Junit5: Compile and test on Java 9

[sormuras]

Moved "Lessons learned..." to Wiki

Lessons learned "JUnit 5 compile and test on Java 9 b162"

Here, only the remaining technical steps resulting from the lessons learned are retained.

Related Issues

• #425 _"Test against JDK 9 modules"_
• #1951 _"Java 12"_

Deliverables

• [x] Upgrade to Gradle 3.5 - done in master with https://github.com/junit-team/junit5/commit/5ef538f7b7e953f89053de6986e254ee737cf918
• [x] Add information tag stating location from where an engine was loaded from - see #788
• [x] Merge main Java 8 and this Java 9 build into a single .travis.yml script -- via https://github.com/junit-team/junit5/commit/a313a075353a573d97b443ba57dd83db0f00f8fd
• [x] Include AssertJ 3.7.0 - remove jitpack from repositories
• [x] Include Gradle Shadow 2.0.0 - remove jfrog.org/artifactory from repositories
• [x] Report and remove workaround for ServiceLoader "bug" when "everybody" has access to build 163

[jbduncan]

I'm curious as to why build warnings should not be treated as errors.

How long should they not be treated as errors, if it's just a temporary measure?

[sormuras]

It's just a temporary measure to not break the build at https://travis-ci.org/junit-team/junit5/builds/219251561#L752 ... which should use AccessibleObject.canAccess as soon as we switch to Java 9.

[sbrannen]

@sormuras, thanks for compiling all of this information! 👍

[sbrannen]

FYI: the Shadow plugin has already been upgraded to ASM 5.2: https://github.com/johnrengelman/shadow/issues/285

[sbrannen]

Why not just suppress deprecation warnings on the call for now?

Or... would that mess up the JDK 8 build? 😜

[sbrannen]

It'll also be nice to be able to switch to trySetAccessible(), but we'll have to wait until JUnit 6 for that probably.

[jbduncan]

Alternatively to waiting for JUnit 6, we might be able to use MR-JAR support coming with Java 9, once we have a clear picture of how MR-JARs fit into Gradle. :)

[sbrannen]

Alternatively to waiting for JUnit 6, we might be able to use MR-JAR support coming with Java 9, once we have a clear picture of how MR-JARs fit into Gradle. :)

Ummmm... yeah... once we can actually (reliably) develop against Java 9 with the whole tool chain (Gradle, all CI servers, IDEs, etc.). So, I'm not going to hold my breath for that at the moment.

[jbduncan]

Of course.

[reinhapa]

I would be happy to help on that as soon you wanted to start on this issue...

[sormuras]

http://mailman.qos.ch/pipermail/announce/2017/000146.html

Ceki just released SLF4J version 1.8.0-alpha0, modularized for Java9/Jigsaw/JPMS/JSR 376, on Maven Central. Let's have look how he did.

[sormuras]

W/ daemon (default): https://travis-ci.org/junit-team/junit5/builds/220482687 works
W/o daemon (--no-daemon): https://travis-ci.org/junit-team/junit5/builds/220480116 fails

Why?

[sormuras]

All three open ends are still not resolvable:

• Travis CI (debian trusty) is still on JDK9-B162, see https://launchpad.net/~webupd8team/+archive/ubuntu/java?field.series_filter=trusty
• @johnrengelman did not release Gradle Shadow 2.0.0 yet, see https://github.com/johnrengelman/shadow/releases
• @joel-costigliola did not release AssertJ 3.7.0 yet, set https://github.com/joel-costigliola/assertj-core/releases

[joel-costigliola]

@sormuras AssertJ 3.7.0 should be released in a week or so.

[sormuras]

Thanks for the information, Joel! 👍

[sormuras]

AssertJ 3.7.0 now used in master via https://github.com/junit-team/junit5/commit/30581c788501729090f3490fe933013420d943fa

[marcphilipp]

@johnrengelman Are there any updates regarding Shadow 2.0.0?

[johnrengelman]

@marcphilipp I'm going to publish it tonight when I get home.

[sormuras]

Shadow 2.0.0 now used in master via https://github.com/junit-team/junit5/commit/febcfe98f5294fd146327ea64163ec4a6465702e