Merge remote-tracking branch 'origin/main' into ajost/context-safe-buffer-cleanup

Author: Andy-Jost

.github/workflows/build-wheel.yml

@@ -166,7 +166,7 @@ jobs:
           cuda-version: ${{ inputs.cuda-version }}
 
       - name: Build cuda.bindings wheel
-        uses: pypa/cibuildwheel@294735312765b09d24a2fbec22660ce817587d55  # v4.1.0
+        uses: pypa/cibuildwheel@4726cd35bb13f7bde50cf2761f2499ac7b3aa32c  # v4.1.1
         with:
           package-dir: ./cuda_bindings/
           output-dir: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
@@ -234,7 +234,7 @@ jobs:
           if-no-files-found: error
 
       - name: Build cuda.core wheel
-        uses: pypa/cibuildwheel@294735312765b09d24a2fbec22660ce817587d55  # v4.1.0
+        uses: pypa/cibuildwheel@4726cd35bb13f7bde50cf2761f2499ac7b3aa32c  # v4.1.1
         with:
           package-dir: ./cuda_core/
           output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
@@ -338,13 +338,16 @@ jobs:
         id: setup-python2
         uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1  # v6.3.0
         with:
-          # TODO: Pin beta.2 precisely until cibuildwheel catches up (b4 broke ABI for Cython);
-          # this precise pin requires the explicit `freethreaded`.
-          # When 3.15 is officially supported we can also remove the `allow-prereleases` override.
-          python-version: ${{ startsWith(matrix.python-version, '3.15') && '3.15.0-beta.2' || matrix.python-version }}
-          freethreaded: ${{ endsWith(matrix.python-version, 't') }}
+          python-version: ${{ matrix.python-version }}
+          # TODO: remove allow-prereleases once 3.15 is officially supported
           allow-prereleases: ${{ startsWith(matrix.python-version, '3.15') }}
 
+      - name: Enable Scientific Python Nightly Wheels for Python 3.15
+        if: ${{ startsWith(matrix.python-version, '3.15') }}
+        run: |
+          echo "PIP_EXTRA_INDEX_URL=https://pypi.anaconda.org/scientific-python-nightly-wheels/simple" >> "$GITHUB_ENV"
+          echo "PIP_ONLY_BINARY=numpy" >> "$GITHUB_ENV"
+
       - name: verify free-threaded build
         if: endsWith(matrix.python-version, 't')
         run: python -c 'import sys; assert not sys._is_gil_enabled()'
@@ -370,64 +373,6 @@ jobs:
             mv "/c/Program Files/Git/usr/bin/link.exe" "/c/Program Files/Git/usr/bin/link.exe.bak"
           fi
 
-      # TODO: remove the numpy pre-build steps once 3.15 is officially supported
-      # (numpy will publish pre-built 3.15 wheels at that point)
-      - name: Download and patch numpy sdist (pre-release Python)
-        if: ${{ startsWith(matrix.python-version, '3.15') }}
-        run: |
-          pip download --no-binary numpy --no-deps "numpy>=1.21.1" -d numpy-sdist/
-          cd numpy-sdist && tar xf numpy-*.tar.gz && rm numpy-*.tar.gz
-          # WAR: numpy 2.4.x ships [tool.cibuildwheel] config that is
-          # incompatible with cibuildwheel v4.0 (cpython-freethreading enable
-          # group, OpenBLAS before-build scripts, etc.). Strip the cibuildwheel
-          # sections but preserve [tool.meson-python] (vendored meson path).
-          python -c "
-          import glob
-          for f in glob.glob('numpy-*/pyproject.toml'):
-              lines, skip = open(f).readlines(), False
-              out = []
-              for line in lines:
-                  hdr = line.strip()
-                  if hdr.startswith('[tool.cibuildwheel') or hdr.startswith('[[tool.cibuildwheel'):
-                      skip = True
-                      continue
-                  if skip and hdr.startswith('[') and 'cibuildwheel' not in hdr:
-                      skip = False
-                  if not skip:
-                      out.append(line)
-              open(f, 'w').writelines(out)
-          "
-          echo "NUMPY_SRC_DIR=$(pwd)/$(ls -d numpy-*/)" >> $GITHUB_ENV
-
-      - name: Build numpy wheel (pre-release Python)
-        if: ${{ startsWith(matrix.python-version, '3.15') }}
-        uses: pypa/cibuildwheel@294735312765b09d24a2fbec22660ce817587d55  # v4.1.0
-        env:
-          CIBW_BUILD: ${{ env.CIBW_BUILD }}
-          CIBW_SKIP: "*-musllinux* *-win32"
-          CIBW_ARCHS_LINUX: "native"
-          CIBW_BUILD_VERBOSITY: 1
-          CIBW_CONFIG_SETTINGS: "setup-args=-Dallow-noblas=true"
-          CIBW_CONFIG_SETTINGS_WINDOWS: "setup-args=--vsenv setup-args=-Dallow-noblas=true"
-          CIBW_BEFORE_BUILD_WINDOWS: "pip install delvewheel"
-          CIBW_REPAIR_WHEEL_COMMAND_WINDOWS: "delvewheel repair -w {dest_dir} {wheel}"
-          CIBW_ENABLE: "cpython-prerelease"
-        with:
-          package-dir: ${{ env.NUMPY_SRC_DIR }}
-          output-dir: numpy-wheel/
-
-      - name: Upload numpy wheel
-        if: ${{ startsWith(matrix.python-version, '3.15') }}
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a  # v7.0.1
-        with:
-          name: numpy-python${{ env.PYTHON_VERSION_FORMATTED }}-${{ inputs.host-platform }}
-          path: numpy-wheel/*.whl
-          if-no-files-found: error
-
-      - name: Install numpy wheel
-        if: ${{ startsWith(matrix.python-version, '3.15') }}
-        run: pip install numpy-wheel/*.whl
-
       - name: Build cuda.bindings Cython tests
         run: |
           pip install ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl --group ./cuda_bindings/pyproject.toml:test
@@ -508,7 +453,7 @@ jobs:
           rmdir $OLD_BASENAME
 
       - name: Build cuda.core wheel
-        uses: pypa/cibuildwheel@294735312765b09d24a2fbec22660ce817587d55  # v4.1.0
+        uses: pypa/cibuildwheel@4726cd35bb13f7bde50cf2761f2499ac7b3aa32c  # v4.1.1
         with:
           package-dir: ./cuda_core/
           output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}

.github/workflows/ci.yml

@@ -255,11 +255,11 @@ jobs:
           GH_TOKEN: ${{ github.token }}
         run: |
           # --paginate fetches all pages; jq outputs one name per line per page;
-          # head -1 takes the first (newest) match since GitHub returns tags
-          # newest-first. Fails hard if no cuda-core-v* tag is found.
+          # sed prints the first (newest) match while consuming all pages, so
+          # gh can complete without SIGPIPE. Fails if no cuda-core-v* tag is found.
           tag="$(gh api "repos/$GITHUB_REPOSITORY/tags" --paginate \
             --jq '.[] | select(.name | startswith("cuda-core-v")) | .name' \
-            | head -1)"
+            | sed -n '1p')"
           if [[ -z "${tag}" ]]; then
             echo "::error::No cuda-core-v* tag found in the repository." >&2
             exit 1
@@ -501,6 +501,38 @@ jobs:
     with:
       is-release: ${{ github.ref_type == 'tag' }}
 
+  precommit-windows:
+    name: Pre-commit on Windows
+    runs-on: windows-latest
+    if: ${{ github.repository_owner == 'nvidia' && !fromJSON(needs.should-skip.outputs.skip) }}
+    needs:
+      - should-skip
+    permissions:
+      contents: read
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
+        with:
+          fetch-depth: 1
+          persist-credentials: false
+
+      - name: Set up Python
+        uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1  # v6.3.0
+        with:
+          python-version: '3.13'
+
+      - name: Install pre-commit
+        shell: bash
+        run: |
+          set -euxo pipefail
+          python -m pip install --upgrade pip pre-commit
+
+      - name: Run pre-commit
+        shell: bash
+        run: |
+          set -euxo pipefail
+          SKIP=lychee pre-commit run --all-files
+
   checks:
     name: Check job status
     if: always()
@@ -514,6 +546,7 @@ jobs:
       - test-linux-aarch64
       - test-windows
       - doc
+      - precommit-windows
     steps:
       - name: Exit
         run: |
@@ -546,6 +579,7 @@ jobs:
           check_result "should-skip"    "success" "${{ needs.should-skip.result }}"
           check_result "detect-changes" "success" "${{ needs.detect-changes.result }}"
           check_result "doc"            "success" "${{ needs.doc.result }}"
+          check_result "precommit-windows" "success" "${{ needs.precommit-windows.result }}"
 
           # [doc-only] flips these from 'success' to 'skipped'
           if [[ "$doc_only" == "true" ]]; then expected="skipped"; else expected="success"; fi

.github/workflows/test-wheel-linux.yml

@@ -296,16 +296,19 @@ jobs:
       - name: Set up Python ${{ matrix.PY_VER }}
         uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1  # v6.3.0
         with:
-          # TODO: Pin beta.2 precisely until cibuildwheel catches up (b4 broke ABI for Cython)
-          # this precise pin requires the explicit `freethreaded`.
-          # When 3.15 is officially supported we can also remove the `allow-prereleases` override.
-          python-version: ${{ startsWith(matrix.PY_VER, '3.15') && '3.15.0-beta.2' || matrix.PY_VER }}
-          freethreaded: ${{ endsWith(matrix.PY_VER, 't') }}
+          python-version: ${{ matrix.PY_VER }}
+          # TODO: remove allow-prereleases once 3.15 is officially supported
           allow-prereleases: ${{ startsWith(matrix.PY_VER, '3.15') }}
         env:
           # we use self-hosted runners on which setup-python behaves weirdly (Python include can't be found)...
           AGENT_TOOLSDIRECTORY: "/opt/hostedtoolcache"
 
+      - name: Enable Scientific Python Nightly Wheels for Python 3.15
+        if: ${{ startsWith(matrix.PY_VER, '3.15') }}
+        run: |
+          echo "PIP_EXTRA_INDEX_URL=https://pypi.anaconda.org/scientific-python-nightly-wheels/simple" >> "$GITHUB_ENV"
+          echo "PIP_ONLY_BINARY=numpy" >> "$GITHUB_ENV"
+
       - name: Set up mini CTK
         if: ${{ matrix.LOCAL_CTK == '1' }}
         uses: ./.github/actions/fetch_ctk
@@ -314,18 +317,6 @@ jobs:
           host-platform: ${{ inputs.host-platform }}
           cuda-version: ${{ matrix.CUDA_VER }}
 
-      # TODO: remove the numpy wheel steps once 3.15 is officially supported
-      - name: Download numpy wheel (pre-release Python)
-        if: ${{ startsWith(matrix.PY_VER, '3.15') }}
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c  # v8.0.1
-        with:
-          name: numpy-python${{ env.PYTHON_VERSION_FORMATTED }}-${{ inputs.host-platform }}
-          path: numpy-wheel
-
-      - name: Install numpy wheel (pre-release Python)
-        if: ${{ startsWith(matrix.PY_VER, '3.15') }}
-        run: pip install numpy-wheel/*.whl
-
       - name: Set up latest cuda_sanitizer_api
         if: ${{ env.SETUP_SANITIZER == '1' }}
         uses: ./.github/actions/fetch_ctk

.github/workflows/test-wheel-windows.yml

@@ -276,13 +276,17 @@ jobs:
       - name: Set up Python ${{ matrix.PY_VER }}
         uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1  # v6.3.0
         with:
-          # TODO: Pin beta.2 precisely until cibuildwheel catches up (b4 broke ABI for Cython)
-          # this precise pin requires the explicit `freethreaded`.
-          # When 3.15 is officially supported we can also remove the `allow-prereleases` override.
-          python-version: ${{ startsWith(matrix.PY_VER, '3.15') && '3.15.0-beta.2' || matrix.PY_VER }}
-          freethreaded: ${{ endsWith(matrix.PY_VER, 't') }}
+          python-version: ${{ matrix.PY_VER }}
+          # TODO: remove allow-prereleases once 3.15 is officially supported
           allow-prereleases: ${{ startsWith(matrix.PY_VER, '3.15') }}
 
+      - name: Enable Scientific Python Nightly Wheels for Python 3.15
+        if: ${{ startsWith(matrix.PY_VER, '3.15') }}
+        shell: bash --noprofile --norc -xeuo pipefail {0}
+        run: |
+          echo "PIP_EXTRA_INDEX_URL=https://pypi.anaconda.org/scientific-python-nightly-wheels/simple" >> "$GITHUB_ENV"
+          echo "PIP_ONLY_BINARY=numpy" >> "$GITHUB_ENV"
+
       - name: Verify LongPathsEnabled
         run: |
           $val = (Get-ItemProperty -Path 'HKLM:\SYSTEM\CurrentControlSet\Control\FileSystem' -Name 'LongPathsEnabled').LongPathsEnabled
@@ -300,19 +304,6 @@ jobs:
           host-platform: ${{ inputs.host-platform }}
           cuda-version: ${{ matrix.CUDA_VER }}
 
-      # TODO: remove the numpy wheel steps once 3.15 is officially supported
-      - name: Download numpy wheel (pre-release Python)
-        if: ${{ startsWith(matrix.PY_VER, '3.15') }}
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c  # v8.0.1
-        with:
-          name: numpy-python${{ env.PYTHON_VERSION_FORMATTED }}-${{ inputs.host-platform }}
-          path: numpy-wheel
-
-      - name: Install numpy wheel (pre-release Python)
-        if: ${{ startsWith(matrix.PY_VER, '3.15') }}
-        shell: bash --noprofile --norc -xeuo pipefail {0}
-        run: pip install numpy-wheel/*.whl
-
       - name: Set up test repetition on nightly runs
         shell: bash --noprofile --norc -xeuo pipefail {0}
         run: echo "PYTEST_ADDOPTS=\"--count=${{ inputs.nruns }}\"" >> "$GITHUB_ENV"

.pre-commit-config.yaml

@@ -69,13 +69,13 @@ repos:
 
       - id: stubgen-pyx-cuda-core
         name: Generate .pyi stubs for cuda_core
-        entry: stubgen-pyx cuda_core/cuda --continue-on-error --include-private
+        entry: python ./toolshed/run_stubgen_pyx.py
         language: python
         files: ^cuda_core/cuda/.*\.(pyx|pxd)$
         pass_filenames: false
         additional_dependencies:
           - stubgen-pyx==0.2.6
-          - Cython==3.2.4
+          - Cython==3.2.9
 
   # Link checking for authored documentation files
   - repo: https://github.com/lycheeverse/lychee
@@ -103,7 +103,7 @@ repos:
     - id: check-yaml
     - id: debug-statements
     - id: end-of-file-fixer
-      exclude: &gen_exclude '^(?:cuda_python/README\.md|cuda_bindings/cuda/bindings/.*\.in?|cuda_bindings/docs/source/module/.*\.rst?|.*\.pyi)$'
+      exclude: &gen_exclude '^(?:cuda_python/README\.md|(?:.*/)?CLAUDE\.md|(?:.*/)?\.git_archival\.txt|cuda_bindings/cuda/bindings/.*\.in?|cuda_bindings/docs/source/module/.*\.rst?|.*\.pyi)$'
     - id: mixed-line-ending
     - id: trailing-whitespace
       exclude: |

CONTRIBUTING.md

@@ -25,6 +25,7 @@ Thank you for your interest in contributing to CUDA Python! Based on the type of
     - [Symptoms of a bad clone](#symptoms-of-a-bad-clone)
   - [Type stubs for cuda.core](#type-stubs-for-cudacore)
   - [Pre-commit](#pre-commit)
+    - [Pre-commit on Windows](#pre-commit-on-windows)
   - [Signing Your Work](#signing-your-work)
   - [Code signing](#code-signing)
   - [Developer Certificate of Origin (DCO)](#developer-certificate-of-origin-dco)
@@ -166,6 +167,17 @@ between commits, leaving stale headers or out-of-date stubs in the history.
 If the hook isn't installed, `pre-commit run` (and CI) will print a visible
 warning reminding you to run `pre-commit install`.
 
+### Pre-commit on Windows
+
+For development on Windows (not WSL), the `lychee` pre-commit task will not work
+when running `pre-commit run --all-files`.  This problem does not occur if you
+install the pre-commit hook and run it automatically as part of your `git
+commit` workflow.  To resolve this, you can either:
+
+1. Run `pre-commit` in Git Bash, rather than directly in PowerShell or cmd
+
+2. Skip it by setting the environment variable `SKIP` to `lychee`.  This would
+   be `$env:SKIP = "lychee"` in PowerShell or `set SKIP=lychee` in cmd.
 
 ## Signing Your Work
 

ci/test-matrix.yml

@@ -128,6 +128,7 @@ windows:
     - { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '12.9.1', LOCAL_CTK: '1', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'TCC' }
     - { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'a100',       GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'MCDM' }
     - { ARCH: 'amd64', PY_VER: '3.14t', CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'a100',       GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'MCDM' }
+    - { ARCH: 'amd64', PY_VER: '3.15',  CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'MCDM' }
     # special runners
     - { ARCH: 'amd64', PY_VER: '3.14',  CUDA_VER: '13.3.0', LOCAL_CTK: '1', GPU: 't4',         GPU_COUNT: '2', DRIVER: 'latest',  DRIVER_MODE: 'TCC' }
     - { ARCH: 'amd64', PY_VER: '3.14',  CUDA_VER: '13.3.0', LOCAL_CTK: '0', GPU: 'h100',       GPU_COUNT: '2', DRIVER: 'latest',  DRIVER_MODE: 'MCDM' }