CASSJAVA-135 Update the build to use Java17 rather than Java8 by absurdfarce · Pull Request #2096 · apache/cassandra-java-driver

absurdfarce · 2026-08-17T16:46:58Z

Change the build to use Java17 for compiling the driver

absurdfarce · 2026-08-17T16:47:47Z

    <module>bom</module>
  </modules>
  <properties>
+    <maven.compiler.release>17</maven.compiler.release>


This change replaces the source + target designations used in earlier Maven builds

absurdfarce · 2026-08-17T16:48:25Z

        <artifactId>maven-compiler-plugin</artifactId>
        <configuration>
          <compilerId>javac-with-errorprone</compilerId>
-          <forceJavacCompilerUse>true</forceJavacCompilerUse>


Replaced by "forceLegacyJavaApi" use below

Actually this is no longer needed now that we've moved to the new mechanism for deploying errorprone with Java17.

absurdfarce · 2026-08-17T16:49:12Z

-            <compilerArg>-Xep:AnnotateFormatMethod:OFF</compilerArg>
-            <compilerArg>-Xep:WildcardImport:WARN</compilerArg>
-            <compilerArg>-XepExcludedPaths:.*/target/(?:generated-sources|generated-test-sources)/.*</compilerArg>
-          </compilerArgs>


The compiler args above were removed because they didn't seem to be supported on Java17 javac. I need to run back through these again to see if I missed something here.

absurdfarce · 2026-08-17T17:01:16Z

+          <forceLegacyJavacApi>true</forceLegacyJavacApi>
          <showWarnings>true</showWarnings>
-          <failOnWarning>true</failOnWarning>
+          <failOnWarning>false</failOnWarning>


I really don't like this change but I couldn't find another way around it.

Without this change the build fails as follows:

[INFO] --- compiler:3.15.0:compile (default-compile) @ java-driver-core --- [INFO] Compiling 799 source files with javac-with-errorprone [debug release 17] to target/classes [INFO] ------------------------------------------------------------- [WARNING] COMPILATION WARNING : [INFO] ------------------------------------------------------------- [WARNING] Supported source version 'RELEASE_8' from annotation processor 'org.apache.tinkerpop.gremlin.process.traversal.dsl.GremlinDslProcessor' less than -source '17' [INFO] 1 warning [INFO] ------------------------------------------------------------- [INFO] ------------------------------------------------------------- [ERROR] COMPILATION ERROR : [INFO] ------------------------------------------------------------- [ERROR] error: warnings found and -Werror specified [INFO] 1 error

Problem here stems from the GremlinDsl usage in the graph test code. The annotation processor for the version of Tinkerpop we use states that it produces Java8 code and apparently Maven considers this a warning.

Thing is that upgrading Tinkerpop is a fairly involved process... and perhaps more importantly newer versions of the annotation processor aren't any different. I'm not opposed to a Tinkerpop upgrade in general but I don't think it'll help much for this specific issue.

I've subsequently worked around the original issue by removing the annotation and bringing in static code (more on that elsewhere) but I'm inclined to keep "failOnWarning" set to false. With this setting we still try to run through the entire test suite meaning a single run can give us more useful info about the full range of test failures for the entire source at a given git commit. Perhaps as importantly it also prevents the build from cratering if a single "sketchy" test happens to act up during the current build.

absurdfarce · 2026-08-17T19:38:49Z

-              </additionalJOptions>
-              <useStandardDocletOptions>false</useStandardDocletOptions>
-            </configuration>
-          </execution>


For now I had to remove the api-plumber-doclet ref entirely.

Using the version that was in the build (version 1.0.0) gives the following errors now:

[INFO] ------------------------------------------------------------------------ [INFO] BUILD FAILURE [INFO] ------------------------------------------------------------------------ [INFO] Total time: 2.348 s [INFO] Finished at: 2026-08-17T12:36:17-05:00 [INFO] ------------------------------------------------------------------------ [ERROR] Failed to execute goal org.apache.maven.plugins:maven-javadoc-plugin:3.2.0:javadoc (check-api-leaks) on project java-driver-guava-shaded: An error has occurred in Javadoc report generation: [ERROR] Exit code: 1 - error: Class com.datastax.oss.doclet.ApiPlumber is not a valid doclet. [ERROR] Note: As of JDK 13, the com.sun.javadoc API is no longer supported. [ERROR] [ERROR] Command line was: /home/mersault/.jabba/jdk/openjdk@17.0.2/bin/javadoc -preventleak com.datastax.oss.driver.internal com.datastax.dse.driver.internal -preventleak com.datastax.oss.driver.shaded -preventleak com.typesafe.config -preventleak com.codahale.metrics -preventleak org.HdrHistogram -preventleak io.netty -preventleak jnr -preventleak com.kenai.constantine -preventleak com.kenai.jffi -preventleak com.kenai.jnr -preventleak net.jpountz -preventleak org.xerial.snappy @options @packages [ERROR] [ERROR] Refer to the generated Javadoc files in '/home/mersault/work/git/cassandra-java-driver-absurdfarce/guava-shaded/target/site/apidocs' dir. [ERROR] [ERROR] -> [Help 1]

Version 2.0.0 of the doclet was intended to address exactly this issue but it also seems to run aground:

[INFO] ------------------------------------------------------------------------ [INFO] BUILD FAILURE [INFO] ------------------------------------------------------------------------ [INFO] Total time: 2.342 s [INFO] Finished at: 2026-08-17T12:38:57-05:00 [INFO] ------------------------------------------------------------------------ [ERROR] Failed to execute goal org.apache.maven.plugins:maven-javadoc-plugin:3.2.0:javadoc (check-api-leaks) on project java-driver-guava-shaded: An error has occurred in Javadoc report generation: [ERROR] Exit code: 2 - error: No source files for package com.datastax.dse.driver.internal [ERROR] 1 error [ERROR] [ERROR] Command line was: /home/mersault/.jabba/jdk/openjdk@17.0.2/bin/javadoc -preventleak com.datastax.oss.driver.internal com.datastax.dse.driver.internal -preventleak com.datastax.oss.driver.shaded -preventleak com.typesafe.config -preventleak com.codahale.metrics -preventleak org.HdrHistogram -preventleak io.netty -preventleak jnr -preventleak com.kenai.constantine -preventleak com.kenai.jffi -preventleak com.kenai.jnr -preventleak net.jpountz -preventleak org.xerial.snappy @options @packages [ERROR] [ERROR] Refer to the generated Javadoc files in '/home/mersault/work/git/cassandra-java-driver-absurdfarce/guava-shaded/target/site/apidocs' dir. [ERROR] [ERROR] -> [Help 1]

I'm wondering if the underlying issue here isn't just that there's no real source in guava-shaded and if that's causing the plugin to barf.

For now I've simply disabled this doclet all together. This is another change I'd like to revisit (and ideally revert) at some point in the future.

…otation processors for integration tests to handle everything except the SocialTraversalDsl case.

…the build

… DSL

…wrong before

absurdfarce · 2026-08-21T06:11:22Z

+    <!-- for DriverBlockHoundIntegrationIT when using JDK 13+, see https://github.com/reactor/BlockHound/issues/33 -->
+    <blockhound.argline>-XX:+AllowRedefinitionToAddDeleteMethods</blockhound.argline>
+    <!-- allow deep reflection for mockito when using JDK 17+, see https://stackoverflow.com/questions/70993863/mockito-can-not-mock-random-in-java-17 -->
+    <mockitoopens.argline>--add-opens=java.base/jdk.internal.util.random=ALL-UNNAMED</mockitoopens.argline>


This was the configured setting for both Java17 and Java21 under the old regime of per-JVM settings. Folding these into the defaults since they're both common now... we can always break them out again if we need to (if Java25 needs something different, say).

absurdfarce · 2026-08-21T06:12:48Z

  environment {
    OS_VERSION = 'ubuntu/focal64/java-driver'
    JABBA_SHELL = '/usr/lib/jabba/jabba.sh'
+    BUILD_JAVA_VERSION = "openjdk@1.17"


A significant simplification of what we had here. We define a single Java version used to build (and drive testing of) the driver here and all invocations elsewhere reference this. We were doing something functionally equivalent by passing the Java version for building the driver in as a param... this just simplifies everything in one place.

absurdfarce · 2026-08-21T06:13:51Z

-    else
-      SKIP_JAVADOCS=false
-    fi
+    jabba use ${BUILD_JAVA_VERSION}


We'll run Maven with the fixed Java version even though individual tests will be run via a (perhaps different) Java version passed in via the "testJavaHome" param.

absurdfarce · 2026-08-21T06:14:41Z

    jabba which 1.8''', returnStdout: true).trim()
+  env.JAVA11_HOME = sh(label: 'Get JAVA11_HOME (in case ccm needs it)',script: '''#!/bin/bash -le
+    . ${JABBA_SHELL}
+    jabba which openjdk@11''', returnStdout: true).trim()


Adding Java11 for sake of completeness. I believe at some point we lose support for Java8 on some Cassandra servers so really we're just trying to get out in front of things here.

absurdfarce · 2026-08-21T06:16:26Z

 import org.apache.tinkerpop.gremlin.process.traversal.dsl.graph.GraphTraversal;
 import org.apache.tinkerpop.gremlin.structure.Vertex;

-@GremlinDsl(traversalSource = "com.datastax.dse.driver.api.core.graph.SocialTraversalSourceDsl")


No longer necessary now that we've brought in static Java source to represent the generated source code... see below for more on this.

absurdfarce · 2026-08-21T06:18:32Z

@@ -0,0 +1,55 @@
+/*


DefaultSocialTraversal, SocialTraversal and SocialTraversalSource were extracted from a local Java8 build which generated these sources via the older GremlinDsl annotation. That annotation explicitly supports only Java8 (and is only used here for testing) so rather than shift to some new test DSL that might be more friendly to Java17 it seemed easiest to just fix the generated source in the repo and move on.

absurdfarce · 2026-08-21T06:20:04Z

 import org.apache.tinkerpop.gremlin.process.traversal.Traversal;
 import org.apache.tinkerpop.gremlin.process.traversal.dsl.graph.GraphTraversal;
 import org.apache.tinkerpop.gremlin.process.traversal.dsl.graph.GraphTraversalSource;
+import org.apache.tinkerpop.gremlin.process.traversal.dsl.graph.__;


All the usages of the anonymous traversal class (the "double underscore") leverage base functionality provided by the Tinkerpop version and don't obviously require the customized GremlinDsl generated version. In order to minimize the amount of generated code we want to preserve in amber here it seemed worthwhile to just drop the anonymous traversal stuff all together.

absurdfarce · 2026-08-21T06:22:10Z

These changes got to a nearly green run on DataStax/IBM Jenkins. There were a few test failures for things like ccm timeouts and approximations of zero in a few places but they all look spurious to me (I'll very likely be creating tickets for one or two of them to avoid them returning).

absurdfarce · 2026-08-21T06:24:18Z

Ping @tolbertam for discovery... for some reason I couldn't add you as a reviewer.

absurdfarce added 2 commits August 14, 2026 16:05

Version bump to 5.0.0-SNAPSHOT

01f1795

Fixes to get base Java17 compilation working

11e47da

absurdfarce added the in progress label Aug 17, 2026

absurdfarce commented Aug 17, 2026

View reviewed changes

absurdfarce added 13 commits August 17, 2026 16:04

Fixed the issue with the API doclet, added back into the build

182a8bf

Jenkins fixes for DataStax build

495e68a

Removing now-unsupported LTS Java versions

df1ec90

Fix errorprone compilation to work with Java17. Also re-introduce ann…

7a438b6

…otation processors for integration tests to handle everything except the SocialTraversalDsl case.

Forgot to change the Jenkins config which, you know, uses Java17 for …

1c6f90d

…the build

Let's try using the right name

c4ce2b3

Bring in generated files for test cases that rely on the social graph…

0569441

… DSL

Update annotation processors for OSGi testing

517a576

Various Maven/Jenkins fixes

569d59b

A few more Jenkinsfile fixes

d696fcd

Still more Jenkins fixes

2e35151

Yet more Jenkinsfile fixes 'cause I love them so

b70bd2f

I think this is the last round of Jenkinsfile fixes... but I've been …

0cec826

…wrong before

absurdfarce changed the title ~~Compile the driver with Java17~~ CASSJAVA-135 Update the build to use Java17 rather than Java8 Aug 21, 2026

absurdfarce commented Aug 21, 2026

View reviewed changes

absurdfarce requested review from SiyaoIsHiding and aratno August 21, 2026 06:23

Uh oh!

Conversation

absurdfarce commented Aug 17, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

absurdfarce commented Aug 21, 2026

Uh oh!

absurdfarce commented Aug 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant