Make it easier to use a different JVM target #297
Conversation
Whilst Jenkins defaults will now be java11 some libraries may want to stay on java8. Doing so makes backports of any issues much easier (hopefully not needed!) and the library can take advantage of all the updated plugins with their fixes. Whilst this is less than optimal for java8 as it still required a java9+ JVM to be used to compile (as that was when the `release` flag was introduced it is much better than either pinning the parent to 1.76 (the last supporting java8), or having to overwrite all the properties to remove the maven.compiler.release and testRelease ones as well as overwriting the enforcer configurations (albeit this is simplier with combine.children)
There was a problem hiding this comment.
On the one hand, this is relatively simple and there have already been a few components that updated this POM only to laboriously override Java levels.
On the other hand, I tend to prefer to keep things simple and just drop Java 8 support from new versions of components. If and when a backport is needed for some emergency bug fix—and 2.346.x is even still under support—it is very simple to create a backport branch and publish fixes from it. In fact I would strongly advise using a backport branch anyway, since trunk can accumulate all sorts of incidental changes unrelated to your emergency which might cause regressions.
In the case motivating this PR, jenkinsci/trilead-api-plugin#55 seems more straightforward. There is no particular reason to offer new versions of the plugin for older Jenkins versions.
|
Amends #209 for the record. (Would have mentioned that in the review comment, but alas GitHub neglects to create a backlink in that case—a bug I thought I reported long ago.) |
Co-authored-by: Jesse Glick <[email protected]>
There was a problem hiding this comment.
I concur with #297 (review). This implicitly encourages consumers to stay on Java 8 by adding first-class API support for it in the form of the javaTarget Jenkins-specific property. Once such an API is exposed, consumers will start expecting generic support for arbitrary Java versions from the parent POM, which is not something we are committing to providing. Seems safer not to expose any such commitment at the API level and to implicitly discourage consumers from staying on Java 8 by making them carry their own workarounds (or a backport branch).
|
Could we enforce a minimum Jenkins dependency of 2.357 for Java 11 targets in the plugin POM to prevent a situation in which a plugin requires Java 11, yet claims compatibility with Jenkins <2.357 which are supposed to still support Java 8? |
portably the wrong place for that discussion but https://github.com/jenkinsci/plugin-pom/blob/master/pom.xml#L1001 is likely the reason this went undetected. |
|
Yes, we went with jenkinsci/plugin-pom#563 currently. Something along the lines of jenkinsci/plugin-pom#478 would be a longer-term change. Probably you can arrange to continue to have |
|
Right, there's a comment explaining it in detail just above, thanks for the reference. Too bad this isn't possible 😦 |
|
I filed #331 with an alternative. |
Whilst Jenkins defaults will now be java11 some libraries may want to
stay on java8.
Doing so makes backports of any issues much easier (hopefully not
needed!) and the library can take advantage of all the updated plugins
with their fixes.
Whilst this is less than optimal for java8 as it still required a java9+
JVM to be used to compile (as that was when the
releaseflag wasintroduced) it is much better than either
overwriting the enforcer configurations (albeit this is simplier with
combine.children)