From e7627e308b43e8143f2a4bdcc5667bc1eca27bee Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 15:57:47 +0000
Subject: [PATCH 01/32] test installing git for windows

---
 .github/workflows/test-wheel-windows.yml | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 95a9cbe6f8..315b5707c2 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -107,6 +107,20 @@ jobs:
           }
           gh --version
 
+      - name: Install Git for Windows
+        # the GPU runner image does not have Git Bash pre-installed...
+        env:
+          # doesn't seem there's an easy way to avoid hard-coding it?
+          GFW_EXE_URL: https://github.com/git-for-windows/git/releases/download/v2.49.0.windows.1/PortableGit-2.49.0-64-bit.7z.exe
+        run: |
+          Invoke-WebRequest -Uri "$env:GFW_EXE_URL" -OutFile "PortableGit.7z.exe"
+          Start-Process 7z -Wait -Verbose -ArgumentList "x PortableGit.7z.exe -oPortableGit"
+          ls -l PortableGit
+          mv PortableGit\git-bash.exe PortableGit\bash.exe
+          echo "$(pwd).path\\PortableGit" >> $env:GITHUB_PATH
+          $env:Path += ";$(pwd).path\\PortableGit"
+          bash --version
+
       - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
         if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
         env:

From 562b558006f84b52f8331a917e3b9027403ed9a9 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 16:02:33 +0000
Subject: [PATCH 02/32] shrink CI

---
 .github/workflows/build-and-test.yml | 150 +++++++++++++--------------
 1 file changed, 75 insertions(+), 75 deletions(-)

diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index a00a940286..d2b04b7890 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -24,15 +24,15 @@ jobs:
       fail-fast: false
       matrix:
         host-platform:
-          - linux-64
-          - linux-aarch64
+#          - linux-64
+#          - linux-aarch64
           - win-64
         python-version:
-          - "3.13"
+#          - "3.13"
           - "3.12"
-          - "3.11"
-          - "3.10"
-          - "3.9"
+#          - "3.11"
+#          - "3.10"
+#          - "3.9"
         cuda-version:
           # Note: this is for build-time only.
           - "12.8.0"
@@ -208,56 +208,56 @@ jobs:
         run: |
           echo "CUDA_VERSION=${{ matrix.cuda-version }}" >> $GITHUB_OUTPUT
 
-  test-linux:
-    strategy:
-      fail-fast: false
-      # TODO: add driver version here
-      matrix:
-        host-platform:
-          - linux-64
-          - linux-aarch64
-        python-version:
-          - "3.13"
-          - "3.12"
-          - "3.11"
-          - "3.10"
-          - "3.9"
-        cuda-version:
-          # Note: this is for test-time only.
-          - "12.8.0"
-          - "12.0.1"
-          - "11.8.0"
-        local-ctk:
-          - 1  # use mini CTK
-          - 0  # use CTK wheels
-        runner:
-          - default
-        exclude:
-          # To test this combo would require nontrivial installation steps.
-          - cuda-version: "12.0.1"
-            local-ctk: 0
-        include:
-          - host-platform: linux-64
-            python-version: "3.12"
-            cuda-version: "12.8.0"
-            local-ctk: 1
-            runner: H100
-    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
-    if: ${{ github.repository_owner == 'nvidia' }}
-    permissions:
-      contents: read  # This is required for actions/checkout
-    needs:
-      - build
-    secrets: inherit
-    uses:
-      ./.github/workflows/test-wheel-linux.yml
-    with:
-      host-platform: ${{ matrix.host-platform }}
-      python-version: ${{ matrix.python-version }}
-      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
-      cuda-version: ${{ matrix.cuda-version }}
-      local-ctk: ${{ matrix.local-ctk}}
-      runner: ${{ matrix.runner }}
+#  test-linux:
+#    strategy:
+#      fail-fast: false
+#      # TODO: add driver version here
+#      matrix:
+#        host-platform:
+#          - linux-64
+#          - linux-aarch64
+#        python-version:
+#          - "3.13"
+#          - "3.12"
+#          - "3.11"
+#          - "3.10"
+#          - "3.9"
+#        cuda-version:
+#          # Note: this is for test-time only.
+#          - "12.8.0"
+#          - "12.0.1"
+#          - "11.8.0"
+#        local-ctk:
+#          - 1  # use mini CTK
+#          - 0  # use CTK wheels
+#        runner:
+#          - default
+#        exclude:
+#          # To test this combo would require nontrivial installation steps.
+#          - cuda-version: "12.0.1"
+#            local-ctk: 0
+#        include:
+#          - host-platform: linux-64
+#            python-version: "3.12"
+#            cuda-version: "12.8.0"
+#            local-ctk: 1
+#            runner: H100
+#    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
+#    if: ${{ github.repository_owner == 'nvidia' }}
+#    permissions:
+#      contents: read  # This is required for actions/checkout
+#    needs:
+#      - build
+#    secrets: inherit
+#    uses:
+#      ./.github/workflows/test-wheel-linux.yml
+#    with:
+#      host-platform: ${{ matrix.host-platform }}
+#      python-version: ${{ matrix.python-version }}
+#      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
+#      cuda-version: ${{ matrix.cuda-version }}
+#      local-ctk: ${{ matrix.local-ctk}}
+#      runner: ${{ matrix.runner }}
 
   test-windows:
     strategy:
@@ -271,10 +271,10 @@ jobs:
         cuda-version:
           # Note: this is for test-time only.
           - "12.8.0"
-          - "11.8.0"
+#          - "11.8.0"
         local-ctk:
           - 1  # use mini CTK
-          - 0  # use CTK wheels
+#          - 0  # use CTK wheels
         runner:
           - default
     name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
@@ -294,21 +294,21 @@ jobs:
       local-ctk: ${{ matrix.local-ctk}}
       runner: ${{ matrix.runner }}
 
-  doc:
-    name: Docs
-    if: ${{ github.repository_owner == 'nvidia' }}
-    # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
-    permissions:
-      id-token: write
-      contents: write
-      pull-requests: write
-    needs:
-      - build
-    secrets: inherit
-    uses:
-      ./.github/workflows/build-docs.yml
-    with:
-      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
+#  doc:
+#    name: Docs
+#    if: ${{ github.repository_owner == 'nvidia' }}
+#    # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
+#    permissions:
+#      id-token: write
+#      contents: write
+#      pull-requests: write
+#    needs:
+#      - build
+#    secrets: inherit
+#    uses:
+#      ./.github/workflows/build-docs.yml
+#    with:
+#      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
 
   checks:
     name: Check job status
@@ -316,9 +316,9 @@ jobs:
       checks: read
     needs:
       - build
-      - test-linux
+#      - test-linux
       - test-windows
-      - doc
+#      - doc
     secrets: inherit
     uses:
       ./.github/workflows/status-check.yml

From c54332b78cfa22877902f5b6ef85337ba5f1bdb6 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 16:33:33 +0000
Subject: [PATCH 03/32] also install 7z...

---
 .github/workflows/test-wheel-windows.yml | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 315b5707c2..441c924f2b 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -37,9 +37,9 @@ jobs:
         with:
           fetch-depth: 0
 
-      - name: Update driver
-        run: |
-          .github/workflows/install_gpu_driver.ps1
+#      - name: Update driver
+#        run: |
+#          .github/workflows/install_gpu_driver.ps1
 
       - name: Ensure GPU is working
         run: nvidia-smi
@@ -107,6 +107,21 @@ jobs:
           }
           gh --version
 
+      - name: Install 7z
+        # the GPU runner image does not have Git Bash pre-installed...
+        env:
+          # doesn't seem there's an easy way to avoid hard-coding it?
+          7Z_URL: https://www.7-zip.org/a/7z2409-x64.msi
+        run: |
+          Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
+          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /qn'
+          $GH_POSSIBLE_PATHS = "C:\\Program Files\\7-Zip", "C:\\Program Files (x86)\\7-Zip"
+          foreach ($p in $GH_POSSIBLE_PATHS) {
+              echo "$p" >> $env:GITHUB_PATH
+              $env:Path += ";$p"
+          }
+          7z --version
+
       - name: Install Git for Windows
         # the GPU runner image does not have Git Bash pre-installed...
         env:

From 270b6816e019bc5ce927632f65fa7fd94c0be766 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 16:37:13 +0000
Subject: [PATCH 04/32] skip cuda.bindings for now

---
 .github/workflows/build-and-test.yml | 92 ++++++++++++++--------------
 1 file changed, 46 insertions(+), 46 deletions(-)

diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index d2b04b7890..a7c82fee03 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -127,52 +127,52 @@ jobs:
           path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
           if-no-files-found: error
 
-      - name: Set up mini CTK
-        uses: ./.github/actions/fetch_ctk
-        continue-on-error: false
-        with:
-          host-platform: ${{ matrix.host-platform }}
-          cuda-version: ${{ matrix.cuda-version }}
-
-      - name: Build cuda.bindings wheel
-        uses: pypa/cibuildwheel@v2.22.0
-        env:
-          CIBW_BUILD: ${{ env.CIBW_BUILD }}
-          CIBW_ARCHS_LINUX: "native"
-          CIBW_BUILD_VERBOSITY: 1
-          # CIBW mounts the host filesystem under /host
-          CIBW_ENVIRONMENT_LINUX: >
-            CUDA_PATH=/host/${{ env.CUDA_PATH }}
-            LIBRARY_PATH=/host/${{ env.CUDA_PATH }}/lib
-            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
-          CIBW_ENVIRONMENT_WINDOWS: >
-            CUDA_HOME="$(cygpath -w ${{ env.CUDA_PATH }})"
-            LIB="${CUDA_HOME}\\lib\\x64;${LIB}"
-            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
-        with:
-          package-dir: ./cuda_bindings/
-          output-dir: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-
-      - name: List the cuda.bindings artifacts directory
-        run: |
-          if [[ "${{ matrix.host-platform }}" == win* ]]; then
-            export CHOWN=chown
-          else
-            export CHOWN="sudo chown"
-          fi
-          $CHOWN -R $(whoami) ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-          ls -lahR ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-
-      - name: Check cuda.bindings wheel
-        run: |
-          twine check ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
-
-      - name: Upload cuda.bindings build artifacts
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
-          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
-          if-no-files-found: error
+#      - name: Set up mini CTK
+#        uses: ./.github/actions/fetch_ctk
+#        continue-on-error: false
+#        with:
+#          host-platform: ${{ matrix.host-platform }}
+#          cuda-version: ${{ matrix.cuda-version }}
+#
+#      - name: Build cuda.bindings wheel
+#        uses: pypa/cibuildwheel@v2.22.0
+#        env:
+#          CIBW_BUILD: ${{ env.CIBW_BUILD }}
+#          CIBW_ARCHS_LINUX: "native"
+#          CIBW_BUILD_VERBOSITY: 1
+#          # CIBW mounts the host filesystem under /host
+#          CIBW_ENVIRONMENT_LINUX: >
+#            CUDA_PATH=/host/${{ env.CUDA_PATH }}
+#            LIBRARY_PATH=/host/${{ env.CUDA_PATH }}/lib
+#            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
+#          CIBW_ENVIRONMENT_WINDOWS: >
+#            CUDA_HOME="$(cygpath -w ${{ env.CUDA_PATH }})"
+#            LIB="${CUDA_HOME}\\lib\\x64;${LIB}"
+#            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
+#        with:
+#          package-dir: ./cuda_bindings/
+#          output-dir: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+#
+#      - name: List the cuda.bindings artifacts directory
+#        run: |
+#          if [[ "${{ matrix.host-platform }}" == win* ]]; then
+#            export CHOWN=chown
+#          else
+#            export CHOWN="sudo chown"
+#          fi
+#          $CHOWN -R $(whoami) ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+#          ls -lahR ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+#
+#      - name: Check cuda.bindings wheel
+#        run: |
+#          twine check ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
+#
+#      - name: Upload cuda.bindings build artifacts
+#        uses: actions/upload-artifact@v4
+#        with:
+#          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
+#          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
+#          if-no-files-found: error
 
       # upload-artifact's "overwrite: true" option has a race condition among parallel
       # jobs, so we let job 0 do the work

From 9e665db5e2355918b19358eb755e3b8419fe595e Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 17:01:27 +0000
Subject: [PATCH 05/32] fix

---
 .github/workflows/test-wheel-windows.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 441c924f2b..25236d6169 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -40,9 +40,9 @@ jobs:
 #      - name: Update driver
 #        run: |
 #          .github/workflows/install_gpu_driver.ps1
-
-      - name: Ensure GPU is working
-        run: nvidia-smi
+#
+#      - name: Ensure GPU is working
+#        run: nvidia-smi
 
       - name: Set environment variables
         run: |

From 38830b20fa2716a26c2f801c9460bad468033f01 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 17:27:58 +0000
Subject: [PATCH 06/32] fix

---
 .github/workflows/test-wheel-windows.yml | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 25236d6169..eb439db845 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -85,12 +85,12 @@ jobs:
           name: cuda-python-wheel
           path: .
 
-      - name: Download cuda.bindings build artifacts
-        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
-        uses: actions/download-artifact@v4
-        with:
-          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
-          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+#      - name: Download cuda.bindings build artifacts
+#        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
+#        uses: actions/download-artifact@v4
+#        with:
+#          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
+#          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
 
       - name: Install gh cli
         # the GPU runner image does not have gh pre-installed...

From 4a2bb6fbf8a4432160f579e26382433b30f0d1f3 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 18:25:36 +0000
Subject: [PATCH 07/32] fix

---
 .github/workflows/test-wheel-windows.yml | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index eb439db845..09c371acbe 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -114,12 +114,10 @@ jobs:
           7Z_URL: https://www.7-zip.org/a/7z2409-x64.msi
         run: |
           Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
-          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /qn'
-          $GH_POSSIBLE_PATHS = "C:\\Program Files\\7-Zip", "C:\\Program Files (x86)\\7-Zip"
-          foreach ($p in $GH_POSSIBLE_PATHS) {
-              echo "$p" >> $env:GITHUB_PATH
-              $env:Path += ";$p"
-          }
+          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR=$((Get-Location).Path)\7-Zip'
+          ls -l 7-Zip
+          echo "$((Get-Location).Path)\7-Zip" >> $env:GITHUB_PATH
+          $env:Path += ";$((Get-Location).Path)\7-Zip"
           7z --version
 
       - name: Install Git for Windows
@@ -132,8 +130,8 @@ jobs:
           Start-Process 7z -Wait -Verbose -ArgumentList "x PortableGit.7z.exe -oPortableGit"
           ls -l PortableGit
           mv PortableGit\git-bash.exe PortableGit\bash.exe
-          echo "$(pwd).path\\PortableGit" >> $env:GITHUB_PATH
-          $env:Path += ";$(pwd).path\\PortableGit"
+          echo "$((Get-Location).Path)\\PortableGit" >> $env:GITHUB_PATH
+          $env:Path += ";$((Get-Location).Path)\\PortableGit"
           bash --version
 
       - name: Download cuda-python & cuda.bindings build artifacts from the prior branch

From 2bb32a939e0ccb81ba9ae5aa4bc4dafe9d352521 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 18:26:58 +0000
Subject: [PATCH 08/32] skip

---
 .github/workflows/build-and-test.yml | 64 ++++++++++++++--------------
 1 file changed, 32 insertions(+), 32 deletions(-)

diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index a7c82fee03..aba5f86401 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -95,38 +95,38 @@ jobs:
         run: |
           env
 
-      - name: Build cuda.core wheel
-        uses: pypa/cibuildwheel@v2.22.0
-        env:
-          CIBW_BUILD: ${{ env.CIBW_BUILD }}
-          CIBW_ARCHS_LINUX: "native"
-          CIBW_BUILD_VERBOSITY: 1
-        with:
-          package-dir: ./cuda_core/
-          output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-
-      - name: List the cuda.core artifacts directory
-        run: |
-          if [[ "${{ matrix.host-platform }}" == win* ]]; then
-            export CHOWN=chown
-          else
-            export CHOWN="sudo chown"
-          fi
-          $CHOWN -R $(whoami) ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-          ls -lahR ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-
-      - name: Check cuda.core wheel
-        run: |
-          pip install twine
-          twine check ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
-
-      - name: Upload cuda.core build artifacts
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
-          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
-          if-no-files-found: error
-
+#      - name: Build cuda.core wheel
+#        uses: pypa/cibuildwheel@v2.22.0
+#        env:
+#          CIBW_BUILD: ${{ env.CIBW_BUILD }}
+#          CIBW_ARCHS_LINUX: "native"
+#          CIBW_BUILD_VERBOSITY: 1
+#        with:
+#          package-dir: ./cuda_core/
+#          output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+#
+#      - name: List the cuda.core artifacts directory
+#        run: |
+#          if [[ "${{ matrix.host-platform }}" == win* ]]; then
+#            export CHOWN=chown
+#          else
+#            export CHOWN="sudo chown"
+#          fi
+#          $CHOWN -R $(whoami) ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+#          ls -lahR ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+#
+#      - name: Check cuda.core wheel
+#        run: |
+#          pip install twine
+#          twine check ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
+#
+#      - name: Upload cuda.core build artifacts
+#        uses: actions/upload-artifact@v4
+#        with:
+#          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
+#          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
+#          if-no-files-found: error
+#
 #      - name: Set up mini CTK
 #        uses: ./.github/actions/fetch_ctk
 #        continue-on-error: false

From dc633dcf04e0cf9002ed3c0e6fa50c4d0cb73b37 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 18:29:19 +0000
Subject: [PATCH 09/32] skip

---
 .github/workflows/build-and-test.yml | 58 ++++++++++++++--------------
 1 file changed, 29 insertions(+), 29 deletions(-)

diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index aba5f86401..698208fbee 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -173,35 +173,35 @@ jobs:
 #          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
 #          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
 #          if-no-files-found: error
-
-      # upload-artifact's "overwrite: true" option has a race condition among parallel
-      # jobs, so we let job 0 do the work
-      - name: Build and check cuda-python wheel
-        if: ${{ strategy.job-index == 0 }}
-        run: |
-          pushd cuda_python
-          pip wheel -v --no-deps .
-          twine check *.whl
-          popd
-
-      - name: List the cuda-python artifacts directory
-        if: ${{ strategy.job-index == 0 }}
-        run: |
-          if [[ "${{ matrix.host-platform }}" == win* ]]; then
-            export CHOWN=chown
-          else
-            export CHOWN="sudo chown"
-          fi
-          $CHOWN -R $(whoami) cuda_python/*.whl
-          ls -lahR cuda_python
-
-      - name: Upload cuda-python build artifacts
-        if: ${{ strategy.job-index == 0 }}
-        uses: actions/upload-artifact@v4
-        with:
-          name: cuda-python-wheel
-          path: cuda_python/*.whl
-          if-no-files-found: error
+#
+#      # upload-artifact's "overwrite: true" option has a race condition among parallel
+#      # jobs, so we let job 0 do the work
+#      - name: Build and check cuda-python wheel
+#        if: ${{ strategy.job-index == 0 }}
+#        run: |
+#          pushd cuda_python
+#          pip wheel -v --no-deps .
+#          twine check *.whl
+#          popd
+#
+#      - name: List the cuda-python artifacts directory
+#        if: ${{ strategy.job-index == 0 }}
+#        run: |
+#          if [[ "${{ matrix.host-platform }}" == win* ]]; then
+#            export CHOWN=chown
+#          else
+#            export CHOWN="sudo chown"
+#          fi
+#          $CHOWN -R $(whoami) cuda_python/*.whl
+#          ls -lahR cuda_python
+#
+#      - name: Upload cuda-python build artifacts
+#        if: ${{ strategy.job-index == 0 }}
+#        uses: actions/upload-artifact@v4
+#        with:
+#          name: cuda-python-wheel
+#          path: cuda_python/*.whl
+#          if-no-files-found: error
 
       - name: Pass environment variables to the next runner
         id: pass_env

From cd69de39639ca3e77feef0a42a7aa6b887af7921 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 18:32:11 +0000
Subject: [PATCH 10/32] more skip

---
 .github/workflows/test-wheel-windows.yml | 44 ++++++++++++------------
 1 file changed, 22 insertions(+), 22 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 09c371acbe..396cbaa4b8 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -78,34 +78,34 @@ jobs:
           "SKIP_CUDA_BINDINGS_TEST=${SKIP_CUDA_BINDINGS_TEST}" >> $env:GITHUB_ENV
           "MINI_CTK_DEPS=${MINI_CTK_DEPS}" >> $env:GITHUB_ENV
 
-      - name: Download cuda-python build artifacts
-        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
-        uses: actions/download-artifact@v4
-        with:
-          name: cuda-python-wheel
-          path: .
-
+#      - name: Download cuda-python build artifacts
+#        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
+#        uses: actions/download-artifact@v4
+#        with:
+#          name: cuda-python-wheel
+#          path: .
+#
 #      - name: Download cuda.bindings build artifacts
 #        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
 #        uses: actions/download-artifact@v4
 #        with:
 #          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
 #          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-
-      - name: Install gh cli
-        # the GPU runner image does not have gh pre-installed...
-        env:
-          # doesn't seem there's an easy way to avoid hard-coding it?
-          GH_MSI_URL: https://github.com/cli/cli/releases/download/v2.67.0/gh_2.67.0_windows_amd64.msi
-        run: |
-          Invoke-WebRequest -Uri "$env:GH_MSI_URL" -OutFile "gh_installer.msi"
-          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "gh_installer.msi" /qn'
-          $GH_POSSIBLE_PATHS = "C:\\Program Files\\GitHub CLI", "C:\\Program Files (x86)\\GitHub CLI"
-          foreach ($p in $GH_POSSIBLE_PATHS) {
-              echo "$p" >> $env:GITHUB_PATH
-              $env:Path += ";$p"
-          }
-          gh --version
+#
+#      - name: Install gh cli
+#        # the GPU runner image does not have gh pre-installed...
+#        env:
+#          # doesn't seem there's an easy way to avoid hard-coding it?
+#          GH_MSI_URL: https://github.com/cli/cli/releases/download/v2.67.0/gh_2.67.0_windows_amd64.msi
+#        run: |
+#          Invoke-WebRequest -Uri "$env:GH_MSI_URL" -OutFile "gh_installer.msi"
+#          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "gh_installer.msi" /qn'
+#          $GH_POSSIBLE_PATHS = "C:\\Program Files\\GitHub CLI", "C:\\Program Files (x86)\\GitHub CLI"
+#          foreach ($p in $GH_POSSIBLE_PATHS) {
+#              echo "$p" >> $env:GITHUB_PATH
+#              $env:Path += ";$p"
+#          }
+#          gh --version
 
       - name: Install 7z
         # the GPU runner image does not have Git Bash pre-installed...

From 52adef23e96012ec11e31bf2672ca7483e6c7a44 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 18:37:28 +0000
Subject: [PATCH 11/32] add quotes

---
 .github/workflows/test-wheel-windows.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 396cbaa4b8..7089411d04 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -114,7 +114,7 @@ jobs:
           7Z_URL: https://www.7-zip.org/a/7z2409-x64.msi
         run: |
           Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
-          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR=$((Get-Location).Path)\7-Zip'
+          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="$((Get-Location).Path)\7-Zip"'
           ls -l 7-Zip
           echo "$((Get-Location).Path)\7-Zip" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\7-Zip"

From d5c7be2f6943d840ec85a602ff1e2fc30ef0e869 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 18:41:02 +0000
Subject: [PATCH 12/32] try this

---
 .github/workflows/test-wheel-windows.yml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 7089411d04..0cfe51cd81 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -115,7 +115,8 @@ jobs:
         run: |
           Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
           Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="$((Get-Location).Path)\7-Zip"'
-          ls -l 7-Zip
+          #ls -l 7-Zip
+          Get-Location
           echo "$((Get-Location).Path)\7-Zip" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\7-Zip"
           7z --version

From efd6036128aeea829d88d47519a56ba73be68783 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 18:50:26 +0000
Subject: [PATCH 13/32] try this

---
 .github/workflows/test-wheel-windows.yml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 0cfe51cd81..8884c2e0b9 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -116,7 +116,8 @@ jobs:
           Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
           Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="$((Get-Location).Path)\7-Zip"'
           #ls -l 7-Zip
-          Get-Location
+          #Get-Location
+          ls
           echo "$((Get-Location).Path)\7-Zip" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\7-Zip"
           7z --version

From 1c8d7a8d162d3cd565f2afdad251b62bb40be5af Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 19:13:25 +0000
Subject: [PATCH 14/32] try this

---
 .github/workflows/test-wheel-windows.yml | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 8884c2e0b9..48eb22a09c 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -114,12 +114,10 @@ jobs:
           7Z_URL: https://www.7-zip.org/a/7z2409-x64.msi
         run: |
           Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
-          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="$((Get-Location).Path)\7-Zip"'
-          #ls -l 7-Zip
-          #Get-Location
-          ls
-          echo "$((Get-Location).Path)\7-Zip" >> $env:GITHUB_PATH
-          $env:Path += ";$((Get-Location).Path)\7-Zip"
+          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="C:\Program Files\7-Zip" /qn'
+          ls -l "C:\\Program Files\7-Zip"
+          echo "C:\\Program Files\7-Zip" >> $env:GITHUB_PATH
+          $env:Path += ";C:\\Program Files\7-Zip"
           7z --version
 
       - name: Install Git for Windows

From 4cb69d65e9a270676faee25ec4a168c2c875fcdc Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 19:31:43 +0000
Subject: [PATCH 15/32] don't escape?

---
 .github/workflows/test-wheel-windows.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 48eb22a09c..23e227ab3b 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -115,9 +115,9 @@ jobs:
         run: |
           Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
           Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="C:\Program Files\7-Zip" /qn'
-          ls -l "C:\\Program Files\7-Zip"
-          echo "C:\\Program Files\7-Zip" >> $env:GITHUB_PATH
-          $env:Path += ";C:\\Program Files\7-Zip"
+          ls -l "C:\Program Files\7-Zip"
+          echo "C:\Program Files\7-Zip" >> $env:GITHUB_PATH
+          $env:Path += ";C:\Program Files\7-Zip"
           7z --version
 
       - name: Install Git for Windows

From 10de33c20b262a8f46f4e78406d1fd0e7756f95d Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 19:38:03 +0000
Subject: [PATCH 16/32] try this

---
 .github/workflows/test-wheel-windows.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 23e227ab3b..e7e6be8032 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -118,6 +118,7 @@ jobs:
           ls -l "C:\Program Files\7-Zip"
           echo "C:\Program Files\7-Zip" >> $env:GITHUB_PATH
           $env:Path += ";C:\Program Files\7-Zip"
+          echo $env:Path
           7z --version
 
       - name: Install Git for Windows

From 3cfc4e3442feb3e76476fe07353d7ec79e3d988c Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 19:42:10 +0000
Subject: [PATCH 17/32] --version does not exist

---
 .github/workflows/test-wheel-windows.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index e7e6be8032..94ce657b77 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -119,7 +119,7 @@ jobs:
           echo "C:\Program Files\7-Zip" >> $env:GITHUB_PATH
           $env:Path += ";C:\Program Files\7-Zip"
           echo $env:Path
-          7z --version
+          7z
 
       - name: Install Git for Windows
         # the GPU runner image does not have Git Bash pre-installed...

From fffe85c9500c4393a4f89b017387a6f19687bf01 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 19:49:20 +0000
Subject: [PATCH 18/32] switch to fetch_ctk

---
 .github/workflows/test-wheel-windows.yml | 110 ++++++++++++-----------
 1 file changed, 59 insertions(+), 51 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 94ce657b77..a3b94dcf78 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -135,66 +135,74 @@ jobs:
           $env:Path += ";$((Get-Location).Path)\\PortableGit"
           bash --version
 
-      - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
-        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          $OLD_BRANCH = Get-Content .github/BACKPORT_BRANCH
-          $OLD_BASENAME = "cuda-bindings-python${env:PYTHON_VERSION_FORMATTED}-cuda*-${{ inputs.host-platform }}*"
-          $runData = gh run list -b $OLD_BRANCH -L 1 -w "CI: Build and test" -s completed -R NVIDIA/cuda-python --json databaseId | ConvertFrom-Json
-          if (-not $runData -or $runData.Length -eq 0 -or -not $runData[0].databaseId -or [string]::IsNullOrEmpty($runData[0].databaseId)) {
-              Write-Host "LATEST_PRIOR_RUN_ID not found!"
-              exit 1
-          }
-          $LATEST_PRIOR_RUN_ID = $runData[0].databaseId
-
-          gh run download $LATEST_PRIOR_RUN_ID -p $OLD_BASENAME -R NVIDIA/cuda-python
-          Get-ChildItem -Path $OLD_BASENAME
-          New-Item -Path "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}" -ItemType Directory -Force
-          Move-Item -Path "$OLD_BASENAME/*.whl" -Destination "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"
-          Remove-Item -Path $OLD_BASENAME -Force
-      
-          gh run download $LATEST_PRIOR_RUN_ID -p cuda-python-wheel -R NVIDIA/cuda-python
-          Get-ChildItem -Path cuda-python-wheel
-          Move-Item -Path "cuda-python-wheel/*.whl" -Destination .
-          Remove-Item -Path cuda-python-wheel -Force
-
-      - name: Display structure of downloaded cuda-python artifacts
-        run: |
-          Get-Location
-          Get-ChildItem -Recurse -Force | Select-Object Mode, LastWriteTime, Length, FullName
-
-      - name: Display structure of downloaded cuda.bindings artifacts
-        run: |
-          Get-Location
-          Get-ChildItem -Recurse -Force $env:CUDA_BINDINGS_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
-
-      - name: Download cuda.core build artifacts
-        uses: actions/download-artifact@v4
-        with:
-          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
-          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-
-      - name: Display structure of downloaded cuda.core build artifacts
-        run: |
-          Get-Location
-          Get-ChildItem -Recurse -Force $env:CUDA_CORE_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
+#      - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
+#        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
+#        env:
+#          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+#        run: |
+#          $OLD_BRANCH = Get-Content .github/BACKPORT_BRANCH
+#          $OLD_BASENAME = "cuda-bindings-python${env:PYTHON_VERSION_FORMATTED}-cuda*-${{ inputs.host-platform }}*"
+#          $runData = gh run list -b $OLD_BRANCH -L 1 -w "CI: Build and test" -s completed -R NVIDIA/cuda-python --json databaseId | ConvertFrom-Json
+#          if (-not $runData -or $runData.Length -eq 0 -or -not $runData[0].databaseId -or [string]::IsNullOrEmpty($runData[0].databaseId)) {
+#              Write-Host "LATEST_PRIOR_RUN_ID not found!"
+#              exit 1
+#          }
+#          $LATEST_PRIOR_RUN_ID = $runData[0].databaseId
+#
+#          gh run download $LATEST_PRIOR_RUN_ID -p $OLD_BASENAME -R NVIDIA/cuda-python
+#          Get-ChildItem -Path $OLD_BASENAME
+#          New-Item -Path "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}" -ItemType Directory -Force
+#          Move-Item -Path "$OLD_BASENAME/*.whl" -Destination "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"
+#          Remove-Item -Path $OLD_BASENAME -Force
+#      
+#          gh run download $LATEST_PRIOR_RUN_ID -p cuda-python-wheel -R NVIDIA/cuda-python
+#          Get-ChildItem -Path cuda-python-wheel
+#          Move-Item -Path "cuda-python-wheel/*.whl" -Destination .
+#          Remove-Item -Path cuda-python-wheel -Force
+#
+#      - name: Display structure of downloaded cuda-python artifacts
+#        run: |
+#          Get-Location
+#          Get-ChildItem -Recurse -Force | Select-Object Mode, LastWriteTime, Length, FullName
+#
+#      - name: Display structure of downloaded cuda.bindings artifacts
+#        run: |
+#          Get-Location
+#          Get-ChildItem -Recurse -Force $env:CUDA_BINDINGS_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
+#
+#      - name: Download cuda.core build artifacts
+#        uses: actions/download-artifact@v4
+#        with:
+#          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
+#          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+#
+#      - name: Display structure of downloaded cuda.core build artifacts
+#        run: |
+#          Get-Location
+#          Get-ChildItem -Recurse -Force $env:CUDA_CORE_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
 
       - name: Set up Python ${{ inputs.python-version }}
         uses: actions/setup-python@v5
         with:
           python-version: ${{ inputs.python-version }}
 
+#      - name: Set up mini CTK
+#        if: ${{ inputs.local-ctk == '1' }}
+#        # Note: The GH-hosted Windows GPU runner does not have Git for Windows pre-installed,
+#        # so we cannot use our own fetch_ctk action unfortunately...
+#        uses: Jimver/cuda-toolkit@v0.2.21
+#        with:
+#          cuda: ${{ inputs.cuda-version }}
+#          method: 'network'
+#          sub-packages: ${{ env.MINI_CTK_DEPS }}
+
       - name: Set up mini CTK
         if: ${{ inputs.local-ctk == '1' }}
-        # Note: The GH-hosted Windows GPU runner does not have Git for Windows pre-installed,
-        # so we cannot use our own fetch_ctk action unfortunately...
-        uses: Jimver/cuda-toolkit@v0.2.21
+        uses: ./.github/actions/fetch_ctk
+        continue-on-error: false
         with:
-          cuda: ${{ inputs.cuda-version }}
-          method: 'network'
-          sub-packages: ${{ env.MINI_CTK_DEPS }}
+          host-platform: ${{ inputs.host-platform }}
+          cuda-version: ${{ inputs.cuda-version }}
 
       - name: Update PATH
         if: ${{ inputs.local-ctk == '1' }}

From daf26ade38ec20efa05d4e40f6841dde9f1e58fc Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 20:19:59 +0000
Subject: [PATCH 19/32] try not to hang

---
 .github/actions/fetch_ctk/action.yml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/.github/actions/fetch_ctk/action.yml b/.github/actions/fetch_ctk/action.yml
index ed0b6d2907..17330a5544 100644
--- a/.github/actions/fetch_ctk/action.yml
+++ b/.github/actions/fetch_ctk/action.yml
@@ -19,11 +19,14 @@ inputs:
     type: string
     default: "cuda_nvcc,cuda_cudart,cuda_nvrtc,cuda_profiler_api,cuda_cccl,cuda_sanitizer_api,libnvjitlink"
 
+defaults:
+  run:
+    shell: bash ${{ (inputs.host-platform != 'win-64' && '--noprofile --norc') || '' }} -xeuo pipefail {0}
+
 runs:
   using: composite
   steps:
     - name: Set up CTK cache variable
-      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         HASH=$(echo -n "${{ inputs.cuda-components }}" | sha256sum | awk '{print $1}')
         echo "CTK_CACHE_KEY=mini-ctk-${{ inputs.cuda-version }}-${{ inputs.host-platform }}-$HASH" >> $GITHUB_ENV
@@ -48,7 +51,6 @@ runs:
 
     - name: Get CUDA components
       if: ${{ steps.ctk-get-cache.outputs.cache-hit != 'true' }}
-      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         CUDA_PATH="./cuda_toolkit"
         mkdir $CUDA_PATH
@@ -126,7 +128,6 @@ runs:
 
     - name: Restore CTK cache
       if: ${{ steps.ctk-get-cache.outputs.cache-hit == 'true' }}
-      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         ls -l
         CUDA_PATH="./cuda_toolkit"
@@ -137,7 +138,6 @@ runs:
         fi
 
     - name: Set output environment variables
-      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         CUDA_PATH=$(realpath "./cuda_toolkit")
         echo "CUDA_PATH=${CUDA_PATH}" >> $GITHUB_ENV

From 0a578befe13b66f7c6f2bdc2dc877ea1dee24ef4 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 20:26:25 +0000
Subject: [PATCH 20/32] try not to hang

---
 .github/actions/fetch_ctk/action.yml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/.github/actions/fetch_ctk/action.yml b/.github/actions/fetch_ctk/action.yml
index 17330a5544..36622d24ff 100644
--- a/.github/actions/fetch_ctk/action.yml
+++ b/.github/actions/fetch_ctk/action.yml
@@ -19,12 +19,11 @@ inputs:
     type: string
     default: "cuda_nvcc,cuda_cudart,cuda_nvrtc,cuda_profiler_api,cuda_cccl,cuda_sanitizer_api,libnvjitlink"
 
-defaults:
-  run:
-    shell: bash ${{ (inputs.host-platform != 'win-64' && '--noprofile --norc') || '' }} -xeuo pipefail {0}
-
 runs:
   using: composite
+  defaults:
+    run:
+      shell: bash ${{ (inputs.host-platform != 'win-64' && '--noprofile --norc') || '' }} -xeuo pipefail {0}
   steps:
     - name: Set up CTK cache variable
       run: |

From f3fa59640cf93e4ed1bc2f3af93dc8da5ddff359 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 2 May 2025 20:31:58 +0000
Subject: [PATCH 21/32] try not to hang; composite action does not support
 setting default shell...

---
 .github/actions/fetch_ctk/action.yml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/.github/actions/fetch_ctk/action.yml b/.github/actions/fetch_ctk/action.yml
index 36622d24ff..7c2ab91a3f 100644
--- a/.github/actions/fetch_ctk/action.yml
+++ b/.github/actions/fetch_ctk/action.yml
@@ -21,11 +21,9 @@ inputs:
 
 runs:
   using: composite
-  defaults:
-    run:
-      shell: bash ${{ (inputs.host-platform != 'win-64' && '--noprofile --norc') || '' }} -xeuo pipefail {0}
   steps:
     - name: Set up CTK cache variable
+      shell: bash -xeuo pipefail {0}
       run: |
         HASH=$(echo -n "${{ inputs.cuda-components }}" | sha256sum | awk '{print $1}')
         echo "CTK_CACHE_KEY=mini-ctk-${{ inputs.cuda-version }}-${{ inputs.host-platform }}-$HASH" >> $GITHUB_ENV
@@ -50,6 +48,7 @@ runs:
 
     - name: Get CUDA components
       if: ${{ steps.ctk-get-cache.outputs.cache-hit != 'true' }}
+      shell: bash -xeuo pipefail {0}
       run: |
         CUDA_PATH="./cuda_toolkit"
         mkdir $CUDA_PATH
@@ -127,6 +126,7 @@ runs:
 
     - name: Restore CTK cache
       if: ${{ steps.ctk-get-cache.outputs.cache-hit == 'true' }}
+      shell: bash -xeuo pipefail {0}
       run: |
         ls -l
         CUDA_PATH="./cuda_toolkit"
@@ -137,6 +137,7 @@ runs:
         fi
 
     - name: Set output environment variables
+      shell: bash -xeuo pipefail {0}
       run: |
         CUDA_PATH=$(realpath "./cuda_toolkit")
         echo "CUDA_PATH=${CUDA_PATH}" >> $GITHUB_ENV

From 0418045d4dbcdbeae304ee59803c3d25a3f7a4df Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Fri, 2 May 2025 22:48:09 -0400
Subject: [PATCH 22/32] self extracting + run post install

---
 .github/workflows/test-wheel-windows.yml | 31 ++++++++++++------------
 1 file changed, 16 insertions(+), 15 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index a3b94dcf78..26656e30a0 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -107,19 +107,19 @@ jobs:
 #          }
 #          gh --version
 
-      - name: Install 7z
-        # the GPU runner image does not have Git Bash pre-installed...
-        env:
-          # doesn't seem there's an easy way to avoid hard-coding it?
-          7Z_URL: https://www.7-zip.org/a/7z2409-x64.msi
-        run: |
-          Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
-          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="C:\Program Files\7-Zip" /qn'
-          ls -l "C:\Program Files\7-Zip"
-          echo "C:\Program Files\7-Zip" >> $env:GITHUB_PATH
-          $env:Path += ";C:\Program Files\7-Zip"
-          echo $env:Path
-          7z
+#      - name: Install 7z
+#        # the GPU runner image does not have Git Bash pre-installed...
+#        env:
+#          # doesn't seem there's an easy way to avoid hard-coding it?
+#          7Z_URL: https://www.7-zip.org/a/7z2409-x64.msi
+#        run: |
+#          Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
+#          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="C:\Program Files\7-Zip" /qn'
+#          ls -l "C:\Program Files\7-Zip"
+#          echo "C:\Program Files\7-Zip" >> $env:GITHUB_PATH
+#          $env:Path += ";C:\Program Files\7-Zip"
+#          echo $env:Path
+#          7z
 
       - name: Install Git for Windows
         # the GPU runner image does not have Git Bash pre-installed...
@@ -128,9 +128,10 @@ jobs:
           GFW_EXE_URL: https://github.com/git-for-windows/git/releases/download/v2.49.0.windows.1/PortableGit-2.49.0-64-bit.7z.exe
         run: |
           Invoke-WebRequest -Uri "$env:GFW_EXE_URL" -OutFile "PortableGit.7z.exe"
-          Start-Process 7z -Wait -Verbose -ArgumentList "x PortableGit.7z.exe -oPortableGit"
+          .\PortableGit.7z.exe -y -gm2 -InstallPath=".\\PortableGit"
           ls -l PortableGit
-          mv PortableGit\git-bash.exe PortableGit\bash.exe
+          .\PortableGit\post-install.bat
+          New-Item -Path PortableGit\bash.exe -ItemType SymbolicLink -Value PortableGit\git-bash.exe
           echo "$((Get-Location).Path)\\PortableGit" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\\PortableGit"
           bash --version

From 869ff140c0a67a3ac8ef263839628ceecafab419 Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Fri, 2 May 2025 23:06:16 -0400
Subject: [PATCH 23/32] block extraction to ensure file visibility

---
 .github/workflows/test-wheel-windows.yml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 26656e30a0..a24484bb47 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -128,9 +128,8 @@ jobs:
           GFW_EXE_URL: https://github.com/git-for-windows/git/releases/download/v2.49.0.windows.1/PortableGit-2.49.0-64-bit.7z.exe
         run: |
           Invoke-WebRequest -Uri "$env:GFW_EXE_URL" -OutFile "PortableGit.7z.exe"
-          .\PortableGit.7z.exe -y -gm2 -InstallPath=".\\PortableGit"
+          Start-Process .\PortableGit.7z.exe -Wait -Verbose -ArgumentList '-y -gm2'
           ls -l PortableGit
-          .\PortableGit\post-install.bat
           New-Item -Path PortableGit\bash.exe -ItemType SymbolicLink -Value PortableGit\git-bash.exe
           echo "$((Get-Location).Path)\\PortableGit" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\\PortableGit"

From ffa239ebb1a951980660489ec1bdcc3a19796873 Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Fri, 2 May 2025 23:38:49 -0400
Subject: [PATCH 24/32] clean up

---
 .github/workflows/build-and-test.yml     | 364 +++++++++++------------
 .github/workflows/test-wheel-windows.yml | 190 +++++-------
 2 files changed, 261 insertions(+), 293 deletions(-)

diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index 698208fbee..a00a940286 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -24,15 +24,15 @@ jobs:
       fail-fast: false
       matrix:
         host-platform:
-#          - linux-64
-#          - linux-aarch64
+          - linux-64
+          - linux-aarch64
           - win-64
         python-version:
-#          - "3.13"
+          - "3.13"
           - "3.12"
-#          - "3.11"
-#          - "3.10"
-#          - "3.9"
+          - "3.11"
+          - "3.10"
+          - "3.9"
         cuda-version:
           # Note: this is for build-time only.
           - "12.8.0"
@@ -95,169 +95,169 @@ jobs:
         run: |
           env
 
-#      - name: Build cuda.core wheel
-#        uses: pypa/cibuildwheel@v2.22.0
-#        env:
-#          CIBW_BUILD: ${{ env.CIBW_BUILD }}
-#          CIBW_ARCHS_LINUX: "native"
-#          CIBW_BUILD_VERBOSITY: 1
-#        with:
-#          package-dir: ./cuda_core/
-#          output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-#
-#      - name: List the cuda.core artifacts directory
-#        run: |
-#          if [[ "${{ matrix.host-platform }}" == win* ]]; then
-#            export CHOWN=chown
-#          else
-#            export CHOWN="sudo chown"
-#          fi
-#          $CHOWN -R $(whoami) ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-#          ls -lahR ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-#
-#      - name: Check cuda.core wheel
-#        run: |
-#          pip install twine
-#          twine check ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
-#
-#      - name: Upload cuda.core build artifacts
-#        uses: actions/upload-artifact@v4
-#        with:
-#          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
-#          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
-#          if-no-files-found: error
-#
-#      - name: Set up mini CTK
-#        uses: ./.github/actions/fetch_ctk
-#        continue-on-error: false
-#        with:
-#          host-platform: ${{ matrix.host-platform }}
-#          cuda-version: ${{ matrix.cuda-version }}
-#
-#      - name: Build cuda.bindings wheel
-#        uses: pypa/cibuildwheel@v2.22.0
-#        env:
-#          CIBW_BUILD: ${{ env.CIBW_BUILD }}
-#          CIBW_ARCHS_LINUX: "native"
-#          CIBW_BUILD_VERBOSITY: 1
-#          # CIBW mounts the host filesystem under /host
-#          CIBW_ENVIRONMENT_LINUX: >
-#            CUDA_PATH=/host/${{ env.CUDA_PATH }}
-#            LIBRARY_PATH=/host/${{ env.CUDA_PATH }}/lib
-#            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
-#          CIBW_ENVIRONMENT_WINDOWS: >
-#            CUDA_HOME="$(cygpath -w ${{ env.CUDA_PATH }})"
-#            LIB="${CUDA_HOME}\\lib\\x64;${LIB}"
-#            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
-#        with:
-#          package-dir: ./cuda_bindings/
-#          output-dir: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-#
-#      - name: List the cuda.bindings artifacts directory
-#        run: |
-#          if [[ "${{ matrix.host-platform }}" == win* ]]; then
-#            export CHOWN=chown
-#          else
-#            export CHOWN="sudo chown"
-#          fi
-#          $CHOWN -R $(whoami) ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-#          ls -lahR ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-#
-#      - name: Check cuda.bindings wheel
-#        run: |
-#          twine check ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
-#
-#      - name: Upload cuda.bindings build artifacts
-#        uses: actions/upload-artifact@v4
-#        with:
-#          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
-#          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
-#          if-no-files-found: error
-#
-#      # upload-artifact's "overwrite: true" option has a race condition among parallel
-#      # jobs, so we let job 0 do the work
-#      - name: Build and check cuda-python wheel
-#        if: ${{ strategy.job-index == 0 }}
-#        run: |
-#          pushd cuda_python
-#          pip wheel -v --no-deps .
-#          twine check *.whl
-#          popd
-#
-#      - name: List the cuda-python artifacts directory
-#        if: ${{ strategy.job-index == 0 }}
-#        run: |
-#          if [[ "${{ matrix.host-platform }}" == win* ]]; then
-#            export CHOWN=chown
-#          else
-#            export CHOWN="sudo chown"
-#          fi
-#          $CHOWN -R $(whoami) cuda_python/*.whl
-#          ls -lahR cuda_python
-#
-#      - name: Upload cuda-python build artifacts
-#        if: ${{ strategy.job-index == 0 }}
-#        uses: actions/upload-artifact@v4
-#        with:
-#          name: cuda-python-wheel
-#          path: cuda_python/*.whl
-#          if-no-files-found: error
+      - name: Build cuda.core wheel
+        uses: pypa/cibuildwheel@v2.22.0
+        env:
+          CIBW_BUILD: ${{ env.CIBW_BUILD }}
+          CIBW_ARCHS_LINUX: "native"
+          CIBW_BUILD_VERBOSITY: 1
+        with:
+          package-dir: ./cuda_core/
+          output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+
+      - name: List the cuda.core artifacts directory
+        run: |
+          if [[ "${{ matrix.host-platform }}" == win* ]]; then
+            export CHOWN=chown
+          else
+            export CHOWN="sudo chown"
+          fi
+          $CHOWN -R $(whoami) ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+          ls -lahR ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+
+      - name: Check cuda.core wheel
+        run: |
+          pip install twine
+          twine check ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
+
+      - name: Upload cuda.core build artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
+          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
+          if-no-files-found: error
+
+      - name: Set up mini CTK
+        uses: ./.github/actions/fetch_ctk
+        continue-on-error: false
+        with:
+          host-platform: ${{ matrix.host-platform }}
+          cuda-version: ${{ matrix.cuda-version }}
+
+      - name: Build cuda.bindings wheel
+        uses: pypa/cibuildwheel@v2.22.0
+        env:
+          CIBW_BUILD: ${{ env.CIBW_BUILD }}
+          CIBW_ARCHS_LINUX: "native"
+          CIBW_BUILD_VERBOSITY: 1
+          # CIBW mounts the host filesystem under /host
+          CIBW_ENVIRONMENT_LINUX: >
+            CUDA_PATH=/host/${{ env.CUDA_PATH }}
+            LIBRARY_PATH=/host/${{ env.CUDA_PATH }}/lib
+            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
+          CIBW_ENVIRONMENT_WINDOWS: >
+            CUDA_HOME="$(cygpath -w ${{ env.CUDA_PATH }})"
+            LIB="${CUDA_HOME}\\lib\\x64;${LIB}"
+            CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
+        with:
+          package-dir: ./cuda_bindings/
+          output-dir: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+
+      - name: List the cuda.bindings artifacts directory
+        run: |
+          if [[ "${{ matrix.host-platform }}" == win* ]]; then
+            export CHOWN=chown
+          else
+            export CHOWN="sudo chown"
+          fi
+          $CHOWN -R $(whoami) ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+          ls -lahR ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+
+      - name: Check cuda.bindings wheel
+        run: |
+          twine check ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
+
+      - name: Upload cuda.bindings build artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
+          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
+          if-no-files-found: error
+
+      # upload-artifact's "overwrite: true" option has a race condition among parallel
+      # jobs, so we let job 0 do the work
+      - name: Build and check cuda-python wheel
+        if: ${{ strategy.job-index == 0 }}
+        run: |
+          pushd cuda_python
+          pip wheel -v --no-deps .
+          twine check *.whl
+          popd
+
+      - name: List the cuda-python artifacts directory
+        if: ${{ strategy.job-index == 0 }}
+        run: |
+          if [[ "${{ matrix.host-platform }}" == win* ]]; then
+            export CHOWN=chown
+          else
+            export CHOWN="sudo chown"
+          fi
+          $CHOWN -R $(whoami) cuda_python/*.whl
+          ls -lahR cuda_python
+
+      - name: Upload cuda-python build artifacts
+        if: ${{ strategy.job-index == 0 }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: cuda-python-wheel
+          path: cuda_python/*.whl
+          if-no-files-found: error
 
       - name: Pass environment variables to the next runner
         id: pass_env
         run: |
           echo "CUDA_VERSION=${{ matrix.cuda-version }}" >> $GITHUB_OUTPUT
 
-#  test-linux:
-#    strategy:
-#      fail-fast: false
-#      # TODO: add driver version here
-#      matrix:
-#        host-platform:
-#          - linux-64
-#          - linux-aarch64
-#        python-version:
-#          - "3.13"
-#          - "3.12"
-#          - "3.11"
-#          - "3.10"
-#          - "3.9"
-#        cuda-version:
-#          # Note: this is for test-time only.
-#          - "12.8.0"
-#          - "12.0.1"
-#          - "11.8.0"
-#        local-ctk:
-#          - 1  # use mini CTK
-#          - 0  # use CTK wheels
-#        runner:
-#          - default
-#        exclude:
-#          # To test this combo would require nontrivial installation steps.
-#          - cuda-version: "12.0.1"
-#            local-ctk: 0
-#        include:
-#          - host-platform: linux-64
-#            python-version: "3.12"
-#            cuda-version: "12.8.0"
-#            local-ctk: 1
-#            runner: H100
-#    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
-#    if: ${{ github.repository_owner == 'nvidia' }}
-#    permissions:
-#      contents: read  # This is required for actions/checkout
-#    needs:
-#      - build
-#    secrets: inherit
-#    uses:
-#      ./.github/workflows/test-wheel-linux.yml
-#    with:
-#      host-platform: ${{ matrix.host-platform }}
-#      python-version: ${{ matrix.python-version }}
-#      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
-#      cuda-version: ${{ matrix.cuda-version }}
-#      local-ctk: ${{ matrix.local-ctk}}
-#      runner: ${{ matrix.runner }}
+  test-linux:
+    strategy:
+      fail-fast: false
+      # TODO: add driver version here
+      matrix:
+        host-platform:
+          - linux-64
+          - linux-aarch64
+        python-version:
+          - "3.13"
+          - "3.12"
+          - "3.11"
+          - "3.10"
+          - "3.9"
+        cuda-version:
+          # Note: this is for test-time only.
+          - "12.8.0"
+          - "12.0.1"
+          - "11.8.0"
+        local-ctk:
+          - 1  # use mini CTK
+          - 0  # use CTK wheels
+        runner:
+          - default
+        exclude:
+          # To test this combo would require nontrivial installation steps.
+          - cuda-version: "12.0.1"
+            local-ctk: 0
+        include:
+          - host-platform: linux-64
+            python-version: "3.12"
+            cuda-version: "12.8.0"
+            local-ctk: 1
+            runner: H100
+    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
+    if: ${{ github.repository_owner == 'nvidia' }}
+    permissions:
+      contents: read  # This is required for actions/checkout
+    needs:
+      - build
+    secrets: inherit
+    uses:
+      ./.github/workflows/test-wheel-linux.yml
+    with:
+      host-platform: ${{ matrix.host-platform }}
+      python-version: ${{ matrix.python-version }}
+      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
+      cuda-version: ${{ matrix.cuda-version }}
+      local-ctk: ${{ matrix.local-ctk}}
+      runner: ${{ matrix.runner }}
 
   test-windows:
     strategy:
@@ -271,10 +271,10 @@ jobs:
         cuda-version:
           # Note: this is for test-time only.
           - "12.8.0"
-#          - "11.8.0"
+          - "11.8.0"
         local-ctk:
           - 1  # use mini CTK
-#          - 0  # use CTK wheels
+          - 0  # use CTK wheels
         runner:
           - default
     name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
@@ -294,21 +294,21 @@ jobs:
       local-ctk: ${{ matrix.local-ctk}}
       runner: ${{ matrix.runner }}
 
-#  doc:
-#    name: Docs
-#    if: ${{ github.repository_owner == 'nvidia' }}
-#    # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
-#    permissions:
-#      id-token: write
-#      contents: write
-#      pull-requests: write
-#    needs:
-#      - build
-#    secrets: inherit
-#    uses:
-#      ./.github/workflows/build-docs.yml
-#    with:
-#      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
+  doc:
+    name: Docs
+    if: ${{ github.repository_owner == 'nvidia' }}
+    # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
+    permissions:
+      id-token: write
+      contents: write
+      pull-requests: write
+    needs:
+      - build
+    secrets: inherit
+    uses:
+      ./.github/workflows/build-docs.yml
+    with:
+      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
 
   checks:
     name: Check job status
@@ -316,9 +316,9 @@ jobs:
       checks: read
     needs:
       - build
-#      - test-linux
+      - test-linux
       - test-windows
-#      - doc
+      - doc
     secrets: inherit
     uses:
       ./.github/workflows/status-check.yml
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index a24484bb47..347a98a8fa 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -37,12 +37,12 @@ jobs:
         with:
           fetch-depth: 0
 
-#      - name: Update driver
-#        run: |
-#          .github/workflows/install_gpu_driver.ps1
-#
-#      - name: Ensure GPU is working
-#        run: nvidia-smi
+      - name: Update driver
+        run: |
+          .github/workflows/install_gpu_driver.ps1
+
+      - name: Ensure GPU is working
+        run: nvidia-smi
 
       - name: Set environment variables
         run: |
@@ -57,14 +57,6 @@ jobs:
             $SKIP_CUDA_BINDINGS_TEST = 0
           }
 
-          if ('${{ inputs.local-ctk }}' -eq '1') {
-            if ($TEST_CUDA_MAJOR -eq '12') {
-              $MINI_CTK_DEPS = '["nvcc", "nvrtc", "nvjitlink", "thrust"]'
-            } else {
-              $MINI_CTK_DEPS = '["nvcc", "nvrtc", "thrust"]'
-            }
-          }
-
           # Make outputs from the previous job as env vars
           $CUDA_CORE_ARTIFACT_BASENAME = "cuda-core-python${PYTHON_VERSION_FORMATTED}-${{ inputs.host-platform }}"
           "PYTHON_VERSION_FORMATTED=${PYTHON_VERSION_FORMATTED}" >> $env:GITHUB_ENV
@@ -76,50 +68,35 @@ jobs:
           "CUDA_BINDINGS_ARTIFACT_NAME=${CUDA_BINDINGS_ARTIFACT_BASENAME}-${{ github.sha }}" >> $env:GITHUB_ENV
           "CUDA_BINDINGS_ARTIFACTS_DIR=$($ExecutionContext.SessionState.Path.GetUnresolvedProviderPathFromPSPath("$REPO_DIR\cuda_bindings\dist"))" >> $env:GITHUB_ENV
           "SKIP_CUDA_BINDINGS_TEST=${SKIP_CUDA_BINDINGS_TEST}" >> $env:GITHUB_ENV
-          "MINI_CTK_DEPS=${MINI_CTK_DEPS}" >> $env:GITHUB_ENV
 
-#      - name: Download cuda-python build artifacts
-#        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
-#        uses: actions/download-artifact@v4
-#        with:
-#          name: cuda-python-wheel
-#          path: .
-#
-#      - name: Download cuda.bindings build artifacts
-#        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
-#        uses: actions/download-artifact@v4
-#        with:
-#          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
-#          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-#
-#      - name: Install gh cli
-#        # the GPU runner image does not have gh pre-installed...
-#        env:
-#          # doesn't seem there's an easy way to avoid hard-coding it?
-#          GH_MSI_URL: https://github.com/cli/cli/releases/download/v2.67.0/gh_2.67.0_windows_amd64.msi
-#        run: |
-#          Invoke-WebRequest -Uri "$env:GH_MSI_URL" -OutFile "gh_installer.msi"
-#          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "gh_installer.msi" /qn'
-#          $GH_POSSIBLE_PATHS = "C:\\Program Files\\GitHub CLI", "C:\\Program Files (x86)\\GitHub CLI"
-#          foreach ($p in $GH_POSSIBLE_PATHS) {
-#              echo "$p" >> $env:GITHUB_PATH
-#              $env:Path += ";$p"
-#          }
-#          gh --version
+      - name: Download cuda-python build artifacts
+        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
+        uses: actions/download-artifact@v4
+        with:
+          name: cuda-python-wheel
+          path: .
+
+      - name: Download cuda.bindings build artifacts
+        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
+        uses: actions/download-artifact@v4
+        with:
+          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
+          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
 
-#      - name: Install 7z
-#        # the GPU runner image does not have Git Bash pre-installed...
-#        env:
-#          # doesn't seem there's an easy way to avoid hard-coding it?
-#          7Z_URL: https://www.7-zip.org/a/7z2409-x64.msi
-#        run: |
-#          Invoke-WebRequest -Uri "$env:7Z_URL" -OutFile "7z_installer.msi"
-#          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "7z_installer.msi" /q INSTALLDIR="C:\Program Files\7-Zip" /qn'
-#          ls -l "C:\Program Files\7-Zip"
-#          echo "C:\Program Files\7-Zip" >> $env:GITHUB_PATH
-#          $env:Path += ";C:\Program Files\7-Zip"
-#          echo $env:Path
-#          7z
+      - name: Install gh cli
+        # the GPU runner image does not have gh pre-installed...
+        env:
+          # doesn't seem there's an easy way to avoid hard-coding it?
+          GH_MSI_URL: https://github.com/cli/cli/releases/download/v2.67.0/gh_2.67.0_windows_amd64.msi
+        run: |
+          Invoke-WebRequest -Uri "$env:GH_MSI_URL" -OutFile "gh_installer.msi"
+          Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "gh_installer.msi" /qn'
+          $GH_POSSIBLE_PATHS = "C:\\Program Files\\GitHub CLI", "C:\\Program Files (x86)\\GitHub CLI"
+          foreach ($p in $GH_POSSIBLE_PATHS) {
+              echo "$p" >> $env:GITHUB_PATH
+              $env:Path += ";$p"
+          }
+          gh --version
 
       - name: Install Git for Windows
         # the GPU runner image does not have Git Bash pre-installed...
@@ -128,6 +105,7 @@ jobs:
           GFW_EXE_URL: https://github.com/git-for-windows/git/releases/download/v2.49.0.windows.1/PortableGit-2.49.0-64-bit.7z.exe
         run: |
           Invoke-WebRequest -Uri "$env:GFW_EXE_URL" -OutFile "PortableGit.7z.exe"
+          # Self-extracting, see https://gitforwindows.org/zip-archives-extracting-the-released-archives.html
           Start-Process .\PortableGit.7z.exe -Wait -Verbose -ArgumentList '-y -gm2'
           ls -l PortableGit
           New-Item -Path PortableGit\bash.exe -ItemType SymbolicLink -Value PortableGit\git-bash.exe
@@ -135,67 +113,57 @@ jobs:
           $env:Path += ";$((Get-Location).Path)\\PortableGit"
           bash --version
 
-#      - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
-#        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
-#        env:
-#          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-#        run: |
-#          $OLD_BRANCH = Get-Content .github/BACKPORT_BRANCH
-#          $OLD_BASENAME = "cuda-bindings-python${env:PYTHON_VERSION_FORMATTED}-cuda*-${{ inputs.host-platform }}*"
-#          $runData = gh run list -b $OLD_BRANCH -L 1 -w "CI: Build and test" -s completed -R NVIDIA/cuda-python --json databaseId | ConvertFrom-Json
-#          if (-not $runData -or $runData.Length -eq 0 -or -not $runData[0].databaseId -or [string]::IsNullOrEmpty($runData[0].databaseId)) {
-#              Write-Host "LATEST_PRIOR_RUN_ID not found!"
-#              exit 1
-#          }
-#          $LATEST_PRIOR_RUN_ID = $runData[0].databaseId
-#
-#          gh run download $LATEST_PRIOR_RUN_ID -p $OLD_BASENAME -R NVIDIA/cuda-python
-#          Get-ChildItem -Path $OLD_BASENAME
-#          New-Item -Path "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}" -ItemType Directory -Force
-#          Move-Item -Path "$OLD_BASENAME/*.whl" -Destination "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"
-#          Remove-Item -Path $OLD_BASENAME -Force
-#      
-#          gh run download $LATEST_PRIOR_RUN_ID -p cuda-python-wheel -R NVIDIA/cuda-python
-#          Get-ChildItem -Path cuda-python-wheel
-#          Move-Item -Path "cuda-python-wheel/*.whl" -Destination .
-#          Remove-Item -Path cuda-python-wheel -Force
-#
-#      - name: Display structure of downloaded cuda-python artifacts
-#        run: |
-#          Get-Location
-#          Get-ChildItem -Recurse -Force | Select-Object Mode, LastWriteTime, Length, FullName
-#
-#      - name: Display structure of downloaded cuda.bindings artifacts
-#        run: |
-#          Get-Location
-#          Get-ChildItem -Recurse -Force $env:CUDA_BINDINGS_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
-#
-#      - name: Download cuda.core build artifacts
-#        uses: actions/download-artifact@v4
-#        with:
-#          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
-#          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-#
-#      - name: Display structure of downloaded cuda.core build artifacts
-#        run: |
-#          Get-Location
-#          Get-ChildItem -Recurse -Force $env:CUDA_CORE_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
+      - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
+        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          $OLD_BRANCH = Get-Content .github/BACKPORT_BRANCH
+          $OLD_BASENAME = "cuda-bindings-python${env:PYTHON_VERSION_FORMATTED}-cuda*-${{ inputs.host-platform }}*"
+          $runData = gh run list -b $OLD_BRANCH -L 1 -w "CI: Build and test" -s completed -R NVIDIA/cuda-python --json databaseId | ConvertFrom-Json
+          if (-not $runData -or $runData.Length -eq 0 -or -not $runData[0].databaseId -or [string]::IsNullOrEmpty($runData[0].databaseId)) {
+              Write-Host "LATEST_PRIOR_RUN_ID not found!"
+              exit 1
+          }
+          $LATEST_PRIOR_RUN_ID = $runData[0].databaseId
+
+          gh run download $LATEST_PRIOR_RUN_ID -p $OLD_BASENAME -R NVIDIA/cuda-python
+          Get-ChildItem -Path $OLD_BASENAME
+          New-Item -Path "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}" -ItemType Directory -Force
+          Move-Item -Path "$OLD_BASENAME/*.whl" -Destination "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"
+          Remove-Item -Path $OLD_BASENAME -Force
+      
+          gh run download $LATEST_PRIOR_RUN_ID -p cuda-python-wheel -R NVIDIA/cuda-python
+          Get-ChildItem -Path cuda-python-wheel
+          Move-Item -Path "cuda-python-wheel/*.whl" -Destination .
+          Remove-Item -Path cuda-python-wheel -Force
+
+      - name: Display structure of downloaded cuda-python artifacts
+        run: |
+          Get-Location
+          Get-ChildItem -Recurse -Force | Select-Object Mode, LastWriteTime, Length, FullName
+
+      - name: Display structure of downloaded cuda.bindings artifacts
+        run: |
+          Get-Location
+          Get-ChildItem -Recurse -Force $env:CUDA_BINDINGS_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
+
+      - name: Download cuda.core build artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
+          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+
+      - name: Display structure of downloaded cuda.core build artifacts
+        run: |
+          Get-Location
+          Get-ChildItem -Recurse -Force $env:CUDA_CORE_ARTIFACTS_DIR | Select-Object Mode, LastWriteTime, Length, FullName
 
       - name: Set up Python ${{ inputs.python-version }}
         uses: actions/setup-python@v5
         with:
           python-version: ${{ inputs.python-version }}
 
-#      - name: Set up mini CTK
-#        if: ${{ inputs.local-ctk == '1' }}
-#        # Note: The GH-hosted Windows GPU runner does not have Git for Windows pre-installed,
-#        # so we cannot use our own fetch_ctk action unfortunately...
-#        uses: Jimver/cuda-toolkit@v0.2.21
-#        with:
-#          cuda: ${{ inputs.cuda-version }}
-#          method: 'network'
-#          sub-packages: ${{ env.MINI_CTK_DEPS }}
-
       - name: Set up mini CTK
         if: ${{ inputs.local-ctk == '1' }}
         uses: ./.github/actions/fetch_ctk

From 1b02e55ccf76b3013180c758239e4750f04b1231 Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Sat, 3 May 2025 00:26:04 -0400
Subject: [PATCH 25/32] ensure cuda related paths are set correctly on windows

---
 .github/actions/fetch_ctk/action.yml     | 12 +++++++++---
 .github/workflows/build-and-test.yml     |  2 +-
 .github/workflows/test-wheel-windows.yml |  8 +-------
 3 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/.github/actions/fetch_ctk/action.yml b/.github/actions/fetch_ctk/action.yml
index 3aff4100ac..47ad49e786 100644
--- a/.github/actions/fetch_ctk/action.yml
+++ b/.github/actions/fetch_ctk/action.yml
@@ -146,8 +146,14 @@ runs:
     - name: Set output environment variables
       shell: bash -xeuo pipefail {0}
       run: |
-        CUDA_PATH=$(realpath "./cuda_toolkit")
+        # mimics actual CTK installation
+        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
+          CUDA_PATH=$(realpath "./cuda_toolkit")
+          echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH:-}:${CUDA_PATH}/lib:${CUDA_PATH}/nvvm/lib64" >> $GITHUB_ENV
+        elif [[ "${{ inputs.host-platform }}" == win* ]]; then
+          CUDA_PATH=$(cygpath -w $(realpath "./cuda_toolkit"))
+          echo "$(cygpath -w ${CUDA_PATH}/bin)" >> $GITHUB_PATH
+          echo "$(cygpath -w $CUDA_PATH/nvvm/bin)" >> $GITHUB_PATH
+        fi
         echo "CUDA_PATH=${CUDA_PATH}" >> $GITHUB_ENV
         echo "CUDA_HOME=${CUDA_PATH}" >> $GITHUB_ENV
-        echo "${CUDA_PATH}/bin" >> $GITHUB_PATH
-        echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH:-}:${CUDA_PATH}/lib:${CUDA_PATH}/nvvm/lib64" >> $GITHUB_ENV
diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index a00a940286..b6d3ce5497 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -146,7 +146,7 @@ jobs:
             LIBRARY_PATH=/host/${{ env.CUDA_PATH }}/lib
             CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
           CIBW_ENVIRONMENT_WINDOWS: >
-            CUDA_HOME="$(cygpath -w ${{ env.CUDA_PATH }})"
+            CUDA_HOME=${{ env.CUDA_PATH }}
             LIB="${CUDA_HOME}\\lib\\x64;${LIB}"
             CUDA_BINDINGS_PARALLEL_LEVEL=${{ env.CUDA_BINDINGS_PARALLEL_LEVEL }}
         with:
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 347a98a8fa..567c281e24 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -100,6 +100,7 @@ jobs:
 
       - name: Install Git for Windows
         # the GPU runner image does not have Git Bash pre-installed...
+        if: ${{ inputs.local-ctk == '1' }}
         env:
           # doesn't seem there's an easy way to avoid hard-coding it?
           GFW_EXE_URL: https://github.com/git-for-windows/git/releases/download/v2.49.0.windows.1/PortableGit-2.49.0-64-bit.7z.exe
@@ -172,13 +173,6 @@ jobs:
           host-platform: ${{ inputs.host-platform }}
           cuda-version: ${{ inputs.cuda-version }}
 
-      - name: Update PATH
-        if: ${{ inputs.local-ctk == '1' }}
-        run: |
-          # mimics actual CTK installation
-          echo $PATH
-          echo "$env:CUDA_PATH\nvvm\bin" >> $env:GITHUB_PATH
-
       - name: Run cuda.bindings tests
         if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0' }}
         run: |

From 7a0e053feec2d65998d5526022c49c32c6c42e97 Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Sat, 3 May 2025 00:50:21 -0400
Subject: [PATCH 26/32] fix normpath (again) to escape slash

---
 .github/actions/fetch_ctk/action.yml | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/.github/actions/fetch_ctk/action.yml b/.github/actions/fetch_ctk/action.yml
index 47ad49e786..03f30eaa73 100644
--- a/.github/actions/fetch_ctk/action.yml
+++ b/.github/actions/fetch_ctk/action.yml
@@ -151,9 +151,12 @@ runs:
           CUDA_PATH=$(realpath "./cuda_toolkit")
           echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH:-}:${CUDA_PATH}/lib:${CUDA_PATH}/nvvm/lib64" >> $GITHUB_ENV
         elif [[ "${{ inputs.host-platform }}" == win* ]]; then
-          CUDA_PATH=$(cygpath -w $(realpath "./cuda_toolkit"))
-          echo "$(cygpath -w ${CUDA_PATH}/bin)" >> $GITHUB_PATH
-          echo "$(cygpath -w $CUDA_PATH/nvvm/bin)" >> $GITHUB_PATH
+          function normpath() {
+            echo "$(echo $(cygpath -w $1) | sed 's/\\/\\\\/g')"
+          }
+          CUDA_PATH=$(normpath $(realpath "./cuda_toolkit"))
+          echo "$(normpath ${CUDA_PATH}/bin)" >> $GITHUB_PATH
+          echo "$(normpath $CUDA_PATH/nvvm/bin)" >> $GITHUB_PATH
         fi
         echo "CUDA_PATH=${CUDA_PATH}" >> $GITHUB_ENV
         echo "CUDA_HOME=${CUDA_PATH}" >> $GITHUB_ENV

From 96ecae04e10ee2ac3763da6c5b6679f7ae4d1387 Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Sat, 3 May 2025 02:22:06 -0400
Subject: [PATCH 27/32] DEBUG

---
 .github/workflows/build-and-test.yml     | 118 +++++++++++------------
 .github/workflows/test-wheel-windows.yml |   8 ++
 2 files changed, 67 insertions(+), 59 deletions(-)

diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index b6d3ce5497..da4809224a 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -24,15 +24,15 @@ jobs:
       fail-fast: false
       matrix:
         host-platform:
-          - linux-64
-          - linux-aarch64
+#          - linux-64
+#          - linux-aarch64
           - win-64
         python-version:
-          - "3.13"
+#          - "3.13"
           - "3.12"
-          - "3.11"
-          - "3.10"
-          - "3.9"
+#          - "3.11"
+#          - "3.10"
+#          - "3.9"
         cuda-version:
           # Note: this is for build-time only.
           - "12.8.0"
@@ -208,56 +208,56 @@ jobs:
         run: |
           echo "CUDA_VERSION=${{ matrix.cuda-version }}" >> $GITHUB_OUTPUT
 
-  test-linux:
-    strategy:
-      fail-fast: false
-      # TODO: add driver version here
-      matrix:
-        host-platform:
-          - linux-64
-          - linux-aarch64
-        python-version:
-          - "3.13"
-          - "3.12"
-          - "3.11"
-          - "3.10"
-          - "3.9"
-        cuda-version:
-          # Note: this is for test-time only.
-          - "12.8.0"
-          - "12.0.1"
-          - "11.8.0"
-        local-ctk:
-          - 1  # use mini CTK
-          - 0  # use CTK wheels
-        runner:
-          - default
-        exclude:
-          # To test this combo would require nontrivial installation steps.
-          - cuda-version: "12.0.1"
-            local-ctk: 0
-        include:
-          - host-platform: linux-64
-            python-version: "3.12"
-            cuda-version: "12.8.0"
-            local-ctk: 1
-            runner: H100
-    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
-    if: ${{ github.repository_owner == 'nvidia' }}
-    permissions:
-      contents: read  # This is required for actions/checkout
-    needs:
-      - build
-    secrets: inherit
-    uses:
-      ./.github/workflows/test-wheel-linux.yml
-    with:
-      host-platform: ${{ matrix.host-platform }}
-      python-version: ${{ matrix.python-version }}
-      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
-      cuda-version: ${{ matrix.cuda-version }}
-      local-ctk: ${{ matrix.local-ctk}}
-      runner: ${{ matrix.runner }}
+#  test-linux:
+#    strategy:
+#      fail-fast: false
+#      # TODO: add driver version here
+#      matrix:
+#        host-platform:
+#          - linux-64
+#          - linux-aarch64
+#        python-version:
+#          - "3.13"
+#          - "3.12"
+#          - "3.11"
+#          - "3.10"
+#          - "3.9"
+#        cuda-version:
+#          # Note: this is for test-time only.
+#          - "12.8.0"
+#          - "12.0.1"
+#          - "11.8.0"
+#        local-ctk:
+#          - 1  # use mini CTK
+#          - 0  # use CTK wheels
+#        runner:
+#          - default
+#        exclude:
+#          # To test this combo would require nontrivial installation steps.
+#          - cuda-version: "12.0.1"
+#            local-ctk: 0
+#        include:
+#          - host-platform: linux-64
+#            python-version: "3.12"
+#            cuda-version: "12.8.0"
+#            local-ctk: 1
+#            runner: H100
+#    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
+#    if: ${{ github.repository_owner == 'nvidia' }}
+#    permissions:
+#      contents: read  # This is required for actions/checkout
+#    needs:
+#      - build
+#    secrets: inherit
+#    uses:
+#      ./.github/workflows/test-wheel-linux.yml
+#    with:
+#      host-platform: ${{ matrix.host-platform }}
+#      python-version: ${{ matrix.python-version }}
+#      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
+#      cuda-version: ${{ matrix.cuda-version }}
+#      local-ctk: ${{ matrix.local-ctk}}
+#      runner: ${{ matrix.runner }}
 
   test-windows:
     strategy:
@@ -270,11 +270,11 @@ jobs:
           - "3.12"
         cuda-version:
           # Note: this is for test-time only.
-          - "12.8.0"
+#          - "12.8.0"
           - "11.8.0"
         local-ctk:
           - 1  # use mini CTK
-          - 0  # use CTK wheels
+#          - 0  # use CTK wheels
         runner:
           - default
     name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
@@ -316,7 +316,7 @@ jobs:
       checks: read
     needs:
       - build
-      - test-linux
+#      - test-linux
       - test-windows
       - doc
     secrets: inherit
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 567c281e24..cdfcadc907 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -173,6 +173,14 @@ jobs:
           host-platform: ${{ inputs.host-platform }}
           cuda-version: ${{ inputs.cuda-version }}
 
+      - name: DEBUG
+        if: ${{ inputs.local-ctk == '1' }}
+        run: |
+          echo $env:CUDA_PATH
+          echo $env:Path
+          Get-Location $env:CUDA_PATH
+          Get-Location $env:CUDA_PATH/bin
+
       - name: Run cuda.bindings tests
         if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0' }}
         run: |

From a73564c5b423033a985e42ab18a44d96a54d3fea Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Sat, 3 May 2025 02:58:44 -0400
Subject: [PATCH 28/32] perhaps git-bash is not the right exe...

---
 .github/workflows/test-wheel-windows.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index cdfcadc907..45ae501ee6 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -109,9 +109,9 @@ jobs:
           # Self-extracting, see https://gitforwindows.org/zip-archives-extracting-the-released-archives.html
           Start-Process .\PortableGit.7z.exe -Wait -Verbose -ArgumentList '-y -gm2'
           ls -l PortableGit
-          New-Item -Path PortableGit\bash.exe -ItemType SymbolicLink -Value PortableGit\git-bash.exe
-          echo "$((Get-Location).Path)\\PortableGit" >> $env:GITHUB_PATH
-          $env:Path += ";$((Get-Location).Path)\\PortableGit"
+          #New-Item -Path PortableGit\bash.exe -ItemType SymbolicLink -Value PortableGit\git-bash.exe
+          echo "$((Get-Location).Path)\\PortableGit\\bin" >> $env:GITHUB_PATH
+          $env:Path += ";$((Get-Location).Path)\\PortableGit\\bin"
           bash --version
 
       - name: Download cuda-python & cuda.bindings build artifacts from the prior branch

From 78d0113525507c1cff8357bb38a733928c04639e Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Sat, 3 May 2025 21:30:26 -0400
Subject: [PATCH 29/32] install zstd for win local ctk test

---
 .github/actions/fetch_ctk/action.yml     |  8 ++++----
 .github/workflows/test-wheel-windows.yml | 16 +++++++++++++++-
 2 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/.github/actions/fetch_ctk/action.yml b/.github/actions/fetch_ctk/action.yml
index 03f30eaa73..cbdd25ad17 100644
--- a/.github/actions/fetch_ctk/action.yml
+++ b/.github/actions/fetch_ctk/action.yml
@@ -23,7 +23,7 @@ runs:
   using: composite
   steps:
     - name: Set up CTK cache variable
-      shell: bash -xeuo pipefail {0}
+      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         HASH=$(echo -n "${{ inputs.cuda-components }}" | sha256sum | awk '{print $1}')
         echo "CTK_CACHE_KEY=mini-ctk-${{ inputs.cuda-version }}-${{ inputs.host-platform }}-$HASH" >> $GITHUB_ENV
@@ -48,7 +48,7 @@ runs:
 
     - name: Get CUDA components
       if: ${{ steps.ctk-get-cache.outputs.cache-hit != 'true' }}
-      shell: bash -xeuo pipefail {0}
+      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         # Everything under this folder is packed and stored in the GitHub Cache space,
         # and unpacked after retrieving from the cache.
@@ -128,7 +128,7 @@ runs:
 
     - name: Restore CTK cache
       if: ${{ steps.ctk-get-cache.outputs.cache-hit == 'true' }}
-      shell: bash -xeuo pipefail {0}
+      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         ls -l
         CACHE_TMP_DIR="./cache_tmp_dir"
@@ -144,7 +144,7 @@ runs:
         fi
 
     - name: Set output environment variables
-      shell: bash -xeuo pipefail {0}
+      shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         # mimics actual CTK installation
         if [[ "${{ inputs.host-platform }}" == linux* ]]; then
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 45ae501ee6..b16c8c5fe8 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -109,11 +109,25 @@ jobs:
           # Self-extracting, see https://gitforwindows.org/zip-archives-extracting-the-released-archives.html
           Start-Process .\PortableGit.7z.exe -Wait -Verbose -ArgumentList '-y -gm2'
           ls -l PortableGit
-          #New-Item -Path PortableGit\bash.exe -ItemType SymbolicLink -Value PortableGit\git-bash.exe
           echo "$((Get-Location).Path)\\PortableGit\\bin" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\\PortableGit\\bin"
           bash --version
 
+      - name: Install zstd
+        # the GPU runner image does not have zstd pre-installed... and it's needed by actions/cache
+        if: ${{ inputs.local-ctk == '1' }}
+        env:
+          # doesn't seem there's an easy way to avoid hard-coding it?
+          ZSTD_URL: https://github.com/facebook/zstd/releases/download/v1.5.7/zstd-v1.5.7-win64.zip
+          ZSTD_DIR: zstd-v1.5.7-win64
+        run: |
+          Invoke-WebRequest -Uri "$env:ZSTD_URL" -OutFile "zstd-win64.zip"
+          Expand-Archive -Path "zstd-win64.zip"
+          ls -l $env:ZSTD_DIR
+          echo "$((Get-Location).Path)\\$env:ZSTD_DIR" >> $env:GITHUB_PATH
+          $env:Path += ";$((Get-Location).Path)\\$env:ZSTD_DIR"
+          zstd -h
+
       - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
         if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
         env:

From 390995ab2760229c2a3c3c60e6f1177e96759ebd Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Sat, 3 May 2025 21:46:07 -0400
Subject: [PATCH 30/32] fix copy pasta

---
 .github/workflows/test-wheel-windows.yml | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index b16c8c5fe8..67d68045d6 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -37,12 +37,12 @@ jobs:
         with:
           fetch-depth: 0
 
-      - name: Update driver
-        run: |
-          .github/workflows/install_gpu_driver.ps1
-
-      - name: Ensure GPU is working
-        run: nvidia-smi
+#      - name: Update driver
+#        run: |
+#          .github/workflows/install_gpu_driver.ps1
+#
+#      - name: Ensure GPU is working
+#        run: nvidia-smi
 
       - name: Set environment variables
         run: |
@@ -122,7 +122,7 @@ jobs:
           ZSTD_DIR: zstd-v1.5.7-win64
         run: |
           Invoke-WebRequest -Uri "$env:ZSTD_URL" -OutFile "zstd-win64.zip"
-          Expand-Archive -Path "zstd-win64.zip"
+          Expand-Archive -Path "zstd-win64.zip" -DestinationPath .
           ls -l $env:ZSTD_DIR
           echo "$((Get-Location).Path)\\$env:ZSTD_DIR" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\\$env:ZSTD_DIR"

From 65e19b230279e6fbc318c20e28a753f3e9628641 Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Sat, 3 May 2025 22:04:16 -0400
Subject: [PATCH 31/32] forward-port fetch_ctk fix

---
 .github/actions/fetch_ctk/action.yml | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/.github/actions/fetch_ctk/action.yml b/.github/actions/fetch_ctk/action.yml
index cbdd25ad17..f6b78701a5 100644
--- a/.github/actions/fetch_ctk/action.yml
+++ b/.github/actions/fetch_ctk/action.yml
@@ -75,6 +75,8 @@ runs:
             unzip $1 -d $_TEMP_DIR_
             cp -r $_TEMP_DIR_/*/* $CACHE_TMP_DIR
             rm -rf $_TEMP_DIR_
+            # see commit NVIDIA/cuda-python@69410f1d9228e775845ef6c8b4a9c7f37ffc68a5
+            chmod 644 $CACHE_TMP_DIR/LICENSE
           }
         fi
         function populate_cuda_path() {

From 6ade91b6bf70883d9c0f5d42dc46cbedfe2c4652 Mon Sep 17 00:00:00 2001
From: Leo Fang <leo80042@gmail.com>
Date: Sat, 3 May 2025 22:21:55 -0400
Subject: [PATCH 32/32] ready for full test!

---
 .github/workflows/build-and-test.yml     | 118 +++++++++++------------
 .github/workflows/test-wheel-windows.yml |  22 ++---
 2 files changed, 66 insertions(+), 74 deletions(-)

diff --git a/.github/workflows/build-and-test.yml b/.github/workflows/build-and-test.yml
index da4809224a..b6d3ce5497 100644
--- a/.github/workflows/build-and-test.yml
+++ b/.github/workflows/build-and-test.yml
@@ -24,15 +24,15 @@ jobs:
       fail-fast: false
       matrix:
         host-platform:
-#          - linux-64
-#          - linux-aarch64
+          - linux-64
+          - linux-aarch64
           - win-64
         python-version:
-#          - "3.13"
+          - "3.13"
           - "3.12"
-#          - "3.11"
-#          - "3.10"
-#          - "3.9"
+          - "3.11"
+          - "3.10"
+          - "3.9"
         cuda-version:
           # Note: this is for build-time only.
           - "12.8.0"
@@ -208,56 +208,56 @@ jobs:
         run: |
           echo "CUDA_VERSION=${{ matrix.cuda-version }}" >> $GITHUB_OUTPUT
 
-#  test-linux:
-#    strategy:
-#      fail-fast: false
-#      # TODO: add driver version here
-#      matrix:
-#        host-platform:
-#          - linux-64
-#          - linux-aarch64
-#        python-version:
-#          - "3.13"
-#          - "3.12"
-#          - "3.11"
-#          - "3.10"
-#          - "3.9"
-#        cuda-version:
-#          # Note: this is for test-time only.
-#          - "12.8.0"
-#          - "12.0.1"
-#          - "11.8.0"
-#        local-ctk:
-#          - 1  # use mini CTK
-#          - 0  # use CTK wheels
-#        runner:
-#          - default
-#        exclude:
-#          # To test this combo would require nontrivial installation steps.
-#          - cuda-version: "12.0.1"
-#            local-ctk: 0
-#        include:
-#          - host-platform: linux-64
-#            python-version: "3.12"
-#            cuda-version: "12.8.0"
-#            local-ctk: 1
-#            runner: H100
-#    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
-#    if: ${{ github.repository_owner == 'nvidia' }}
-#    permissions:
-#      contents: read  # This is required for actions/checkout
-#    needs:
-#      - build
-#    secrets: inherit
-#    uses:
-#      ./.github/workflows/test-wheel-linux.yml
-#    with:
-#      host-platform: ${{ matrix.host-platform }}
-#      python-version: ${{ matrix.python-version }}
-#      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
-#      cuda-version: ${{ matrix.cuda-version }}
-#      local-ctk: ${{ matrix.local-ctk}}
-#      runner: ${{ matrix.runner }}
+  test-linux:
+    strategy:
+      fail-fast: false
+      # TODO: add driver version here
+      matrix:
+        host-platform:
+          - linux-64
+          - linux-aarch64
+        python-version:
+          - "3.13"
+          - "3.12"
+          - "3.11"
+          - "3.10"
+          - "3.9"
+        cuda-version:
+          # Note: this is for test-time only.
+          - "12.8.0"
+          - "12.0.1"
+          - "11.8.0"
+        local-ctk:
+          - 1  # use mini CTK
+          - 0  # use CTK wheels
+        runner:
+          - default
+        exclude:
+          # To test this combo would require nontrivial installation steps.
+          - cuda-version: "12.0.1"
+            local-ctk: 0
+        include:
+          - host-platform: linux-64
+            python-version: "3.12"
+            cuda-version: "12.8.0"
+            local-ctk: 1
+            runner: H100
+    name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
+    if: ${{ github.repository_owner == 'nvidia' }}
+    permissions:
+      contents: read  # This is required for actions/checkout
+    needs:
+      - build
+    secrets: inherit
+    uses:
+      ./.github/workflows/test-wheel-linux.yml
+    with:
+      host-platform: ${{ matrix.host-platform }}
+      python-version: ${{ matrix.python-version }}
+      build-ctk-ver: ${{ needs.build.outputs.BUILD_CTK_VER }}
+      cuda-version: ${{ matrix.cuda-version }}
+      local-ctk: ${{ matrix.local-ctk}}
+      runner: ${{ matrix.runner }}
 
   test-windows:
     strategy:
@@ -270,11 +270,11 @@ jobs:
           - "3.12"
         cuda-version:
           # Note: this is for test-time only.
-#          - "12.8.0"
+          - "12.8.0"
           - "11.8.0"
         local-ctk:
           - 1  # use mini CTK
-#          - 0  # use CTK wheels
+          - 0  # use CTK wheels
         runner:
           - default
     name: Test (${{ matrix.host-platform }}, Python ${{ matrix.python-version }}, CUDA ${{ matrix.cuda-version }}, Runner ${{ matrix.runner }}, ${{ (matrix.local-ctk == '1' && 'local CTK') || 'CTK wheels' }})
@@ -316,7 +316,7 @@ jobs:
       checks: read
     needs:
       - build
-#      - test-linux
+      - test-linux
       - test-windows
       - doc
     secrets: inherit
diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 67d68045d6..c741fe1616 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -37,12 +37,12 @@ jobs:
         with:
           fetch-depth: 0
 
-#      - name: Update driver
-#        run: |
-#          .github/workflows/install_gpu_driver.ps1
-#
-#      - name: Ensure GPU is working
-#        run: nvidia-smi
+      - name: Update driver
+        run: |
+          .github/workflows/install_gpu_driver.ps1
+
+      - name: Ensure GPU is working
+        run: nvidia-smi
 
       - name: Set environment variables
         run: |
@@ -126,7 +126,7 @@ jobs:
           ls -l $env:ZSTD_DIR
           echo "$((Get-Location).Path)\\$env:ZSTD_DIR" >> $env:GITHUB_PATH
           $env:Path += ";$((Get-Location).Path)\\$env:ZSTD_DIR"
-          zstd -h
+          zstd --version
 
       - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
         if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
@@ -187,14 +187,6 @@ jobs:
           host-platform: ${{ inputs.host-platform }}
           cuda-version: ${{ inputs.cuda-version }}
 
-      - name: DEBUG
-        if: ${{ inputs.local-ctk == '1' }}
-        run: |
-          echo $env:CUDA_PATH
-          echo $env:Path
-          Get-Location $env:CUDA_PATH
-          Get-Location $env:CUDA_PATH/bin
-
       - name: Run cuda.bindings tests
         if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0' }}
         run: |