JarInputStream fails to get Manifest from published mockito-core

[hdecarne]

Using mockito-core in a JPMS based gradle build fails due to a known but odd behaviour of JarInputStream. The latter only provides access to the Jar's manifest if it is the 1st or 2nd entry in the Jar. As discussed in:

• https://bugs.openjdk.java.net/browse/JDK-8031748
• https://bugs.openjdk.java.net/browse/JDK-8211917

As gradle uses the JarInputStream to get the manifest and to determine whether the Automatic-Module-Name attribute is set, it fails to add mockito-core to the module path (see GRADLE #15270. Same issue was observed for Quarkus #5399. The attached example code example.zip shows this issue.

An optimal solution would be to make sure the manifest file is kept at the Jar's beginning during the publication process.

Used versions:

• mockito-core-3.6.0.jar
• java version "11.0.6" 2020-01-14 LTS
  Java(TM) SE Runtime Environment 18.9 (build 11.0.6+8-LTS)
  Java HotSpot(TM) 64-Bit Server VM 18.9 (build 11.0.6+8-LTS, mixed mode)

[TimvdLippe]

@mockitoguy might be able to provide more insight, but I think we create our Jar on this line:

mockito/gradle/java-publication.gradle

Line 32 in 78b22fc

javaLibrary(MavenPublication) { //name of the publication

We would welcome PRs for this issue. Hopefully the link provides enough context to figure out how to resolve the problem?

[joeshannon]

The commit message for fde5abb includes [ci maven-central-release] - should this cause a release to be published to maven central?
I can't see a new release there at the moment, not sure if there is usually a delay?

[TimvdLippe]

It should have, but it seems like it incorrectly concluded it didn't include the message. Filed #2110

[mockitoguy]

On it.