Copilot commented on code in PR #12278:
URL: https://github.com/apache/gluten/pull/12278#discussion_r3470654681


##########
.github/workflows/util/delta-spark-ut/compare-test-results.py:
##########
@@ -0,0 +1,467 @@
+#!/usr/bin/env python3
+#
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Gate / seed / aggregate the Delta-on-Gluten unit test results.
+
+Running delta-io/delta's ScalaTest suite against the Gluten Velox bundle
+produces many *expected* failures (Gluten does not yet support every Delta
+code path). To keep the red/green signal meaningful while we fix those
+failures incrementally, we maintain a committed baseline of known failing
+tests (``known-failures.txt``) and compare each CI run against it.
+
+This script has three modes:
+
+``enforce`` (default, per shard)
+    Parse the JUnit XML produced by ``sbt spark/test`` (ScalaTest ``-u``
+    reporter) and compare against the baseline:
+
+      * regression -- a test that FAILED but is NOT in the baseline. These
+        fail the build: a previously-passing test just started failing.
+      * expected   -- a test that failed and IS in the baseline. Ignored.
+      * fixed      -- a baseline test that now PASSES. By default these also
+        fail the build (``--fail-on-fixed true``) so the baseline stays honest
+        and contributors remove entries as they fix them.
+
+    If the baseline is empty (not yet bootstrapped) the mode automatically
+    degrades to ``seed`` so the first run is never spuriously red.
+
+``seed`` (bootstrap / ``update_baseline``)
+    Never fails. Just writes the current shard's failing tests so the baseline
+    can be (re)generated from a real run.
+
+``aggregate`` (final job)
+    Merge every shard's ``--failures-out`` / ``--ran-out`` file into a single,
+    sorted, ready-to-commit ``known-failures.txt`` and report stale baseline
+    entries (tests no longer present in any shard).
+
+Baseline file format (``known-failures.txt``)::
+
+    # comment lines start with '#'
+    <fully.qualified.SuiteName>#<test display name>
+
+The suite is always a JVM class name (dot-separated, never starts with '#'),
+so a line whose first non-space character is '#' is unambiguously a comment,
+and the FIRST '#' after the suite separates suite from the (possibly
+'#'-containing) test name.
+
+Only the Python standard library is used so the script runs in the bare
+centos image used by the Delta UT pipeline with no ``pip install``.
+"""
+
+import argparse
+import glob
+import os
+import sys
+import xml.etree.ElementTree as ET
+
+# Synthetic "test name" recorded when a whole suite aborts (e.g. beforeAll
+# throws) so that the JUnit XML reports a suite-level error with no per-test
+# <testcase>. Without this, a suite that used to pass but now aborts entirely
+# would record zero failing testcases and the regression would be missed.
+SUITE_ABORTED = "<suite aborted>"
+
+SEP = "#"
+
+
+def eprint(*args, **kwargs):
+    print(*args, file=sys.stderr, **kwargs)
+
+
+# --------------------------------------------------------------------------- #
+# Baseline (known-failures.txt) parsing / formatting
+# --------------------------------------------------------------------------- #
+def format_entry(suite, test):
+    return "{}{}{}".format(suite, SEP, test)
+
+
+def parse_entry(line):
+    """Parse a 'suite#test' line into (suite, test) or return None for 
blanks/comments."""
+    stripped = line.strip()
+    if not stripped or stripped.startswith("#"):
+        return None
+    idx = stripped.find(SEP)
+    if idx < 0:
+        # No separator: treat the whole line as a suite-level entry.
+        return (stripped, SUITE_ABORTED)
+    return (stripped[:idx], stripped[idx + len(SEP) :])
+
+
+def load_entries(path):
+    """Load a set of (suite, test) tuples from a baseline/shard-list file."""
+    entries = set()
+    if not path or not os.path.exists(path):
+        return entries
+    with open(path, "r", encoding="utf-8") as fh:
+        for line in fh:
+            parsed = parse_entry(line)
+            if parsed is not None:
+                entries.add(parsed)
+    return entries
+
+
+def write_entries(path, entries, header=None):
+    """Write a sorted set of (suite, test) tuples to a file."""
+    os.makedirs(os.path.dirname(os.path.abspath(path)) or ".", exist_ok=True)
+    with open(path, "w", encoding="utf-8") as fh:
+        if header:
+            for hl in header.splitlines():
+                fh.write(hl.rstrip() + "\n")
+        for suite, test in sorted(entries):
+            # Defensive: collapse any stray newlines so each entry stays on 
one line.
+            safe_test = test.replace("\r", " ").replace("\n", " ")
+            fh.write(format_entry(suite, safe_test) + "\n")
+
+
+# --------------------------------------------------------------------------- #
+# JUnit XML parsing
+# --------------------------------------------------------------------------- #
+def _iter_testsuites(root):
+    """Yield every <testsuite> element regardless of whether the file root is
+    <testsuites> (wrapper) or a single <testsuite>."""
+    tag = root.tag.split("}")[-1]  # strip any namespace
+    if tag == "testsuites":
+        for child in root:
+            if child.tag.split("}")[-1] == "testsuite":
+                yield child
+    elif tag == "testsuite":
+        yield root
+
+
+def _child_local_tags(elem):
+    return {c.tag.split("}")[-1] for c in elem}
+
+
+def parse_reports(reports_dir):
+    """Walk reports_dir for JUnit XML and classify every test.
+
+    Returns (passed, failed, skipped) sets of (suite, test) tuples. A test is
+    'failed' if its <testcase> has a <failure> or <error> child, 'skipped' if
+    it has a <skipped> child, otherwise 'passed'. Suite-level aborts (a
+    <testsuite> reporting errors/failures with no failing <testcase>) are
+    recorded as a synthetic (suite, SUITE_ABORTED) failure.
+    """
+    passed, failed, skipped = set(), set(), set()
+
+    xml_files = []
+    # ScalaTest's -u reporter and Maven surefire both write `TEST-<suite>.xml`
+    # under a `target/.../*-reports/` dir. Restrict the secondary glob to
+    # `target/` so we never parse Delta's own XML *test resources* (which live
+    # under src/test/resources and are not reports). The <testsuite>-root guard
+    # below is a final safety net.
+    for pattern in ("**/TEST-*.xml", "**/target/**/*.xml"):
+        xml_files.extend(glob.glob(os.path.join(reports_dir, pattern), 
recursive=True))
+    xml_files = sorted(set(xml_files))
+
+    parsed_any = False
+    for xml_file in xml_files:
+        try:
+            tree = ET.parse(xml_file)
+        except ET.ParseError as exc:
+            eprint("WARNING: could not parse {}: {}".format(xml_file, exc))
+            continue
+        root = tree.getroot()
+        root_tag = root.tag.split("}")[-1]
+        if root_tag not in ("testsuites", "testsuite"):
+            continue  # not a JUnit report
+
+        for ts in _iter_testsuites(root):
+            parsed_any = True
+            suite_name = ts.get("name") or ""
+            suite_has_failing_tc = False
+            for tc in ts:
+                if tc.tag.split("}")[-1] != "testcase":
+                    continue
+                suite = tc.get("classname") or suite_name
+                name = tc.get("name") or ""
+                key = (suite, name)
+                tags = _child_local_tags(tc)
+                if "failure" in tags or "error" in tags:
+                    failed.add(key)
+                    suite_has_failing_tc = True
+                elif "skipped" in tags:
+                    skipped.add(key)
+                else:
+                    passed.add(key)
+
+            # Suite-level abort: counters say something failed but no testcase
+            # carried the failure (the suite blew up in beforeAll/constructor).
+            # Record a
+            # synthetic entry so the regression is visible.
+            try:
+                errors = int(ts.get("errors", "0") or "0")
+                failures = int(ts.get("failures", "0") or "0")
+            except ValueError:
+                errors = failures = 0
+            if (errors + failures) > 0 and not suite_has_failing_tc:
+                failed.add((suite_name, SUITE_ABORTED))
+
+    if not parsed_any:
+        eprint(
+            "WARNING: no JUnit <testsuite> elements found under 
{}".format(reports_dir)
+        )
+
+    # A test can't be both passed and failed; failure wins. Skipped only counts
+    # if the test was not otherwise seen (e.g. retried).
+    passed -= failed
+    skipped -= failed
+    skipped -= passed
+    return passed, failed, skipped
+
+
+# --------------------------------------------------------------------------- #
+# Reporting helpers
+# --------------------------------------------------------------------------- #
+def _summary_sink():
+    """Return a writer that mirrors to GITHUB_STEP_SUMMARY when available."""
+    path = os.environ.get("GITHUB_STEP_SUMMARY")
+    handle = open(path, "a", encoding="utf-8") if path else None
+
+    def write(line=""):
+        print(line)
+        if handle:
+            handle.write(line + "\n")
+
+    return write, handle
+
+
+def _print_block(write, title, entries, limit=50):
+    write("")
+    write("### {} ({})".format(title, len(entries)))
+    if not entries:
+        return
+    write("")
+    write("```")
+    for i, (suite, test) in enumerate(sorted(entries)):
+        if i >= limit:
+            write("... and {} more".format(len(entries) - limit))
+            break
+        write(format_entry(suite, test))
+    write("```")
+
+
+# --------------------------------------------------------------------------- #
+# Modes
+# --------------------------------------------------------------------------- #
+def run_enforce(args):
+    baseline = load_entries(args.known_failures)
+    passed, failed, skipped = parse_reports(args.reports_dir)
+
+    # Always emit this shard's artifacts for the aggregation job.
+    if args.failures_out:
+        write_entries(args.failures_out, failed)
+    if args.ran_out:
+        write_entries(args.ran_out, passed | failed)
+

Review Comment:
   `--ran-out` is currently written as `passed | failed` only, which excludes 
tests that were reported as `<skipped>` in JUnit XML. In aggregate mode, 
`union_ran` is used to detect stale baseline entries (`baseline - union_ran`), 
so a test that still exists but is skipped will be incorrectly flagged as stale 
(and can also affect global "fixed" calculations). Include skipped tests in 
`ran-out` so the aggregate job reflects "seen this run" rather than only 
pass/fail.



##########
.github/workflows/delta_spark_ut.yml:
##########
@@ -0,0 +1,659 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# Runs Delta Lake's `spark` sbt module unit tests against a Gluten Velox bundle
+# that is built from the source in this repository. The pipeline:
+#
+#   1. Builds the Velox/Gluten native libraries (centos-7 + vcpkg, x86_64).
+#   2. Builds the Gluten Java/Scala jars and assembles the
+#      `gluten-velox-bundle-spark<spark>_<scala>-linux_amd64-<version>.jar`
+#      fat jar for Spark 4.1 + Scala 2.13 + Java 17 with the Delta profile.
+#   3. Clones delta-io/delta at the requested release tag (default `v4.2.0`),
+#      drops the bundle jar into `spark-unified/lib/` only (NOT `spark/lib/`
+#      -- see setup-delta.sh for the unmanagedJars scoping rationale),
+#      patches Delta's `DeltaSQLCommandTest` to register the Gluten plugin,
+#      and runs `sbt spark/test` sharded across the matrix.
+#
+# Limited to Velox + x86 to keep the matrix simple, per the pipeline's purpose
+# of validating Gluten changes against the latest Delta release.
+
+name: Delta Spark UT (Gluten)
+
+on:
+  workflow_dispatch:
+    inputs:
+      delta_ref:
+        description: 'delta-io/delta git ref (tag/branch/SHA) to test against'
+        required: true
+        default: 'v4.2.0'
+      spark_version:
+        description: 'Delta `-DsparkVersion` value (must match the Gluten -P 
profile below)'
+        required: true
+        default: '4.1'
+      test_parallelism:
+        description: 'Forked test JVMs per shard (TEST_PARALLELISM_COUNT)'
+        required: true
+        default: '1'
+      update_baseline:
+        description: 'Seed/refresh the known-failures baseline instead of 
enforcing it'
+        type: boolean
+        required: false
+        default: false
+      fail_on_fixed:
+        description: 'Fail when a baseline test now passes (keeps the baseline 
honest)'
+        type: boolean
+        required: false
+        default: true
+  pull_request:
+    paths:
+      - '.github/workflows/delta_spark_ut.yml'
+      - '.github/workflows/util/delta-spark-ut/**'
+      - 'gluten-delta/**'
+      - 'backends-velox/src-delta40/**/DeltaSQLCommandTest.scala'
+
+env:
+  ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
+  MVN_CMD: 'build/mvn -ntp'
+  CCACHE_DIR: "${{ github.workspace }}/.ccache"
+  # Gluten profile / bundle naming for the build-gluten-bundle and
+  # delta-spark-test jobs. Spark 4.1 + Scala 2.13 + JDK 17 matches Delta 
v4.2.0's
+  # default Spark version (4.1.0) from project/CrossSparkVersions.scala.
+  GLUTEN_SPARK_PROFILE: 'spark-4.1'
+  GLUTEN_SCALA_PROFILE: 'scala-2.13'
+  GLUTEN_JAVA_PROFILE: 'java-17'
+  GLUTEN_BUNDLE_SPARK_VERSION: '4.1'
+  GLUTEN_BUNDLE_SCALA_VERSION: '2.13'
+  # Default values used when the workflow is triggered by pull_request
+  # (where `inputs.*` is empty). Keep these in sync with the workflow_dispatch
+  # defaults above.
+  DELTA_REF_DEFAULT: 'v4.2.0'
+  DELTA_SPARK_VERSION_DEFAULT: '4.1'
+  DELTA_TEST_PARALLELISM_DEFAULT: '1'
+  # Default mode for pull_request runs (where inputs.* is empty): enforce the
+  # committed baseline and fail when a baseline test starts passing. Override
+  # via the workflow_dispatch inputs above.
+  DELTA_UPDATE_BASELINE_DEFAULT: 'false'
+  DELTA_FAIL_ON_FIXED_DEFAULT: 'true'
+  DELTA_SCALA_VERSION: '2.13.16'
+  # Number of shards in the delta-spark-test matrix. Must equal the length of
+  # the `shard` matrix below.
+  #
+  # EXPERIMENT: temporarily set to 4 (production value is 16) to measure the
+  # wall-clock vs concurrent-runner tradeoff. Sharding is by SUITE, so the 
total
+  # test work (~1250 shard-minutes, measured) is fixed regardless of shard 
count
+  # -- more shards = shorter wall-clock but more concurrent runners; fewer 
shards
+  # = longer per-shard time at the same total runner-hours. Measured: 16 shards
+  # ~= 135 min wall-clock (max shard ~110 min). 4 shards carry ~1/4 of the 
suites
+  # each (~300+ min/shard), hence the raised job timeout below. (8-way 
previously
+  # hit the timeout on the slow merge/CDC/DV suites, which is why prod is 16.)
+  # TEST_PARALLELISM_COUNT=1 because each forked test JVM uses ~4G (2G heap +
+  # 2G off-heap) and >1, atop the 4G sbt launcher, would OOM the runner.
+  DELTA_NUM_SHARDS: '4'
+
+concurrency:
+  group: ${{ github.repository }}-${{ github.head_ref || github.sha }}-${{ 
github.workflow }}
+  cancel-in-progress: true
+
+jobs:
+  build-native-lib-centos-7:
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - name: Get Ccache
+        uses: actions/cache/restore@v4
+        with:
+          path: '${{ env.CCACHE_DIR }}'
+          key: ccache-delta-spark-ut-centos7-release-default-${{github.sha}}
+          restore-keys: |
+            ccache-delta-spark-ut-centos7-release-default
+            ccache-centos7-release-default
+      - name: Build Gluten native libraries
+        run: |
+          docker run -v $GITHUB_WORKSPACE:/work -w /work 
apache/gluten:vcpkg-centos-7-gcc13 bash -c "
+            set -e
+            yum install tzdata -y
+            df -a
+            cd /work
+            export CCACHE_DIR=/work/.ccache
+            export CCACHE_MAXSIZE=1G
+            mkdir -p /work/.ccache
+            ccache -sz
+            bash dev/ci-velox-buildstatic-centos-7.sh
+            ccache -s
+            mkdir -p /work/.m2/repository/org/apache/arrow/
+            cp -r /root/.m2/repository/org/apache/arrow/* 
/work/.m2/repository/org/apache/arrow/
+          "
+      - name: Save Ccache
+        if: always()
+        uses: actions/cache/save@v4
+        with:
+          path: '${{ env.CCACHE_DIR }}'
+          key: ccache-delta-spark-ut-centos7-release-default-${{github.sha}}
+      - uses: actions/upload-artifact@v4
+        with:
+          name: delta-spark-ut-native-lib-centos-7-${{github.sha}}
+          path: ./cpp/build/
+          if-no-files-found: error
+      - uses: actions/upload-artifact@v4
+        with:
+          name: delta-spark-ut-arrow-jars-centos-7-${{github.sha}}
+          path: .m2/repository/org/apache/arrow/
+          if-no-files-found: error
+
+  build-gluten-bundle:
+    needs: build-native-lib-centos-7
+    runs-on: ubuntu-22.04
+    container: apache/gluten:centos-9-jdk17
+    steps:
+      - uses: actions/checkout@v4
+      - name: Download native artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: delta-spark-ut-native-lib-centos-7-${{github.sha}}
+          path: ./cpp/build/
+      - name: Download Arrow jars
+        uses: actions/download-artifact@v4
+        with:
+          name: delta-spark-ut-arrow-jars-centos-7-${{github.sha}}
+          path: /root/.m2/repository/org/apache/arrow/
+      - name: Cache Maven repository
+        uses: actions/cache@v4
+        with:
+          path: /root/.m2/repository
+          key: m2-delta-spark-ut-bundle-${{ env.GLUTEN_SPARK_PROFILE }}-${{ 
env.GLUTEN_SCALA_PROFILE }}-${{ hashFiles('pom.xml', '**/pom.xml') }}
+          restore-keys: |
+            m2-delta-spark-ut-bundle-${{ env.GLUTEN_SPARK_PROFILE }}-${{ 
env.GLUTEN_SCALA_PROFILE }}-
+            m2-delta-spark-ut-bundle-
+      - name: Build Gluten Velox + Delta bundle
+        run: |
+          set -euo pipefail
+          yum install -y java-17-openjdk-devel
+          export JAVA_HOME=/usr/lib/jvm/java-17-openjdk
+          export PATH=$JAVA_HOME/bin:$PATH
+          java -version
+          cd "$GITHUB_WORKSPACE"
+          # `install` (not `package`) so the gluten-delta artifact is in the 
local
+          # m2 repo before the `package/` shaded jar is built. 
`Dmaven.compiler.release=17`
+          # overrides any user settings.xml that may pin release=1.8 for Java 
17 builds.
+          $MVN_CMD clean install \
+            -P${{ env.GLUTEN_SPARK_PROFILE }} \
+            -P${{ env.GLUTEN_SCALA_PROFILE }} \
+            -P${{ env.GLUTEN_JAVA_PROFILE }} \
+            -Pbackends-velox -Pdelta \
+            -DskipTests -Dmaven.compiler.release=17
+      - name: Stage bundle jar
+        run: |
+          set -euo pipefail
+          mkdir -p bundle-out
+          # Match the renamed fat jar produced by package/pom.xml's 
copy-fat-jar
+          # exec. The version part may bump (e.g. 1.7.0-SNAPSHOT -> 
1.8.0-SNAPSHOT),
+          # so glob the version suffix.
+          jar=$(ls package/target/gluten-velox-bundle-spark${{ 
env.GLUTEN_BUNDLE_SPARK_VERSION }}_${{ env.GLUTEN_BUNDLE_SCALA_VERSION 
}}-linux_amd64-*.jar | head -n 1)
+          if [ -z "$jar" ] || [ ! -f "$jar" ]; then
+            echo "ERROR: Could not find Gluten bundle jar under 
package/target/" >&2
+            ls -la package/target/ || true
+            exit 1
+          fi
+          cp "$jar" bundle-out/
+          ls -lh bundle-out/
+      - uses: actions/upload-artifact@v4
+        with:
+          name: delta-spark-ut-gluten-bundle-${{github.sha}}
+          path: bundle-out/gluten-velox-bundle-spark*_*-linux_amd64-*.jar
+          if-no-files-found: error
+
+  delta-spark-test:
+    needs: build-gluten-bundle
+    runs-on: ubuntu-22.04
+    container: apache/gluten:centos-9-jdk17
+    # EXPERIMENT (4-shard): raised from 350 so a 4-shard run (~300+ min/shard)
+    # can finish and be measured instead of timing out. Restore 350 with 16 
shards.
+    timeout-minutes: 480

Review Comment:
   `timeout-minutes: 480` exceeds the documented max timeout for GitHub-hosted 
runners (6 hours / 360 minutes). On hosted runners this will still time out at 
the platform limit, which can be surprising when reading the workflow. Consider 
setting this to 360 (or below) and using sharding/parallelism to control wall 
time.



##########
.github/workflows/delta_spark_ut.yml:
##########
@@ -0,0 +1,659 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# Runs Delta Lake's `spark` sbt module unit tests against a Gluten Velox bundle
+# that is built from the source in this repository. The pipeline:
+#
+#   1. Builds the Velox/Gluten native libraries (centos-7 + vcpkg, x86_64).
+#   2. Builds the Gluten Java/Scala jars and assembles the
+#      `gluten-velox-bundle-spark<spark>_<scala>-linux_amd64-<version>.jar`
+#      fat jar for Spark 4.1 + Scala 2.13 + Java 17 with the Delta profile.
+#   3. Clones delta-io/delta at the requested release tag (default `v4.2.0`),
+#      drops the bundle jar into `spark-unified/lib/` only (NOT `spark/lib/`
+#      -- see setup-delta.sh for the unmanagedJars scoping rationale),
+#      patches Delta's `DeltaSQLCommandTest` to register the Gluten plugin,
+#      and runs `sbt spark/test` sharded across the matrix.
+#
+# Limited to Velox + x86 to keep the matrix simple, per the pipeline's purpose
+# of validating Gluten changes against the latest Delta release.
+
+name: Delta Spark UT (Gluten)
+
+on:
+  workflow_dispatch:
+    inputs:
+      delta_ref:
+        description: 'delta-io/delta git ref (tag/branch/SHA) to test against'
+        required: true
+        default: 'v4.2.0'
+      spark_version:
+        description: 'Delta `-DsparkVersion` value (must match the Gluten -P 
profile below)'
+        required: true
+        default: '4.1'
+      test_parallelism:
+        description: 'Forked test JVMs per shard (TEST_PARALLELISM_COUNT)'
+        required: true
+        default: '1'
+      update_baseline:
+        description: 'Seed/refresh the known-failures baseline instead of 
enforcing it'
+        type: boolean
+        required: false
+        default: false
+      fail_on_fixed:
+        description: 'Fail when a baseline test now passes (keeps the baseline 
honest)'
+        type: boolean
+        required: false
+        default: true
+  pull_request:
+    paths:
+      - '.github/workflows/delta_spark_ut.yml'
+      - '.github/workflows/util/delta-spark-ut/**'
+      - 'gluten-delta/**'
+      - 'backends-velox/src-delta40/**/DeltaSQLCommandTest.scala'
+
+env:
+  ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
+  MVN_CMD: 'build/mvn -ntp'
+  CCACHE_DIR: "${{ github.workspace }}/.ccache"
+  # Gluten profile / bundle naming for the build-gluten-bundle and
+  # delta-spark-test jobs. Spark 4.1 + Scala 2.13 + JDK 17 matches Delta 
v4.2.0's
+  # default Spark version (4.1.0) from project/CrossSparkVersions.scala.
+  GLUTEN_SPARK_PROFILE: 'spark-4.1'
+  GLUTEN_SCALA_PROFILE: 'scala-2.13'
+  GLUTEN_JAVA_PROFILE: 'java-17'
+  GLUTEN_BUNDLE_SPARK_VERSION: '4.1'
+  GLUTEN_BUNDLE_SCALA_VERSION: '2.13'
+  # Default values used when the workflow is triggered by pull_request
+  # (where `inputs.*` is empty). Keep these in sync with the workflow_dispatch
+  # defaults above.
+  DELTA_REF_DEFAULT: 'v4.2.0'
+  DELTA_SPARK_VERSION_DEFAULT: '4.1'
+  DELTA_TEST_PARALLELISM_DEFAULT: '1'
+  # Default mode for pull_request runs (where inputs.* is empty): enforce the
+  # committed baseline and fail when a baseline test starts passing. Override
+  # via the workflow_dispatch inputs above.
+  DELTA_UPDATE_BASELINE_DEFAULT: 'false'
+  DELTA_FAIL_ON_FIXED_DEFAULT: 'true'
+  DELTA_SCALA_VERSION: '2.13.16'
+  # Number of shards in the delta-spark-test matrix. Must equal the length of
+  # the `shard` matrix below.
+  #
+  # EXPERIMENT: temporarily set to 4 (production value is 16) to measure the
+  # wall-clock vs concurrent-runner tradeoff. Sharding is by SUITE, so the 
total
+  # test work (~1250 shard-minutes, measured) is fixed regardless of shard 
count
+  # -- more shards = shorter wall-clock but more concurrent runners; fewer 
shards
+  # = longer per-shard time at the same total runner-hours. Measured: 16 shards
+  # ~= 135 min wall-clock (max shard ~110 min). 4 shards carry ~1/4 of the 
suites
+  # each (~300+ min/shard), hence the raised job timeout below. (8-way 
previously
+  # hit the timeout on the slow merge/CDC/DV suites, which is why prod is 16.)
+  # TEST_PARALLELISM_COUNT=1 because each forked test JVM uses ~4G (2G heap +
+  # 2G off-heap) and >1, atop the 4G sbt launcher, would OOM the runner.
+  DELTA_NUM_SHARDS: '4'
+

Review Comment:
   The workflow is configured to run with 4 shards (`DELTA_NUM_SHARDS: '4'` and 
`shard: [0, 1, 2, 3]`), but the PR description and the header comments describe 
sharding across 16 shards. Please align the implementation and documentation 
(either restore 16 shards in the workflow, or update the PR description/header 
comments so expectations match what will run in CI).



##########
.github/workflows/util/delta-spark-ut/setup-delta.sh:
##########
@@ -0,0 +1,177 @@
+#!/usr/bin/env bash
+
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+#
+# Prepares a delta-io/delta clone for running its `spark` module tests with the
+# Gluten (Velox) bundle jar on the classpath.
+#
+# Usage:
+#   setup-delta.sh <delta_ref> <delta_dir> <gluten_bundle_jar> 
<gluten_repo_root>
+#
+# Arguments:
+#   delta_ref           - git ref (tag/branch/sha) to check out (e.g. v4.2.0)
+#   delta_dir           - destination directory for the Delta clone
+#   gluten_bundle_jar   - path to the gluten-velox-bundle fat jar
+#   gluten_repo_root    - path to the Gluten repository root (used to locate
+#                         
backends-velox/src-delta40/.../DeltaSQLCommandTest.scala)
+#
+
+set -euo pipefail
+
+if [ "$#" -ne 4 ]; then
+  echo "Usage: $0 <delta_ref> <delta_dir> <gluten_bundle_jar> 
<gluten_repo_root>" >&2
+  exit 1
+fi
+
+DELTA_REF="$1"
+DELTA_DIR="$2"
+GLUTEN_BUNDLE_JAR="$3"
+GLUTEN_ROOT="$4"
+
+if [ ! -f "$GLUTEN_BUNDLE_JAR" ]; then
+  echo "Gluten bundle jar not found: $GLUTEN_BUNDLE_JAR" >&2
+  exit 1
+fi
+
+# Reuse the existing DeltaSQLCommandTest from Gluten's backends-velox module
+# rather than maintaining a separate copy. This file is compiled as part of the
+# unified `spark` project's Test scope, which has the Gluten bundle on its
+# classpath (via spark-unified/lib/), so the typed GlutenConfig / 
VeloxDeltaConfig
+# imports resolve correctly.
+PATCH_SOURCE="$GLUTEN_ROOT/backends-velox/src-delta40/test/scala/org/apache/spark/sql/delta/test/DeltaSQLCommandTest.scala"
+if [ ! -f "$PATCH_SOURCE" ]; then
+  echo "Gluten DeltaSQLCommandTest not found: $PATCH_SOURCE" >&2
+  exit 1
+fi
+
+echo "::group::Cloning delta-io/delta @ ${DELTA_REF}"
+# Shallow clone the requested tag/branch. Fall back to full clone when the ref 
is a SHA.
+if ! git clone --depth 1 --branch "$DELTA_REF" 
https://github.com/delta-io/delta.git "$DELTA_DIR"; then
+  echo "Shallow clone of ref '${DELTA_REF}' failed, falling back to full 
clone."
+  rm -rf "$DELTA_DIR"
+  git clone https://github.com/delta-io/delta.git "$DELTA_DIR"
+  git -C "$DELTA_DIR" checkout "$DELTA_REF"
+fi

Review Comment:
   The fallback path after a failed shallow clone runs `rm -rf "$DELTA_DIR"` 
without any safety checks. If the caller accidentally passes an empty value or 
`/`, this would delete an unintended path. Add a simple guard before `rm -rf` 
to refuse unsafe values.



-- 
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]

Reply via email to