Add Spark 4.0.0 Build Profile and Other Supporting Changes [databrick…

…s] (#10994) * POM changes for Spark 4.0.0 Signed-off-by: Raza Jafri <rjafri@nvidia.com> * validate buildver and scala versions * more pom changes * fixed the scala-2.12 comment * more fixes for scala-2.13 pom * addressed comments * add in shim check to account for 400 * add 400 for premerge tests against jdk 17 * temporarily remove 400 from snapshotScala213 * fixed 2.13 pom * Remove 400 from jdk17 as it will compile with Scala 2.12 * github workflow changes * added quotes to pom-directory * update version defs to include scala 213 jdk 17 * Cross-compile all shims from JDK17 to JDK8 Eliminate Logging inheritance to prevent shimming of unshimmable API classes Signed-off-by: Gera Shegalov <gera@apache.org> * dummy * undo api pom change Signed-off-by: Gera Shegalov <gera@apache.org> * Add preview1 to the allowed shim versions Signed-off-by: Gera Shegalov <gera@apache.org> * Scala 2.13 to require JDK17 Signed-off-by: Gera Shegalov <gera@apache.org> * Removed unused import left over from razajafri#3 * Setup JAVA_HOME before caching * Only upgrade the Scala plugin for Scala 2.13 * Regenerate Scala 2.13 poms * Remove 330 from JDK17 builds for Scala 2.12 * Revert "Remove 330 from JDK17 builds for Scala 2.12" This reverts commit 1faabd4. * Downgrade scala.plugin.version for cloudera * Updated comment to include the issue * Upgrading the scala.maven.plugin version to 4.9.1 which is the same as Spark 4.0.0 * Downgrade scala-maven-plugin for Cloudera * revert mvn verify changes * Avoid cache for JDK 17 * removed cache dep from scala 213 * Added Scala 2.13 specific checks * Handle the change for UnaryPositive now extending RuntimeReplaceable * Removing 330 from jdk17.buildvers as we only support Scala2.13 and fixing the enviornment variable in version-defs.sh that we read for building against JDK17 with Scala 213 * Update Scala 2.13 poms * fixed scala2.13 verify to actually use the scala2.13/pom.xml * Added missing csv files * Skip Opcode tests There is a bytecode incompatibility which is why we are skipping these until we add support for it. For details please see the following two issues #11174 #10203 * upmerged and fixed the new compile error introduced * addressed review comments * Removed jdk17 cloudera check and moved it inside the 321,330 and 332 cloudera profiles * fixed upmerge conflicts * reverted renaming of id * Fixed HiveGenericUDFShim * addressed review comments * reverted the debugging code * generated Scala 2.13 poms --------- Signed-off-by: Raza Jafri <rjafri@nvidia.com> Signed-off-by: Gera Shegalov <gera@apache.org> Co-authored-by: Gera Shegalov <gera@apache.org>
NVIDIA · Jul 16, 2024 · a41a616 · a41a616
1 parent 156ee51
commit a41a616
Show file tree

Hide file tree

Showing 34 changed files with 1,641 additions and 140 deletions.
diff --git a/.github/workflows/mvn-verify-check.yml b/.github/workflows/mvn-verify-check.yml
@@ -40,10 +40,9 @@ jobs:
     runs-on: ubuntu-latest
     outputs:
       dailyCacheKey: ${{ steps.generateCacheKey.outputs.dailyCacheKey }}
-      defaultSparkVersion: ${{ steps.allShimVersionsStep.outputs.defaultSparkVersion }}
-      sparkTailVersions: ${{ steps.allShimVersionsStep.outputs.tailVersions }}
-      sparkJDKVersions: ${{ steps.allShimVersionsStep.outputs.jdkVersions }}
-      scala213Versions: ${{ steps.allShimVersionsStep.outputs.scala213Versions }}
+      defaultSparkVersion: ${{ steps.all212ShimVersionsStep.outputs.defaultSparkVersion }}
+      sparkTailVersions: ${{ steps.all212ShimVersionsStep.outputs.tailVersions }}
+      sparkJDKVersions: ${{ steps.all212ShimVersionsStep.outputs.jdkVersions }}
     steps:
       - uses: actions/checkout@v4 # refs/pull/:prNumber/merge
       - uses: actions/setup-java@v4
@@ -69,7 +68,7 @@ jobs:
           set -x
           max_retry=3; delay=30; i=1
           while true; do
-            for pom in pom.xml scala2.13/pom.xml
+            for pom in pom.xml 
             do
               mvn ${{ env.COMMON_MVN_FLAGS }} --file $pom help:evaluate -pl dist \
                 -Dexpression=included_buildvers \
@@ -89,7 +88,7 @@ jobs:
           }
           done
       - name: all shim versions
-        id: allShimVersionsStep
+        id: all212ShimVersionsStep
         run: |
           set -x
           . jenkins/version-def.sh
@@ -113,30 +112,12 @@ jobs:
           jdkHeadVersionArrBody=$(printf ",{\"spark-version\":\"%s\",\"java-version\":8}" "${SPARK_BASE_SHIM_VERSION}")
           # jdk11
           jdk11VersionArrBody=$(printf ",{\"spark-version\":\"%s\",\"java-version\":11}" "${SPARK_SHIM_VERSIONS_JDK11[@]}")
-          # jdk17
-          jdk17VersionArrBody=$(printf ",{\"spark-version\":\"%s\",\"java-version\":17}" "${SPARK_SHIM_VERSIONS_JDK17[@]}")
           # jdk
-          jdkVersionArrBody=$jdkHeadVersionArrBody$jdk11VersionArrBody$jdk17VersionArrBody
+          jdkVersionArrBody=$jdkHeadVersionArrBody$jdk11VersionArrBody
           jdkVersionArrBody=${jdkVersionArrBody:1}
           jdkVersionJsonStr=$(printf {\"include\":[%s]} $jdkVersionArrBody)
           echo "jdkVersions=$jdkVersionJsonStr" >> $GITHUB_OUTPUT
 
-          SCALA_BINARY_VER=2.13
-          . jenkins/version-def.sh
-          svArrBodyNoSnapshot=$(printf ",{\"spark-version\":\"%s\",\"isSnapshot\":false}" "${SPARK_SHIM_VERSIONS_NOSNAPSHOTS[@]}")
-          svArrBodyNoSnapshot=${svArrBodyNoSnapshot:1}
-          # get private artifact version
-          privateVer=$(mvn help:evaluate -q -pl dist -Dexpression=spark-rapids-private.version -DforceStdout)
-          # do not add empty snapshot versions or when private version is released one (does not include snapshot shims)
-          if [[ ${#SPARK_SHIM_VERSIONS_SNAPSHOTS_ONLY[@]} -gt 0 && $privateVer == *"-SNAPSHOT" ]]; then
-            svArrBodySnapshot=$(printf ",{\"spark-version\":\"%s\",\"isSnapshot\":true}" "${SPARK_SHIM_VERSIONS_SNAPSHOTS_ONLY[@]}")
-            svArrBodySnapshot=${svArrBodySnapshot:1}
-            svJsonStr=$(printf {\"include\":[%s]} $svArrBodyNoSnapshot,$svArrBodySnapshot)
-          else
-            svJsonStr=$(printf {\"include\":[%s]} $svArrBodyNoSnapshot)
-          fi
-
-          echo "scala213Versions=$svJsonStr" >> $GITHUB_OUTPUT
 
   package-tests:
     needs: cache-dependencies
@@ -187,27 +168,51 @@ jobs:
           }
           done
 
+  set-scala213-versions:
+    runs-on: ubuntu-latest
+    outputs:
+      scala213Versions: ${{ steps.all213ShimVersionsStep.outputs.scala213Versions }}
+      sparkJDK17Versions: ${{ steps.all213ShimVersionsStep.outputs.jdkVersions }}
+    steps:
+      - uses: actions/checkout@v4 # refs/pull/:prNumber/merge
+
+      - id: all213ShimVersionsStep
+        run: |
+          set -x
+          SCALA_BINARY_VER=2.13
+          . jenkins/version-def.sh
+          svArrBodyNoSnapshot=$(printf ",{\"spark-version\":\"%s\",\"isSnapshot\":false}" "${SPARK_SHIM_VERSIONS_NOSNAPSHOTS[@]}")
+          svArrBodyNoSnapshot=${svArrBodyNoSnapshot:1}
+          # get private artifact version
+          privateVer=$(mvn help:evaluate -q -pl dist -Dexpression=spark-rapids-private.version -DforceStdout)
+          svJsonStr=$(printf {\"include\":[%s]} $svArrBodyNoSnapshot)
+
+          echo "scala213Versions=$svJsonStr" >> $GITHUB_OUTPUT
+
+          # jdk17
+          jdk17VersionArrBody=$(printf ",{\"spark-version\":\"%s\",\"java-version\":17}" "${SPARK_SHIM_VERSIONS_JDK17_SCALA213[@]}")
+
+          jdkVersionArrBody=$jdk17VersionArrBody
+          jdkVersionArrBody=${jdkVersionArrBody:1}
+          jdkVersionJsonStr=$(printf {\"include\":[%s]} $jdkVersionArrBody)
+          echo "jdkVersions=$jdkVersionJsonStr" >> $GITHUB_OUTPUT
+
   package-tests-scala213:
-    needs: cache-dependencies
+    needs: set-scala213-versions
     continue-on-error: ${{ matrix.isSnapshot }}
     strategy:
-      matrix: ${{ fromJSON(needs.cache-dependencies.outputs.scala213Versions) }}
+      matrix: ${{ fromJSON(needs.set-scala213-versions.outputs.scala213Versions) }}
       fail-fast: false
     runs-on: ubuntu-latest
     steps:
+
       - uses: actions/checkout@v4 # refs/pull/:prNumber/merge
 
       - name: Setup Java and Maven Env
         uses: actions/setup-java@v4
         with:
           distribution: adopt
-          java-version: 8
-
-      - name: Cache local Maven repository
-        uses: actions/cache@v4
-        with:
-          path: ~/.m2
-          key: ${{ needs.cache-dependencies.outputs.dailyCacheKey }}
+          java-version: 17
 
       - name: check runtime before tests
         run: |
@@ -218,7 +223,7 @@ jobs:
         run: |
           # https://github.com/NVIDIA/spark-rapids/issues/8847
           # specify expected versions
-          export JAVA_HOME=${JAVA_HOME_8_X64}
+          export JAVA_HOME=${JAVA_HOME_17_X64}
           export PATH=${JAVA_HOME}/bin:${PATH}
           java -version && mvn --version && echo "ENV JAVA_HOME: $JAVA_HOME, PATH: $PATH"
           # verify Scala 2.13 build files
@@ -246,8 +251,57 @@ jobs:
           }
           done
 
+  verify-213-modules:
+    needs: set-scala213-versions
+    runs-on: ubuntu-latest
+    strategy:
+      matrix: ${{ fromJSON(needs.set-scala213-versions.outputs.sparkJDK17Versions) }}
+    steps:
+      - uses: actions/checkout@v4 # refs/pull/:prNumber/merge
+
+      - name: Setup Java and Maven Env
+        uses: actions/setup-java@v4
+        with:
+          distribution: adopt
+          java-version: 17
+
+      - name: check runtime before tests
+        run: |
+          env | grep JAVA
+          java -version && mvn --version && echo "ENV JAVA_HOME: $JAVA_HOME, PATH: $PATH"
+
+      - name: Build JDK
+        run: |
+          # https://github.com/NVIDIA/spark-rapids/issues/8847
+          # specify expected versions
+          export JAVA_HOME=${JAVA_HOME_${{ matrix.java-version }}_X64}
+          export PATH=${JAVA_HOME}/bin:${PATH}
+          java -version && mvn --version && echo "ENV JAVA_HOME: $JAVA_HOME, PATH: $PATH"
+           # verify Scala 2.13 build files
+          ./build/make-scala-version-build-files.sh 2.13
+          # verify git status
+          if [ -n "$(echo -n $(git status -s | grep 'scala2.13'))" ]; then
+              git add -N scala2.13/* && git diff 'scala2.13/*'
+              echo "Generated Scala 2.13 build files don't match what's in repository"
+              exit 1
+          fi
+          # change to Scala 2.13 Directory
+          cd scala2.13 
+          # test command, will retry for 3 times if failed.
+          max_retry=3; delay=30; i=1
+          while true; do
+            mvn verify \
+              -P "individual,pre-merge" -Dbuildver=${{ matrix.spark-version }} \
+              ${{ env.COMMON_MVN_FLAGS }} && break || {
+            if [[ $i -le $max_retry ]]; then
+              echo "mvn command failed. Retry $i/$max_retry."; ((i++)); sleep $delay; ((delay=delay*2))
+            else
+              echo "mvn command failed. Exit 1"; exit 1
+            fi
+          }
+          done
 
-  verify-all-modules:
+  verify-all-212-modules:
     needs: cache-dependencies
     runs-on: ubuntu-latest
     strategy:

diff --git a/aggregator/pom.xml b/aggregator/pom.xml
@@ -715,5 +715,24 @@
                 </dependency>
             </dependencies>
         </profile>
+        <!-- #if scala-2.13 --><!--
+        <profile>
+            <id>release400</id>
+            <activation>
+                <property>
+                    <name>buildver</name>
+                    <value>400</value>
+                </property>
+            </activation>
+            <dependencies>
+                <dependency>
+                    <groupId>com.nvidia</groupId>
+                    <artifactId>rapids-4-spark-delta-stub_${scala.binary.version}</artifactId>
+                    <version>${project.version}</version>
+                    <classifier>${spark.version.classifier}</classifier>
+                </dependency>
+            </dependencies>
+        </profile>
+        --><!-- #endif scala-2.13 -->
     </profiles>
 </project>
diff --git a/build/buildall b/build/buildall
@@ -161,7 +161,6 @@ if [[ "$DIST_PROFILE" == *Scala213 ]]; then
   SCALA213=1
 fi
 
-
 # include options to mvn command
 export MVN="mvn -Dmaven.wagon.http.retryHandler.count=3 ${MVN_OPT}"
 
@@ -196,7 +195,7 @@ case $DIST_PROFILE in
     SPARK_SHIM_VERSIONS=($(versionsFromDistProfile "minimumFeatureVersionMix"))
     ;;
 
-  3*)
+  [34]*)
     <<< $DIST_PROFILE IFS="," read -ra SPARK_SHIM_VERSIONS
     INCLUDED_BUILDVERS_OPT="-Dincluded_buildvers=$DIST_PROFILE"
     unset DIST_PROFILE

diff --git a/build/shimplify.py b/build/shimplify.py
@@ -188,6 +188,7 @@ def __csv_as_arr(str_val):
 __dirs_to_derive_shims = sorted(__csv_ant_prop_as_arr('shimplify.dirs'))
 
 __all_shims_arr = sorted(__csv_ant_prop_as_arr('all.buildvers'))
+__allScala213_shims_arr = sorted(__csv_ant_prop_as_arr('allScala213.buildvers'))
 
 __log = logging.getLogger('shimplify')
 __log.setLevel(logging.DEBUG if __should_trace else logging.INFO)
@@ -372,7 +373,8 @@ def __generate_symlinks():
 
 def __map_version_array(shim_json_string):
     shim_ver = str(json.loads(shim_json_string).get('spark'))
-    assert shim_ver in __all_shims_arr, "all.buildvers in pom.xml does not contain %s" % shim_ver
+    assert shim_ver in __all_shims_arr or shim_ver in __allScala213_shims_arr, "all.buildvers or " \
+        "allScala213.buildvers in pom.xml does not contain %s" % shim_ver
     return shim_ver
 
 def __traverse_source_tree_of_all_shims(src_type, func):

diff --git a/dist/pom.xml b/dist/pom.xml
@@ -110,6 +110,14 @@
                 </included_buildvers>
             </properties>
         </profile>
+        <profile>
+            <id>jdk17-scala213-test</id>
+            <properties>
+                <included_buildvers>
+                    ${jdk17.scala213.buildvers}
+                </included_buildvers>
+            </properties>
+        </profile>
         <profile>
             <id>jdk17-test</id>
             <properties>

diff --git a/dist/scripts/binary-dedupe.sh b/dist/scripts/binary-dedupe.sh
@@ -168,12 +168,12 @@ function verify_same_sha_for_unshimmed() {
   # TODO currently RapidsShuffleManager is "removed" from /spark* by construction in
   # dist pom.xml via ant. We could delegate this logic to this script
   # and make both simmpler
-  if [[ ! "$class_file_quoted" =~ (com/nvidia/spark/rapids/spark[34].*/.*ShuffleManager.class|org/apache/spark/sql/rapids/shims/spark[34].*/ProxyRapidsShuffleInternalManager.class) ]]; then
+  if [[ ! "$class_file_quoted" =~ com/nvidia/spark/rapids/spark[34].*/.*ShuffleManager.class ]]; then
 
-    if ! grep -q "/spark.\+/$class_file_quoted" "$SPARK_SHARED_TXT"; then
-      echo >&2 "$class_file is not bitwise-identical across shims"
-      exit 255
-    fi
+      if ! grep -q "/spark.\+/$class_file_quoted" "$SPARK_SHARED_TXT"; then
+        echo >&2 "$class_file is not bitwise-identical across shims"
+        exit 255
+      fi
   fi
 }
 

diff --git a/jdk-profiles/pom.xml b/jdk-profiles/pom.xml
@@ -31,17 +31,45 @@
     <version>24.08.0-SNAPSHOT</version>
     <profiles>
         <profile>
-            <id>jdk9plus</id>
-            <properties>
-                <scala.plugin.version>4.6.1</scala.plugin.version>
-                <maven.compiler.source>${java.specification.version}</maven.compiler.source>
-                <maven.compiler.release>${maven.compiler.source}</maven.compiler.release>
-                <maven.compiler.target>${maven.compiler.source}</maven.compiler.target>
-            </properties>
+            <id>jdk8</id>
             <activation>
+                <jdk>8</jdk>
+            </activation>
+            <build>
+                <pluginManagement>
+                    <plugins>
+                        <plugin>
+                            <groupId>net.alchim31.maven</groupId>
+                            <artifactId>scala-maven-plugin</artifactId>
+                            <version>${scala.plugin.version}</version>
+                            <configuration>
+                                <target>${java.major.version}</target>
+                            </configuration>
+                        </plugin>
+                    </plugins>
+                </pluginManagement>
+            </build>
+        </profile>
+        <profile>
+            <id>jdk9plus</id>
+             <activation>
                 <!-- activate for all java versions after 9 -->
                 <jdk>[9,)</jdk>
             </activation>
+            <build>
+                <pluginManagement>
+                    <plugins>
+                        <plugin>
+                            <groupId>net.alchim31.maven</groupId>
+                            <artifactId>scala-maven-plugin</artifactId>
+                            <version>${scala.plugin.version}</version>
+                            <configuration>
+                                <release>${java.major.version}</release>
+                            </configuration>
+                        </plugin>
+                    </plugins>
+                </pluginManagement>
+            </build>
         </profile>
     </profiles>
 </project>
diff --git a/jenkins/version-def.sh b/jenkins/version-def.sh
@@ -125,6 +125,9 @@ SPARK_SHIM_VERSIONS_JDK11=("${SPARK_SHIM_VERSIONS_ARR[@]}")
 # jdk17 cases
 set_env_var_SPARK_SHIM_VERSIONS_ARR -Pjdk17-test
 SPARK_SHIM_VERSIONS_JDK17=("${SPARK_SHIM_VERSIONS_ARR[@]}")
+# jdk17 scala213 cases
+set_env_var_SPARK_SHIM_VERSIONS_ARR -Pjdk17-scala213-test
+SPARK_SHIM_VERSIONS_JDK17_SCALA213=("${SPARK_SHIM_VERSIONS_ARR[@]}")
 # databricks shims
 set_env_var_SPARK_SHIM_VERSIONS_ARR -Pdatabricks
 SPARK_SHIM_VERSIONS_DATABRICKS=("${SPARK_SHIM_VERSIONS_ARR[@]}")