JarFile is not closed when finding main class from archive

[SebTardif]

What Problem This Solves

FindMainClass opens a JarFile when @classesRoot points at an archive and passes it to MainClassFinder.findSingleMainClass(JarFile, ...). That API does not close the jar; the caller owns the resource. The Ant task never closed it, leaving a file handle open for the rest of the JVM lifetime (or until GC finalization).

Almost every other new JarFile(...) call site in this repository already uses try-with-resources (Gradle/Maven plugins, loader tools, buildSrc checks, etc.).

Evidence

Red-green unit proof:

# Red (buggy code still present)
./gradlew :build-plugin:spring-boot-antlib:test --tests \
  org.springframework.boot.ant.FindMainClassTests.closesJarFileWhenSearchingArchive
FindMainClassTests > closesJarFileWhenSearchingArchive(Path) FAILED
    java.lang.AssertionError at FindMainClassTests.java:71
# assertion: getJarEntry after task.execute should throw IllegalStateException when jar is closed

# Green (try-with-resources applied)
BUILD SUCCESSFUL

Also verified: checkFormatMain/checkFormatTest, checkstyleMain/checkstyleTest.

Summary

Wrap the archive path in try-with-resources so the jar is closed after the main-class search returns.

Related

• Same leak class as prior cleanups: Close FileOutputStream delegate in InspectingOutputStream #50639 (InspectingOutputStream), Close URLClassLoader in ArchitectureCheck #50626 (URLClassLoader)
• Introduced on the archive branch of FindMainClass in 1595286e04 (2016-11-29; preferred @SpringBootApplication search). The unclosed new JarFile pattern has been present since the Ant task gained jar support.
• No open competing PR for this path.

Test plan

• Red: FindMainClassTests.closesJarFileWhenSearchingArchive fails without close
• Green: same test passes with try-with-resources
• Format + Checkstyle on main and test sources

[snicoll]

@SebTardif thanks again. I have a feeling you open a new PR every time one gets closed. This back and forth is quite time consuming on our end. Can't we gather stream cleanup in an issue rather than creating so many different PRs? I am asking if you have piled up more changes.

[SebTardif]

@snicoll No further stream/resource cleanup changes are lined up.

[snicoll]

@SebTardif did you mean to close this one?

[snicoll]

@SebTardif while I appreciate taking the time to write a test for this, I don't think it is worth it considering the change and the additional librairies the test brings. Thanks again for the PR!