HIVE-29027: Move to JDK-21

[ayushtkn]

Tasks

• Hive Dev Box to support JDK-21 --> Built custom image with jdk-21 support & pushed to DockerHub
• DockerFiles to use JDK-21 -> Bump JDK version in docker images
• System.setSecurityManager removal in JDK-21 --> Replaced System.exit() & setSecurityManager with ExitUtil (Hadoop Solution) Moves to HIVE-29056
• Removal of Thread.stop() -> Thread.stop is removed in JDK-21 --> Replaced with interrupt & will thread handle (Replaced in a deprecated test only method otherwise interrupt ain't a direct replacement of stop)
• Removal of Reflections hack -> JDK-21 is more strict it doesn't let hack the Final fields -> test class had field Final removed that final from test class & dropped the hack
• JWT tests modifying System env --> Not supported in JDK-21, tried some hacks to prevent not dropping those tests, may not be exact copy but better than dropping
• Decimal/Float precession change in JDK-21 --> JDK-21 had changes around the precession, leads to query output regen, haven't found any hack to use the old behaviour, as of now can't help it but adapt. https://inside.java/2022/09/23/quality-heads-up/
  https://bugs.openjdk.org/browse/JDK-4511638
• Kudu tests failing -> Tests failing since Kudu uses internal java classes removed in JDK-21 raised KUDU-3670. --> Replaced the util which runs the Kudu Cluster with directly running the Kudu cluster

[aturoczy]

Instant approve :)

[ayushtkn]

Down to some 24 Failures, of which 2 flaky (they passed in previous build). 22 are all from Kudu Handler. Have created a ticket KUDU-3670.

[deniskuzZ]

Trino is already on JDK-24, doesn't even compile with lower versions

[ayushtkn]

Green Build with JDK-21 :-)

[deniskuzZ]

134 files changed 🥇

[simhadri-g]

Awesome, will this be in hive v5.0 ? :)

[deniskuzZ]

@ayushtkn, did you grep for

    <maven.compiler.source>8</maven.compiler.source>
    <maven.compiler.target>8</maven.compiler.target>

see https://issues.apache.org/jira/browse/HIVE-29018

[ayushtkn]

If so, how about creating a separate ticket for it and marking it as a blocking dependency for this one? It already has standalone value, and splitting it out would make this PR much clearer.

Thanx @abstractdog for the review. The ExitUtils change is now separated

[abstractdog]

PR looks good to me—I'm open to giving a +1.

The only potentially debatable part is the precision change, which results in many q.out diffs. However, looking at the underlying Java change:
🔗 https://github.com/openjdk/jdk/pull/3402/files
which was eventually committed here:
🔗 openjdk/jdk@72bcf2a

The fact that even Java’s golden files changed is pretty telling—so it is what it is.
For the Hive release, we can note that this is due to JDK-4511638.

assuming that @ayushtkn will consider my comment regarding systemlambda vs. system stubs, LGTM

[abstractdog]

LGTM +1
Would you be so kind, @ayushtkn, to wait another 24 hours in case anyone has objections?

[sonarqubecloud]

Quality Gate passed

Issues
2 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud