Merge branch 'main' into rparolin/tma_feature

Author: rparolin

.github/RELEASE-core.md

@@ -123,41 +123,21 @@ Pushing the tag triggers a CI run automatically. Monitor it in the
 
 ## Upload wheels to PyPI
 
-This is a two-stage process: first publish to TestPyPI, verify, then
-publish to PyPI.
-
-### Stage 1: TestPyPI
+This is a single `CI: Release` workflow run with two sequential stages:
+publish to TestPyPI, then publish the same wheel set to PyPI.
 
 1. Go to **Actions > CI: Release** and run the workflow with:
    - **Component**: `cuda-core`
    - **The release git tag**: `cuda-core-v0.6.0`
-   - **The GHA run ID that generated validated artifacts**: This is the
-     run ID of the successful tag-triggered CI run from the previous step.
-     You can find it in the URL when viewing the run in the Actions tab
-     (e.g. `https://github.com/NVIDIA/cuda-python/actions/runs/123456789`
-     — the run ID is `123456789`).
-   - **build-ctk-ver**: the `cuda.build.version` from
-     [`ci/versions.yml`](../ci/versions.yml) (e.g. `13.1.1`)
-   - **Which wheel index to publish to**: `testpypi`
-
-2. Wait for the workflow to complete.
-
-3. Verify the TestPyPI upload by installing and running tests from a
-   checked-out copy of the repository:
-
-   ```bash
-   pip install -i https://test.pypi.org/simple/ \
-       --extra-index-url https://pypi.org/simple/ \
-       cuda-core==0.6.0
-   cd cuda_core/tests && pytest
-   ```
 
-### Stage 2: PyPI
+   The workflow automatically looks up the successful tag-triggered CI run
+   for the selected release tag.
 
-Once TestPyPI verification passes, rerun the same workflow with:
-- **Which wheel index to publish to**: `pypi`
+2. Wait for the workflow to complete. It will:
+   - publish the selected wheels to TestPyPI
+   - publish the same wheel set to PyPI
 
-After completion, verify:
+3. After completion, verify the final PyPI upload:
 
 ```bash
 pip install cuda-core==0.6.0

.github/copy-pr-bot.yaml

@@ -5,6 +5,6 @@
 # https://docs.gha-runners.nvidia.com/apps/copy-pr-bot/
 
 enabled: true
-# always require manual CI triggering, ignoring signed commits
+# auto-sync ready PRs so CI starts immediately; keep draft PRs manual
 auto_sync_draft: false
-auto_sync_ready: false
+auto_sync_ready: true

.github/workflows/build-docs.yml

@@ -7,9 +7,6 @@ name: "CI: Build and update docs"
 on:
   workflow_call:
     inputs:
-      build-ctk-ver:
-        type: string
-        required: true
       component:
         description: "Component(s) to build docs for"
         required: false
@@ -47,18 +44,21 @@ jobs:
       run:
         shell: bash -el {0}
     steps:
-      - name: validate build-ctk
-        run: |
-          if [ ! "${{ inputs.build-ctk-ver }}" =~ ^[0-9]+\.[0-9]+\.[0-9]+$ ]; then
-            echo "error: `build-ctk-ver` ${{ inputs.build-ctk-ver }} version does not match MAJOR.MINOR.MICRO" >&2
-            exit 1
-          fi
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
         with:
           fetch-depth: 0
           ref: ${{ inputs.git-tag }}
 
+      - name: Read build CTK version
+        run: |
+          BUILD_CTK_VER=$(yq '.cuda.build.version' ci/versions.yml)
+          if [[ ! "${BUILD_CTK_VER}" =~ ^[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
+            echo "error: derived CTK build version ${BUILD_CTK_VER} does not match MAJOR.MINOR.MICRO" >&2
+            exit 1
+          fi
+          echo "BUILD_CTK_VER=${BUILD_CTK_VER}" >> "$GITHUB_ENV"
+
       # TODO: This workflow runs on GH-hosted runner and cannot use the proxy cache
 
       - name: Set up miniforge
@@ -82,7 +82,7 @@ jobs:
         uses: ./.github/actions/fetch_ctk
         with:
           host-platform: linux-64
-          cuda-version: ${{ inputs.build-ctk-ver }}
+          cuda-version: ${{ env.BUILD_CTK_VER }}
 
       - name: Set environment variables
         run: |
@@ -103,7 +103,7 @@ jobs:
           echo "CUDA_CORE_ARTIFACT_BASENAME=${CUDA_CORE_ARTIFACT_BASENAME}" >> $GITHUB_ENV
           echo "CUDA_CORE_ARTIFACT_NAME=${CUDA_CORE_ARTIFACT_BASENAME}-${FILE_HASH}" >> $GITHUB_ENV
           echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_core/dist")" >> $GITHUB_ENV
-          CUDA_BINDINGS_ARTIFACT_BASENAME="cuda-bindings-python${PYTHON_VERSION_FORMATTED}-cuda${{ inputs.build-ctk-ver }}-linux-64"
+          CUDA_BINDINGS_ARTIFACT_BASENAME="cuda-bindings-python${PYTHON_VERSION_FORMATTED}-cuda${BUILD_CTK_VER}-linux-64"
           echo "CUDA_BINDINGS_ARTIFACT_BASENAME=${CUDA_BINDINGS_ARTIFACT_BASENAME}" >> $GITHUB_ENV
           echo "CUDA_BINDINGS_ARTIFACT_NAME=${CUDA_BINDINGS_ARTIFACT_BASENAME}-${FILE_HASH}" >> $GITHUB_ENV
           echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_bindings/dist")" >> $GITHUB_ENV

.github/workflows/ci.yml

@@ -220,7 +220,6 @@ jobs:
     secrets: inherit
     uses: ./.github/workflows/build-docs.yml
     with:
-      build-ctk-ver: ${{ needs.ci-vars.outputs.CUDA_BUILD_VER }}
       is-release: ${{ github.ref_type == 'tag' }}
 
   checks:

.github/workflows/coverage.yml

@@ -11,14 +11,27 @@ on:
 
 env:
   PY_VER: "3.14"
-  CUDA_VER: "13.1.0"
   LOCAL_CTK: "1"
   GPU: "a100"
   DRIVER: "latest"
 
 jobs:
+  coverage-vars:
+    runs-on: ubuntu-latest
+    outputs:
+      CUDA_VER: ${{ steps.get-vars.outputs.cuda_ver }}
+    steps:
+      - name: Checkout ${{ github.event.repository.name }}
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+      - name: Get CUDA version
+        id: get-vars
+        run: |
+          cuda_ver=$(yq '.cuda.build.version' ci/versions.yml)
+          echo "cuda_ver=$cuda_ver" >> $GITHUB_OUTPUT
+
   coverage-linux:
     name: Coverage (Linux)
+    needs: [coverage-vars]
     runs-on: "linux-amd64-gpu-a100-latest-1"
     permissions:
       id-token: write
@@ -29,6 +42,7 @@ jobs:
     env:
       HOST_PLATFORM: "linux-64"
       ARCH: "x86_64"
+      CUDA_VER: ${{ needs.coverage-vars.outputs.CUDA_VER }}
     # Our self-hosted runners require a container
     # TODO: use a different (nvidia?) container
     container:
@@ -164,6 +178,7 @@ jobs:
   # Build Windows wheels on GitHub-hosted runner (has VS, no GPU)
   build-wheel-windows:
     name: Build Wheels (Windows)
+    needs: [coverage-vars]
     runs-on: windows-2022
     permissions:
       contents: read
@@ -173,6 +188,7 @@ jobs:
     env:
       HOST_PLATFORM: "win-64"
       CUDA_PYTHON_COVERAGE: "1"
+      CUDA_VER: ${{ needs.coverage-vars.outputs.CUDA_VER }}
     steps:
       - name: Checkout ${{ github.event.repository.name }}
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
@@ -231,7 +247,7 @@ jobs:
   # Run coverage tests on self-hosted GPU runner (no VS needed, installs pre-built wheels)
   coverage-windows:
     name: Coverage (Windows)
-    needs: [build-wheel-windows]
+    needs: [coverage-vars, build-wheel-windows]
     runs-on: "windows-amd64-gpu-a100-latest-1"
     permissions:
       id-token: write
@@ -240,6 +256,7 @@ jobs:
       HOST_PLATFORM: "win-64"
       ARCH: "amd64"
       CUDA_PYTHON_COVERAGE: "1"
+      CUDA_VER: ${{ needs.coverage-vars.outputs.CUDA_VER }}
     defaults:
       run:
         shell: bash --noprofile --norc -xeuo pipefail {0}

.github/workflows/release-cuda-pathfinder.yml

@@ -40,7 +40,6 @@ jobs:
       tag: ${{ steps.vars.outputs.tag }}
       version: ${{ steps.vars.outputs.version }}
       run-id: ${{ steps.detect-run.outputs.run-id }}
-      ctk-ver: ${{ steps.ctk.outputs.ctk-ver }}
     steps:
       - name: Verify running on default branch
         run: |
@@ -77,12 +76,6 @@ jobs:
             exit 1
           fi
 
-      - name: Read CTK build version
-        id: ctk
-        run: |
-          ctk_ver=$(yq '.cuda.build.version' ci/versions.yml)
-          echo "ctk-ver=${ctk_ver}" >> "$GITHUB_OUTPUT"
-
       - name: Detect CI run ID
         id: detect-run
         env:
@@ -146,7 +139,6 @@ jobs:
     secrets: inherit
     uses: ./.github/workflows/build-docs.yml
     with:
-      build-ctk-ver: ${{ needs.prepare.outputs.ctk-ver }}
       component: cuda-pathfinder
       git-tag: ${{ needs.prepare.outputs.tag }}
       run-id: ${{ needs.prepare.outputs.run-id }}

.github/workflows/release.yml

@@ -4,7 +4,9 @@
 
 name: "CI: Release"
 
-# Manually-triggered release workflow. Creates a release draft if one doesn't exist for the given tag, or uses existing draft.
+# Manually-triggered release workflow. Creates a release draft if one doesn't exist
+# for the given tag, or uses an existing draft, then publishes the selected wheels
+# to TestPyPI followed by PyPI.
 
 on:
   workflow_dispatch:
@@ -28,16 +30,6 @@ on:
         required: false
         type: string
         default: ""
-      build-ctk-ver:
-        type: string
-        required: true
-      wheel-dst:
-        description: "Which wheel index to publish to?"
-        required: true
-        type: choice
-        options:
-          - testpypi
-          - pypi
 
 defaults:
   run:
@@ -60,15 +52,10 @@ jobs:
         env:
           GH_TOKEN: ${{ github.token }}
         run: |
-          if [[ -n "${{ inputs.run-id }}" ]]; then
-            echo "Using provided run ID: ${{ inputs.run-id }}"
-            echo "run-id=${{ inputs.run-id }}" >> $GITHUB_OUTPUT
-          else
-            echo "Auto-detecting successful tag-triggered run ID for tag: ${{ inputs.git-tag }}"
-            RUN_ID=$(./ci/tools/lookup-run-id "${{ inputs.git-tag }}" "${{ github.repository }}")
-            echo "Auto-detected run ID: $RUN_ID"
-            echo "run-id=$RUN_ID" >> $GITHUB_OUTPUT
-          fi
+          echo "Auto-detecting successful tag-triggered run ID for tag: ${{ inputs.git-tag }}"
+          RUN_ID=$(./ci/tools/lookup-run-id "${{ inputs.git-tag }}" "${{ github.repository }}")
+          echo "Auto-detected run ID: $RUN_ID"
+          echo "run-id=$RUN_ID" >> "$GITHUB_OUTPUT"
 
   check-tag:
     runs-on: ubuntu-latest
@@ -82,17 +69,11 @@ jobs:
         env:
           GH_TOKEN: ${{ github.token }}
         run: |
-          tags=
-          for i in $(gh release list -R ${{ github.repository }} --json tagName --jq '.[]| .tagName'); do
-            tags+=( $i )
-          done
-          is_draft=
-          for i in $(gh release list -R ${{ github.repository }} --json isDraft --jq '.[]| .isDraft'); do
-            is_draft+=( $i )
-          done
+          mapfile -t tags < <(gh release list -R "${{ github.repository }}" --json tagName --jq '.[] | .tagName')
+          mapfile -t is_draft < <(gh release list -R "${{ github.repository }}" --json isDraft --jq '.[] | .isDraft')
 
           found=0
-          for idx in ${!tags[@]}; do
+          for idx in "${!tags[@]}"; do
             if [[ "${tags[$idx]}" == "${{ inputs.git-tag }}" ]]; then
               echo "found existing release for ${{ inputs.git-tag }}"
               found=1
@@ -122,7 +103,6 @@ jobs:
     secrets: inherit
     uses: ./.github/workflows/build-docs.yml
     with:
-      build-ctk-ver: ${{ inputs.build-ctk-ver }}
       component: ${{ inputs.component }}
       git-tag: ${{ inputs.git-tag }}
       run-id: ${{ needs.determine-run-id.outputs.run-id }}
@@ -143,16 +123,16 @@ jobs:
       run-id: ${{ needs.determine-run-id.outputs.run-id }}
       component: ${{ inputs.component }}
 
-  publish-wheels:
-    name: Publish wheels
+  publish-testpypi:
+    name: Publish wheels to TestPyPI
     runs-on: ubuntu-latest
     needs:
       - check-tag
       - determine-run-id
       - doc
     environment:
-      name: ${{ inputs.wheel-dst }}
-      url: https://${{ (inputs.wheel-dst == 'testpypi' && 'test.') || '' }}pypi.org/p/${{ inputs.component }}/
+      name: testpypi
+      url: https://test.pypi.org/${{ inputs.component != 'all' && format('p/{0}/', inputs.component) || '' }}
     permissions:
       id-token: write
     steps:
@@ -169,14 +149,37 @@ jobs:
         run: |
           ./ci/tools/validate-release-wheels "${{ inputs.git-tag }}" "${{ inputs.component }}" "dist"
 
-      - name: Publish package distributions to PyPI
-        if: ${{ inputs.wheel-dst == 'pypi' }}
-        uses: pypa/gh-action-pypi-publish@ed0c53931b1dc9bd32cbe73a98c7f6766f8a527e  # v1.13.0
-
       - name: Publish package distributions to TestPyPI
-        if: ${{ inputs.wheel-dst == 'testpypi' }}
         uses: pypa/gh-action-pypi-publish@ed0c53931b1dc9bd32cbe73a98c7f6766f8a527e  # v1.13.0
         with:
           repository-url: https://test.pypi.org/legacy/
 
+  publish-pypi:
+    name: Publish wheels to PyPI
+    runs-on: ubuntu-latest
+    needs:
+      - determine-run-id
+      - publish-testpypi
+    environment:
+      name: pypi
+      url: https://pypi.org/${{ inputs.component != 'all' && format('p/{0}/', inputs.component) || '' }}
+    permissions:
+      id-token: write
+    steps:
+      - name: Checkout Source
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+
+      - name: Download component wheels
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          ./ci/tools/download-wheels "${{ needs.determine-run-id.outputs.run-id }}" "${{ inputs.component }}" "${{ github.repository }}" "dist"
+
+      - name: Validate wheel versions for release tag
+        run: |
+          ./ci/tools/validate-release-wheels "${{ inputs.git-tag }}" "${{ inputs.component }}" "dist"
+
+      - name: Publish package distributions to PyPI
+        uses: pypa/gh-action-pypi-publish@ed0c53931b1dc9bd32cbe73a98c7f6766f8a527e  # v1.13.0
+
   # TODO: add another job to make the release leave the draft state?

.pre-commit-config.yaml

@@ -30,7 +30,7 @@ repos:
         language: python
         additional_dependencies:
           - https://files.pythonhosted.org/packages/cc/20/ff623b09d963f88bfde16306a54e12ee5ea43e9b597108672ff3a408aad6/pathspec-0.12.1-py3-none-any.whl
-        exclude: '(.*pixi\.lock)|(\.git_archival\.txt)'
+        exclude: '(.*pixi\.lock)|(\.git_archival\.txt)|(.*\.patch$)'
         args: ["--fix"]
 
       - id: no-markdown-in-docs-source
@@ -59,7 +59,13 @@ repos:
       exclude: &gen_exclude '^(?:cuda_python/README\.md|cuda_bindings/cuda/bindings/.*\.in?|cuda_bindings/docs/source/module/.*\.rst?)$'
     - id: mixed-line-ending
     - id: trailing-whitespace
-      exclude: *gen_exclude
+      exclude: |
+        (?x)^(?:
+          cuda_python/README\.md|
+          cuda_bindings/cuda/bindings/.*\.in?|
+          cuda_bindings/docs/source/module/.*\.rst?|
+          .*\.patch$
+        )$
 
   # Checking for common mistakes
   - repo: https://github.com/pre-commit/pygrep-hooks