Fix lazy quantifier priority handling in Glushkov regex engine

.github/workflows/build.yaml

@@ -471,3 +471,73 @@ jobs:
     with:
       push: true
       cuda: '["12.9", "13.3"]'
+  # Build the cuDF Java JAR for every Maven classifier.
+  java-build:
+    needs: [telemetry-setup]
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - { cuda: "12.9", cuda_major: "12", arch: "x86_64", runner: "linux-amd64-cpu16" }
+          - { cuda: "13.3", cuda_major: "13", arch: "x86_64", runner: "linux-amd64-cpu16" }
+          - { cuda: "12.9", cuda_major: "12", arch: "aarch64", runner: "linux-arm64-cpu16" }
+          - { cuda: "13.3", cuda_major: "13", arch: "aarch64", runner: "linux-arm64-cpu16" }
+    runs-on: ${{ matrix.runner }}
+    permissions:
+      contents: read
+    steps:
+      - name: Checkout code repo
+        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4.3.1
+        with:
+          ref: ${{ inputs.sha }}
+          fetch-depth: 0
+          persist-credentials: false
+      - name: Build static libcudf
+        run: |
+          ./java/ci/build_static_libcudf.sh \
+            --output-dir "${RUNNER_TEMP}/libcudf" \
+            --cuda-version "${{ matrix.cuda }}"
+      - name: Build cuDF Java JAR
+        run: |
+          ./java/ci/build_cudf_java_jar.sh \
+            --libcudf-dir "${RUNNER_TEMP}/libcudf" \
+            --output-dir "${RUNNER_TEMP}/jars" \
+            --cuda-version "${{ matrix.cuda }}"
+      - name: Upload per-entry JAR artifact
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
+        with:
+          name: cudf_java_cuda${{ matrix.cuda_major }}_${{ matrix.arch }}
+          # Ship only the JARs and POMs. Exclude the per-classifier Maven build scratch dir.
+          path: |
+            ${{ runner.temp }}/jars
+            !${{ runner.temp }}/jars/.mvn-temp-target
+          if-no-files-found: error
+  # Assemble the per-classifier JARs into one Maven-repository layout.
+  java-gather:
+    needs: [java-build]
+    runs-on: linux-amd64-cpu4
+    permissions:
+      contents: read
+    steps:
+      - name: Checkout code repo
+        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4.3.1
+        with:
+          ref: ${{ inputs.sha }}
+          persist-credentials: false
+      - name: Download per-entry JAR artifacts
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
+        with:
+          pattern: cudf_java_cuda*
+          path: ${{ runner.temp }}/jars
+          merge-multiple: true
+      - name: Assemble Maven repository layout
+        run: |
+          ./java/ci/assemble_maven_repo.sh \
+            --jars-dir "${RUNNER_TEMP}/jars" \
+            --output-dir "${RUNNER_TEMP}/maven-repo"
+      - name: Upload combined Maven repository artifact
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
+        with:
+          name: cudf_java_maven_repo
+          path: ${{ runner.temp }}/maven-repo
+          if-no-files-found: error

ci/run_cudf_polars_polars_tests.sh

@@ -4,6 +4,8 @@
 
 set -euo pipefail
 
+TIMEOUT_TOOL_PATH="$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/timeout_with_stack.py
+
 # Support invoking run_cudf_polars_pytests.sh outside the script directory
 # Assumption, polars has been cloned in the root of the repo.
 cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../polars/
@@ -60,16 +62,18 @@ DESELECTED_TESTS_STR=$(printf -- " --deselect %s" "${DESELECTED_TESTS[@]}")
 # Don't quote the `DESELECTED_...` variable because `pytest` can't handle
 # multiple quoted arguments inline
 # shellcheck disable=SC2086
+# Fail fast (-x) because failed tests pollute the state
 echo "Run polars tests with injected in-memory GPU engine"
-python -m pytest \
+python "${TIMEOUT_TOOL_PATH}" --enable-python 3600 \
+   python -m pytest \
        --import-mode=importlib \
        --cache-clear \
+       -x \
        -m "" \
        -p cudf_polars.testing.inject_gpu_engine \
        -n 4 \
        --dist=worksteal \
        --tb=native \
-       --timeout=240 \
        --durations 10 --durations-min 10 \
        -ra \
        $DESELECTED_TESTS_STR \
@@ -81,17 +85,18 @@ python -m pytest \
 echo "Run polars tests with injected SPMD GPU engine, small blocksize"
 CUDF_POLARS__EXECUTOR__TARGET_PARTITION_SIZE=805306368 \
 CUDF_POLARS__EXECUTOR__FALLBACK_MODE=silent \
-    python -m pytest \
+python "${TIMEOUT_TOOL_PATH}" --enable-python 3600 \
+   python -m pytest \
        --import-mode=importlib \
        --cache-clear \
+       -x \
        -v \
        -m "" \
        -p cudf_polars.testing.inject_gpu_engine \
        -W ignore::ResourceWarning \
        -n 4 \
        --dist=worksteal \
        --tb=native \
-       --timeout=240 \
        --durations 10 --durations-min 10 \
        -ra \
        $DESELECTED_TESTS_STR \

ci/run_cudf_polars_pytests.sh

@@ -1,11 +1,14 @@
 #!/bin/bash
-# SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 set -euo pipefail
 
 # It is essential to cd into python/cudf_polars as `pytest-xdist` + `coverage` seem to work only at this directory level.
 # Support invoking run_cudf_polars_pytests.sh outside the script directory
+TIMEOUT_TOOL_PATH="$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/timeout_with_stack.py
+
 cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../python/cudf_polars/
 
-python -m pytest --cache-clear "$@" tests
+python "${TIMEOUT_TOOL_PATH}" --enable-python 3600 \
+       python -m pytest --cache-clear "$@" tests

ci/test_cuml_compat.sh

@@ -23,9 +23,7 @@ rapids-logger "pytest cuml cuDF-compat subset"
 
 timeout 15m python -m pytest \
   --cache-clear \
-  "${CUML_TESTS_DIR}/test_array.py" \
   "${CUML_TESTS_DIR}/test_compose.py" \
-  "${CUML_TESTS_DIR}/test_input_utils.py" \
   "${CUML_TESTS_DIR}/test_kneighbors_classifier.py" \
   "${CUML_TESTS_DIR}/test_kneighbors_regressor.py" \
   "${CUML_TESTS_DIR}/test_label_encoder.py" \

ci/test_narwhals.sh

@@ -61,8 +61,13 @@ test_get_level \
 "
 
 # test_dtypes: narwhals' dtype mapping changed with polars 1.40 (reports Object where the test expects Int8).
+# test_len_over_2369: cudf now routes transform("size") to the group-level size
+#   (https://github.com/rapidsai/cudf/issues/18491 is fixed), so narwhals' own
+#   xfail for the cudf constructor is stale and the test strict-XPASSes; deselect
+#   until narwhals removes the xfail.
 TESTS_THAT_NEED_NARWHALS_FIX_FOR_CUDF=" \
 test_dtypes or \
+test_len_over_2369[cudf] or \
 test_to_numpy[cudf] or \
 test_fill_null_strategies_with_limit_as_none[cudf] or \
 test_fill_null_series_limit_as_none[cudf] or \
@@ -163,7 +168,6 @@ rapids-logger "Run narwhals tests for cuDF Pandas"
 # test_is_finite_expr & test_is_finite_series: https://github.com/rapidsai/cudf/issues/18257
 # test_maybe_convert_dtypes_pandas: https://github.com/rapidsai/cudf/issues/14149
 # test_log_dtype_pandas: cudf is promoting the type to float64
-# test_len_over_2369: It fails during fallback. The error is 'DataFrame' object has no attribute 'to_frame'
 # test_all_ignore_nulls, test_allh_kleene, and test_anyh_kleene: https://github.com/rapidsai/cudf/issues/19417
 # test_offset_by_date_pandas: https://github.com/rapidsai/cudf/issues/19418
 # test_select_boolean_cols and test_select_boolean_cols_multi_group_by: https://github.com/rapidsai/cudf/issues/19421
@@ -173,7 +177,6 @@ test_is_finite_expr or \
 test_is_finite_series or \
 test_maybe_convert_dtypes_pandas or \
 test_log_dtype_pandas or \
-test_len_over_2369 or \
 test_all_ignore_nulls or \
 test_allh_kleene or \
 test_anyh_kleene or \

ci/test_wheel_cudf_polars.sh

@@ -89,12 +89,14 @@ for version in "${VERSIONS[@]}"; do
         COVERAGE_ARGS=(--no-cov)
     fi
 
+    # Fail fast (-x) rather than trying to continue because failed tests pollute the state
     ./ci/run_cudf_polars_pytests.sh \
         -vv \
         "${COVERAGE_ARGS[@]}" \
         --numprocesses=4 \
         --dist=worksteal \
         --durations 10 --durations-min 10 \
+        -x \
         -ra \
         --junitxml="${RAPIDS_TESTS_DIR}/junit-cudf-polars-${version}.xml"
 
@@ -106,6 +108,7 @@ for version in "${VERSIONS[@]}"; do
         EXITCODE=1
         FAILED+=("${version}")
         rapids-logger "Tests failed for polars ${version}.*"
+        break
     else
         PASSED+=("${version}")
         rapids-logger "Tests passed for polars ${version}.*"