[main] Warn when @PluginBuilderAttribute field lacks public setter (#3195 port)

[vpelikh]

This ports the annotation processor validation from #3195 (2.x) to main (3.x).

What it does

The PluginProcessor annotation processor now validates that every @PluginBuilderAttribute field in a plugin builder class has a public setXxx() or withXxx() method that:

1. Takes exactly 1 parameter matching the field type
2. Returns a type assignable to the enclosing builder class
3. Is declared public

If no such setter is found, a compilation ERROR is emitted:

The field `myField` does not have a public setter. Note that
@SuppressWarnings("log4j.public.setter") can be used on the field to
suppress the compilation error.

Suppressing false positives

Use @SuppressWarnings("log4j.public.setter") on the field to suppress the warning for legitimate cases (e.g., fields set via constructor or reflection).

Files changed

• PluginProcessor.java — Added setter validation logic in processBuilderAttributeFields()
• PluginProcessorPublicSetterTest.java — Test with 3 cases: public setter OK, missing setter → error, suppressed → no error
• FakePluginPublicSetter.java.source — Test plugin for the above test
• SocketPerformancePreferences.java — Added @SuppressWarnings to 3 fields (setters return void, not builder type)
• StringMatchFilter.java — Added @SuppressWarnings (setMatchString name doesn't match field text)
• Rfc5424Layout.java — Added @SuppressWarnings (field enterpriseNumber has no setter)

Note

Unlike the 2.x version, the 3.x PluginProcessor uses @SupportedAnnotationTypes({"org.apache.logging.log4j.plugins.Plugin", "org.apache.logging.log4j.plugins.PluginBuilderAttribute", "org.apache.logging.log4j.core.config.plugins.PluginBuilderAttribute"}) and iterates by FQN, because 3.x has two @PluginBuilderAttribute variants.

[ramanathan1504]

@vpelikh

However, your new processor works so well that it actually caught a pre-existing violation in main! When building locally, the compile fails in log4j-jdbc because JdbcAppender is missing public setters:

[ERROR] .../log4j/jdbc/appender/JdbcAppender.java:[61,25] The field immediateFail does not have a public setter.
[ERROR] .../log4j/jdbc/appender/JdbcAppender.java:[81,22] The field reconnectIntervalMillis does not have a public setter.

Could you please update JdbcAppender.java in this PR (either by adding the missing setters or suppressing the warning) so the CI builds successfully? Once the build is green, I'll happily approve and merge!

[ramanathan1504]

@vpelikh, Also, I noticed Spotless formatting on 3 PR as well. When you push your updates, please make sure to run ./mvnw spotless:apply Follow the BUILDING.adoc

[vpelikh]

Thank you for the review, @ramanathan1504 ! Great catch – you're absolutely right, the new validation uncovered real issues in JdbcAppender.

I've just pushed two commits that:

1. Fix the setter return types – changed setImmediateFail() and setReconnectIntervalMillis() from void to B (the builder type), so they now satisfy the annotation processor's requirements.
2. Apply Spotless formatting – ran ./mvnw spotless:apply to fix all formatting violations across the affected files.

The build should be green now. Please take another look when you have a moment – happy to adjust anything else if needed.

[vpelikh]

@ramanathan1504 gentle ping, PR ready for re-review. Thanks!

[ramanathan1504]

@ramanathan1504 gentle ping, PR ready for re-review. Thanks!

Noted!. I will Try to close within a week. Thanks

[ramanathan1504]

@vpelikh both my earlier points are fixed. Nice catch on the withXxx guard too; 2.x only checks startsWith("set"), so its with branch is dead code.

One blocker: the new test leaves example/ and a PluginService file in the module directory. The @AfterEach deletes Log4j2Plugins.dat — that's the 2.x artifact name, it doesn't exist in 3.x — so it cleans up a file that's never created and misses the two that are. A @TempDir output path fixes it.

Also worth a look: adding PluginBuilderAttribute to @SupportedAnnotationTypes isn't needed — 2.x resolves it with getTypeElement() inside process(), and internal/Annotations.java already owns both FQNs. And the StringMatchFilter suppression collides with #4153, which renames that setter.

[vpelikh]

@ramanathan1504, thanks for review! Addresed your comments.