Annotation processor not always detected when using <type>processor</type>

[desruisseaux]

In Maven 4, the recommended way to add a JAR on the annotation processor classpath is like below:

<project>
  <dependencies>
    <dependency>
      <groupId>org.apache.maven.plugins.compiler.it</groupId>
      <artifactId>annotation-processor</artifactId>
      <version>1.0-SNAPSHOT</version>
      <type>processor</type>
    </dependency>
  </dependencies>
</project>

However, Maven Compiler Plugin 4.0.0-beta-3 does not automatically detects the processor if it is declared only as above. This commit fixes that issue and adds an integration test, together with opportunistic cleaning in other integration tests.

In the main time, as a workaround, adding the following configuration seems to work, where org.foo.MyAnnotationProcessor is the value declared in the META-INF/services/javax.annotation.processing.Processor file. Note that only the processor class is declared, not the artifact coordinates:

<project>
  <build>
    <plugins>
      <plugin>
        <artifactId>maven-compiler-plugin</artifactId>
        <configuration>
          <annotationProcessors>
            <annotationProcessor>org.foo.MyAnnotationProcessor</annotationProcessor>
          </annotationProcessors>
        </configuration>
      </plugin>
    </plugins>
  </build>
</project>

[bmarwell]

I think no one will care in the future about this. This is appropriate for this PR / issue to mention once, but for future use it is probably uninteresting.

[desruisseaux]

The purpose of this file is to document the differences between the 3.x and 4.0 branches. The compiler-one-output-file-test2 directory is still present in the 3.x branch. Because some tests have been added in the 3.x branch without being ported yet to the 4.0 branch (e.g., MCOMPILER-538), in the future we may want to compare the two branches for checking which tests present in one branch are missing in the other branch. This note is for remembering to not port that directory.

[bmarwell]

OK, I thought that should be obvious when browsing through the history, as the history would now have a reference to an issue. But now that you mention it, your commits do not have an issue linked, it seems.

[desruisseaux]

Actually, I'm lazily mixing two works in this same pull request. The work that really solves the issue described in this issue is the "Fix: annotation processor not always detected when specified with processor" commit. But since this commit adds a new integration test, this pull request opportunistically does some cleanup in those tests first. Those cleanups are in separated commits, with the intend to not squash them. The README.md file is added as part of this cleanup, but it describes changes that were actually done long ago.

[Bukama]

Actually, I'm lazily mixing two works in this same pull request.

then why all those renames :D :D :D 💋

[desruisseaux]

It is because some integration tests were using the install goal. I replaced some install by verify in this commit but a few remain if I remember right. The use of install goal in integration tests pollutes the ~/.m2/repository/ directory. By using the org.apache.maven.plugins.compiler.it groupId for all tests, we ensure that if some install goals are used again for whatever reason, at least we have a single place to cleanup in the .m2/repository directory.

[Bukama]

Actually, I'm lazily mixing two works in this same pull request.

then why all those renames :D :D :D 💋