Add Java 15 CI #21713

dreis2211 · 2020-06-05T06:34:04Z

Hi,

this PR tackles #21604 as discussed. Fortunately, even the automatic JDK upgrade should already work.

Before merging this, #21605 needs to be merged. Apart from that I haven't seen any test failures when I run with the custom buildJavaHome property being set to my local JDK 15 installation.

Cheers,
Christoph

Closes spring-projectsgh-21604

snicoll · 2020-06-05T10:14:04Z

Thanks for the PR @dreis2211

Apart from that I haven't seen any test failures when I run with the custom buildJavaHome property being set to my local JDK 15 installation.

This was required when Gradle does not support the target JDK which does not seem to be case with Java15 at this point. I don't see this PR using that feature either so I am a bit confused.

Switching my local JDK to Java 15ea and building master leads to some failures. Do you have time to look at those?

dreis2211 · 2020-06-05T10:17:56Z

@snicoll You're absolutely right. I was under the impression that I'm getting the same results with the custom home property being set.
Let me try with your approach.

dreis2211 · 2020-06-05T11:03:17Z

@snicoll It would have also worked with the custom java home being set, but my tasks were cached and thus not all executed.

I see one additional failure in SampleAntApplicationIT.runJar + the already mentioned multi-release JAR test that is covered by #21605 .

Does that match your local build?

snicoll · 2020-06-05T11:07:02Z

No, it does not. I have a compilation failure:

> Task :spring-boot-project:spring-boot-tools:spring-boot-loader:compileJava FAILED
/Users/snicoll/workspace/work/spring-boot/master/spring-boot-project/spring-boot-tools/spring-boot-loader/src/main/java/org/springframework/boot/loader/jar/StringSequence.java:75: error: isEmpty() in StringSequence cannot implement isEmpty() in CharSequence
        boolean isEmpty() {

and one test failure before I gave up:

MappingsEndpointReactiveDocumentationTests > mappings() FAILED
    org.springframework.restdocs.payload.FieldTypesDoNotMatchException at MappingsEndpointReactiveDocumentationTests.java:126

This in a branch of your first PR I am about to merge. Maybe something is cached and shouldn't be?

$ java -version
openjdk version "15-ea" 2020-09-15
OpenJDK Runtime Environment (build 15-ea+26-1287)
OpenJDK 64-Bit Server VM (build 15-ea+26-1287, mixed mode, sharing)

dreis2211 · 2020-06-05T11:10:22Z

openjdk version "15-ea" 2020-09-15
OpenJDK Runtime Environment (build 15-ea+17-717)
OpenJDK 64-Bit Server VM (build 15-ea+17-717, mixed mode, sharing)

I guess the first one can be explained by CharSequence.isEmpty being only available in later builds. I'll try the latest JDK 15 build then.

In JDK 15 the concept of hidden classes was introduced, which also affects Lambdas in so far that Class.getCanonicalName() will return null for those. Using Class.getName() works around that behavior change. See spring-projectsgh-21604

JDK 15 introduces isEmpty() on CharSequence which clashes with the one declared in StringSequence because it is not public. See spring-projectsgh-21604

dreis2211 · 2020-06-05T15:29:05Z

I've pushed two additional commits that fix the syntax error and the failure in MappingsEndpointReactiveDocumentationTests

dreis2211 · 2020-06-05T16:25:42Z

I see some intermittent failures for tests that all seem to be related to timing. I wonder if https://bugs.openjdk.java.net/browse/JDK-8242504 is related as this got introduced recently as well.
EDIT: After checking the code, I don't think it is related as it should not affect my Mac.

dreis2211 · 2020-06-06T13:04:17Z

I've also seen some flaky tests that are annotated with @WebEndpointTest that prematurely close the connection.

dreis2211 · 2020-06-10T18:03:07Z

I just tried with JDK 14 and see similar flaky failures there as well. Particularly, the "performance" tests ConfigurationPropertySourcesTests seem to fail on JDK 14 on my machine, too. So I wonder if we might already be in the position to consider JDK 15 as done and "simply" have to improve flaky tests across the board.

dreis2211 · 2020-06-11T06:07:21Z

DevToolsIntegrationTests > createAControllerAndThenAddARequestMapping seems to also fail on JDK 8 occasionally already.

snicoll

I think we're almost done here but there is one change I'd like us to reconsider. I've added a comment

snicoll · 2020-06-11T06:49:34Z

.../java/org/springframework/boot/actuate/web/mappings/reactive/HandlerFunctionDescription.java

@@ -29,7 +29,7 @@
 	private final String className;

 	HandlerFunctionDescription(HandlerFunction<?> handlerFunction) {
-		this.className = handlerFunction.getClass().getCanonicalName();
+		this.className = handlerFunction.getClass().getName();


I don't think we should do this. Perhaps as a fallback if getCanonicalName is empty? It's a change of format so it's a breaking change.

Or perhaps we could adapt the test? I haven't looked in details yet so thinking out loud.

I wrote to the core-libs mailing list a week ago, I think, and got this reply.

A hidden class has no canonical name [1] and it can't be nominally referenced [...] For diagnosiability, Class::getName or Class::toString can be used.

As HandlerFunction is often a lambda, it will probably often return null on JDK 15 at least. I don't have a problem with the fallback idea. In reality, there will be only a change for inner classes though. See the following example output from JDK 14 of canonical vs name:

com.example.demo.Example$$Lambda$14/0x0000000800b66c40 com.example.demo.Example$$Lambda$14/0x0000000800b66c40 com.example.demo.AFunction com.example.demo.AFunction com.example.demo.Example.InnerFunction com.example.demo.Example$InnerFunction

Give me a minute to implement the fallback.

See spring-projectsgh-21604

dreis2211 · 2020-06-15T14:47:35Z

Trying with 15-ea-27 introduced a change for missing Javadocs: https://bugs.openjdk.java.net/browse/JDK-8242607

That produces quite a lot of warnings like the following:

/Users/christoph.dreis/projects/spring-boot/spring-boot-project/spring-boot-tools/spring-boot-buildpack-platform/src/main/java/org/springframework/boot/buildpack/platform/docker/transport/DockerEngineException.java:35: warning: no comment
        private final String reasonPhrase;

It's not breaking the builds, so I guess it's fine for this PR, but probably needs to be tackled in another one.

snicoll · 2020-06-15T14:58:48Z

Ah ah. I was building locally and I've already polished that checkstyle issue here Christoph.

dreis2211 · 2020-06-15T15:00:08Z

Sorry - I've noticed this one just now. :/

In JDK 15 the concept of hidden classes was introduced, which also affects Lambdas in so far that Class.getCanonicalName() will return null for those. This commit uses Class.getName() as a fallback when no canonical name is available. See gh-21713

JDK 15 introduces isEmpty() on CharSequence which clashes with the one declared in StringSequence because it is not public. See gh-21713

See gh-21713

Add Java 15 CI

2af8f78

Closes spring-projectsgh-21604

spring-projects-issues added the status: waiting-for-triage An issue we've not yet triaged label Jun 5, 2020

snicoll self-assigned this Jun 5, 2020

snicoll added type: task A general task and removed status: waiting-for-triage An issue we've not yet triaged labels Jun 5, 2020

snicoll added this to the 2.4.0-M1 milestone Jun 5, 2020

snicoll added the status: waiting-for-feedback We need additional information before we can continue label Jun 5, 2020

spring-projects-issues added status: feedback-provided Feedback has been provided and removed status: waiting-for-feedback We need additional information before we can continue labels Jun 5, 2020

snicoll added status: waiting-for-feedback We need additional information before we can continue and removed status: feedback-provided Feedback has been provided labels Jun 5, 2020

spring-projects-issues added status: feedback-provided Feedback has been provided and removed status: waiting-for-feedback We need additional information before we can continue labels Jun 5, 2020

snicoll added status: waiting-for-feedback We need additional information before we can continue and removed status: feedback-provided Feedback has been provided labels Jun 5, 2020

dreis2211 added 2 commits June 5, 2020 17:20

Use Class.getName() in HandlerFunctionDescription

69b197b

In JDK 15 the concept of hidden classes was introduced, which also affects Lambdas in so far that Class.getCanonicalName() will return null for those. Using Class.getName() works around that behavior change. See spring-projectsgh-21604

Allow StringSequence.isEmpty() to be compatible with JDK 15

d21e15f

JDK 15 introduces isEmpty() on CharSequence which clashes with the one declared in StringSequence because it is not public. See spring-projectsgh-21604

spring-projects-issues added status: feedback-provided Feedback has been provided and removed status: waiting-for-feedback We need additional information before we can continue labels Jun 5, 2020

snicoll reviewed Jun 11, 2020

View reviewed changes

Only fallback to Class.getName() in HandlerFunctionDescription

603a1b8

See spring-projectsgh-21604

Fix checkstyle issue

f6a9f7c

snicoll removed the status: feedback-provided Feedback has been provided label Jun 15, 2020

snicoll pushed a commit that referenced this pull request Jun 15, 2020

Allow StringSequence.isEmpty() to be compatible with JDK 15

3d27391

JDK 15 introduces isEmpty() on CharSequence which clashes with the one declared in StringSequence because it is not public. See gh-21713

snicoll pushed a commit that referenced this pull request Jun 15, 2020

Add Java 15 CI

85218db

See gh-21713

snicoll added a commit that referenced this pull request Jun 15, 2020

Polish "Add Java 15 CI"

e87c319

See gh-21713

snicoll closed this in 75dd965 Jun 15, 2020

Add Java 15 CI #21713

Add Java 15 CI #21713

Uh oh!

Conversation

dreis2211 commented Jun 5, 2020

Uh oh!

snicoll commented Jun 5, 2020

Uh oh!

dreis2211 commented Jun 5, 2020

Uh oh!

dreis2211 commented Jun 5, 2020

Uh oh!

snicoll commented Jun 5, 2020

Uh oh!

dreis2211 commented Jun 5, 2020

Uh oh!

dreis2211 commented Jun 5, 2020

Uh oh!

dreis2211 commented Jun 5, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dreis2211 commented Jun 6, 2020

Uh oh!

dreis2211 commented Jun 10, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dreis2211 commented Jun 11, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

snicoll left a comment

Choose a reason for hiding this comment

Uh oh!

snicoll Jun 11, 2020

Choose a reason for hiding this comment

Uh oh!

dreis2211 Jun 11, 2020

Choose a reason for hiding this comment

Uh oh!

dreis2211 Jun 11, 2020

Choose a reason for hiding this comment

Uh oh!

dreis2211 commented Jun 15, 2020

Uh oh!

snicoll commented Jun 15, 2020

Uh oh!

dreis2211 commented Jun 15, 2020

Uh oh!

Uh oh!

dreis2211 commented Jun 5, 2020 •

edited

Loading

dreis2211 commented Jun 10, 2020 •

edited

Loading

dreis2211 commented Jun 11, 2020 •

edited

Loading