felipepessoto commented on code in PR #13560:
URL: https://github.com/apache/gluten/pull/13560#discussion_r4543208685
##########
.github/workflows/velox_backend_x86.yml:
##########
@@ -157,6 +159,7 @@ jobs:
echo "cpp=$(match '^(cpp/|ep/build-velox/|dev/)')"
>> $GITHUB_OUTPUT
echo "java=$(match
'^(\.github/workflows/|pom\.xml|backends-velox/|gluten-(uniffle|celeborn|ras|core|substrait|arrow|delta|iceberg|hudi|paimon)/|gluten-ut/(common/|test/|src-iceberg-upstream/|pom\.xml)|package/|build/mvn)')"
>> $GITHUB_OUTPUT
+ echo "jni_test=$(match
'^(\.github/workflows/velox_backend_x86\.yml|pom\.xml|gluten-(arrow|core)/pom\.xml|build/mvn|gluten-arrow/src/test/java/org/apache/gluten/jni/.*|gluten-arrow/src/main/java/org/apache/gluten/exception/JniExceptionDescriber\.java|gluten-core/src/main/java/org/apache/gluten/exception/GlutenException\.java)$')"
>> $GITHUB_OUTPUT
Review Comment:
Including the root POM is intentional: it defines the inherited Surefire
configuration, test JVM arguments, JUnit version, and Java profiles used by
these tests. Excluding that file would also skip JNI coverage when those
settings change.
The file-level filter does run the job for unrelated edits within that POM,
so there is a CI-cost trade-off. I recommend retaining this conservative
coverage rather than adding content-sensitive detection or splitting the test
job again. Leaving this as a CI-cost discussion, not a correctness fix.
##########
.github/workflows/velox_backend_x86.yml:
##########
@@ -1061,7 +1064,9 @@ jobs:
cpp-test-udf-test:
needs: detect-changes
- if: needs.detect-changes.outputs.cpp == 'true'
+ if: >-
+ needs.detect-changes.outputs.cpp == 'true' ||
+ needs.detect-changes.outputs.jni_test == 'true'
Review Comment:
This is identical to the [root-POM trigger
comment](https://github.com/apache/gluten/pull/13560#discussion_r4542920936).
Keeping the rationale and CI-cost discussion in that thread and resolving this
duplicate. No trigger behavior has been changed.
##########
gluten-arrow/pom.xml:
##########
@@ -278,4 +278,47 @@
<testOutputDirectory>target/scala-${scala.binary.version}/test-classes</testOutputDirectory>
</build>
+ <profiles>
+ <profile>
+ <id>exclude-jni-test</id>
+ <activation>
+ <activeByDefault>true</activeByDefault>
+ </activation>
+ <build>
+ <plugins>
+ <plugin>
+ <groupId>org.apache.maven.plugins</groupId>
+ <artifactId>maven-surefire-plugin</artifactId>
+ <configuration>
+ <excludes combine.children="append">
+ <!-- The jni-test profile requires the separately built native
test bridge. -->
+ <exclude>**/JavaExceptionJniTest.java</exclude>
+ </excludes>
+ </configuration>
+ </plugin>
+ </plugins>
+ </build>
+ </profile>
Review Comment:
This repeats the [previously verified profile-activation
finding](https://github.com/apache/gluten/pull/13560#discussion_r4541164647).
`activeByDefault` is deactivated by another active profile in the same POM, not
by profiles in the parent POM. The listed Spark/backend/Java profiles are in
the parent.
Both POMs on the rebased head still match the executed regression: with the
parent profiles enabled, normal mode excludes the JNI fixture and needs no
bridge; explicitly enabling `jni-test` runs it. Restoring a blanket exclusion
override would reintroduce the inherited-exclusions problem already fixed.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]