Uh oh!
There was an error while loading. Please reload this page.
[improvement](regression) Align regression framework arrow/netty with fe-core and target JDK17 - #64677
Merged
Merged
Conversation
… fe-core and target JDK17 Bump the regression-test framework's dependency versions to match fe-core (fe/pom.xml): arrow 17.0.0 -> 19.0.0 and netty-all 4.1.104.Final -> 4.2.15.Final. Arrow 18+ requires JDK 17, so the framework's maven.compiler.source/target is raised from 1.8 to 17 accordingly. This is a dependency-alignment / JDK17 change only. It does not fix the intermittent arrow_flight_sql_p0 close-time buffer leak, which has a separate root cause.
hello-stephen
commented
Jun 22, 2026
Contributor
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
hello-stephen
commented
Jun 22, 2026
Contributor
run buildall |
hello-stephen
commented
Jun 24, 2026
Contributor
run buildall |
…19 driver loads The Arrow Flight SQL JDBC driver bumped from arrow 17 (Java 8 bytecode, class v52) to arrow 19 (Java 11 bytecode, class v55), which a JDK 8 runtime cannot load. Align the regression compile + runtime with fe-core (JDK17): - prepare_java_udf no longer forces JDK 8 for the whole build, so the framework compiles with the ambient JDK (>=17 in CI). - run-regression-test.sh switches to JDK 8 only for the java-udf-src module (which must stay Java 8 for the BE embedded JVM), falling back to the ambient JDK on dev boxes without a JDK 8. - cloud_p0 / vault_p0 run.sh execute the suite under JDK 17 instead of JDK 8, enabling the arrow 19 / netty 4.2.15 alignment (run-regression-test.sh already adds --add-opens=java.base/java.nio for JDK17+).
CalvinKirs
commented
Jun 29, 2026
MemberAuthor
run buildall |
hello-stephen
commented
Jun 29, 2026
Contributor
TPC-H: Total hot run time: 29458 ms |
hello-stephen
commented
Jun 29, 2026
Contributor
TPC-DS: Total hot run time: 171540 ms |
hello-stephen
commented
Jun 29, 2026
Contributor
ClickBench: Total hot run time: 25.21 s |
…rage vault test On JDK 17 the Groovy property access `list.size` on a List<List> result spreads as a GPath expression over the rows and columns instead of returning the row count, throwing "No such property: size for class java.lang.String" when it hits a String column. JDK 8 tolerated it. Use the explicit size() method call, matching the existing vaultInfos.size() usage in the same test, so it works under JDK 17.
CalvinKirs
commented
Jun 29, 2026
MemberAuthor
run buildall |
CalvinKirs
commented
Jun 30, 2026
MemberAuthor
run buildall |
hello-stephen
commented
Jun 30, 2026
Contributor
TPC-H: Total hot run time: 29286 ms |
hello-stephen
commented
Jun 30, 2026
Contributor
TPC-DS: Total hot run time: 172019 ms |
hello-stephen
commented
Jun 30, 2026
Contributor
ClickBench: Total hot run time: 25.47 s |
hello-stephen
commented
Jul 1, 2026
Contributor
/review |
Contributor
There was a problem hiding this comment.
Automated review summary
I reviewed the six changed files in PR 64677: the regression framework POM, cloud/vault pipeline runners, common Java UDF prep, the vault Groovy regression suite, and run-regression-test.sh. I did not find a substantiated issue requiring inline comments.
Critical checkpoint conclusions:
- Goal/test proof: The change aligns the regression framework and suite runtime with JDK17/Arrow 19 while preserving the Java UDF module's JDK8 build path. PR head reports successful COMPILE, P0 Regression, NonConcurrent Regression, External Regression, cloud_p0, vault_p0, CheckStyle, dependency-review, License Check, and secret checks.
- Scope/focus: The modification is small and limited to regression framework/runtime plumbing plus the Groovy row-count
size()fix. - Concurrency/lifecycle: No new concurrent runtime path or persistent lifecycle state is introduced. The shell
JAVA_HOMEswitch for Java UDF compilation is restored before the framework runner computes the runtime Java command. - Config/compatibility: No Doris config item, storage format, FE/BE protocol, or rolling-upgrade behavior is changed. Runtime compatibility is handled by running the regression framework under JDK17 and limiting the JDK8 switch to the Java UDF module.
- Parallel paths: cloud_p0 and vault_p0 run paths both switch to JDK17; their deploy paths already set JDK17 before Java UDF prep and FE/BE startup.
- Tests/results: No expected
.outfile is added or modified. Existing CI status covers the relevant compile and regression jobs. - Observability/performance/security: No new observability need, hot Doris runtime path, or security-sensitive behavior is introduced.
Validation performed:
- Read the required review prompt, code-review skill, root AGENTS.md, existing review threads, review focus, changed-file list, and shared ledger.
- Ran
bash -non the changed shell scripts successfully. - Ran Maven framework
validateunder JDK17 successfully. - Checked the regression framework dependency tree for Arrow/Netty; it resolved Arrow Flight SQL JDBC 19.0.0 and Netty 4.2.15 modules.
- Full local package/run validation was not attempted because this checkout lacks
thirdparty/installed/bin/thrift; PR head CI/TeamCity reports the relevant compile and regression jobs successful.
Subagent conclusions:
- optimizer-rewrite: no candidates; the PR does not alter optimizer/rewrite code paths, and the Nereids-named vault test exercises metadata/show-command behavior.
- tests-session-config: no candidates; shell syntax, JDK propagation, and CI/status coverage looked sound.
- Convergence round 1 ended with both live subagents replying
NO_NEW_VALUABLE_FINDINGSfor the same empty inline-comment set.
User focus: no additional focus was provided.
morningman
approved these changes
Jul 1, 2026
Uh oh!
There was an error while loading. Please reload this page.
63 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Proposed changes
Align the regression-test framework's dependency versions with fe-core (
fe/pom.xml):arrow.version:17.0.0->19.0.0netty-all:4.1.104.Final->4.2.15.Finalmaven.compiler.source/target:1.8->17(Arrow 18+ requires JDK 17, and the regression build already runs on JDK 17)This is a dependency-alignment / JDK17-target change only. It keeps the framework's Arrow Flight SQL JDBC driver and netty in lockstep with fe-core instead of drifting on older versions.
Scope / risk
regression-test/framework/pom.xmlis touched.maven.compiler.targetto 17 affects the compiled output of the whole regression framework, so this needs full regression CI to validate, not just a local build.arrow_flight_sql_p0close-time buffer leak — that has a separate root cause and will be handled in a follow-up.