the PR #345 is ready for review and fixes this.