Merge branch 'main' into oom-early-reservation-workaround

Author: Andy-Jost

.github/workflows/coverage.yml

@@ -235,13 +235,34 @@ jobs:
           cd cuda_bindings
           ../.venv/Scripts/pip wheel -v --no-deps . -w ../wheels/
 
+      # Pin cuda-bindings to the wheel built above; PIP_PRE, which is what makes
+      # that .devN wheel visible, would otherwise let a PyPI pre-release win.
       - name: Build cuda.core wheel
         run: |
           export PIP_FIND_LINKS="$(pwd)/wheels"
           export PIP_PRE=1
+          bindings_whl="$(ls ./wheels/cuda_bindings-*.whl | head -1)"
+          bindings_ver="$(basename "$bindings_whl" | cut -d- -f2)"
+          echo "cuda-bindings==${bindings_ver%%+*}" > "$GITHUB_WORKSPACE/constraints.txt"
+          cat "$GITHUB_WORKSPACE/constraints.txt"
+          export PIP_CONSTRAINT="$GITHUB_WORKSPACE/constraints.txt"
           cd cuda_core
           ../.venv/Scripts/pip wheel -v --no-deps . -w ../wheels/
 
+      # Vendor the DLLs these wheels were built against, the way cibuildwheel
+      # does for every other Windows build.  --namespace-pkg is needed because
+      # `cuda` is a namespace package.
+      - name: Repair the Windows wheels
+        run: |
+          .venv/Scripts/pip install delvewheel
+          mkdir -p wheels-repaired
+          for whl in ./wheels/cuda_bindings-*.whl ./wheels/cuda_core-*.whl; do
+            .venv/Scripts/delvewheel repair --namespace-pkg cuda \
+              --exclude "torch_cpu.dll;torch_python.dll" \
+              -w ./wheels-repaired "$whl"
+          done
+          mv -f ./wheels-repaired/*.whl ./wheels/
+
       - name: List wheel artifacts
         run: |
           echo "=== Windows wheel artifacts ==="

ci/tools/merge_cuda_core_wheels.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 
-# SPDX-FileCopyrightText: Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # SPDX-License-Identifier: Apache-2.0
 
@@ -27,10 +27,9 @@
 import tempfile
 import zipfile
 from pathlib import Path
-from typing import List
 
 
-def run_command(cmd: List[str], cwd: Path | None = None, env: dict = os.environ) -> subprocess.CompletedProcess:
+def run_command(cmd: list[str], cwd: Path | None = None, env: dict = os.environ) -> subprocess.CompletedProcess:
     """Run a command with error handling."""
     print(f"Running: {' '.join(cmd)}")
     if cwd:
@@ -78,7 +77,7 @@ def print_wheel_directory_structure(wheel_path: Path, filter_prefix: str = "cuda
         print(f"Warning: Could not list wheel contents: {e}", file=sys.stderr)
 
 
-def merge_wheels(wheels: List[Path], output_dir: Path, show_wheel_contents: bool = True) -> Path:
+def merge_wheels(wheels: list[Path], output_dir: Path, show_wheel_contents: bool = True) -> Path:
     """Merge multiple wheels into a single wheel with version-specific binaries."""
     print("\n=== Merging wheels ===", file=sys.stderr)
     print(f"Input wheels: {[w.name for w in wheels]}", file=sys.stderr)

cuda_core/tests/helpers/__init__.py

@@ -3,7 +3,6 @@
 
 import functools
 import os
-from typing import Union
 
 from cuda.core._utils.cuda_utils import handle_return
 from cuda.pathfinder import get_cuda_path_or_home
@@ -23,7 +22,7 @@
 
 
 @functools.cache
-def supports_ipc_mempool(device_id: Union[int, object]) -> bool:
+def supports_ipc_mempool(device_id: int | object) -> bool:
     """Return True if mempool IPC via POSIX file descriptor is supported.
 
     Uses cuDeviceGetAttribute(CU_DEVICE_ATTRIBUTE_MEMPOOL_SUPPORTED_HANDLE_TYPES)

cuda_pathfinder/tests/test_ctk_root_discovery.py

@@ -6,6 +6,7 @@
 import subprocess
 import sys
 import textwrap
+from pathlib import Path
 
 import pytest
 
@@ -427,7 +428,7 @@ def test_resolve_ctk_root_via_canary_none_when_probe_fails(mocker):
 def test_resolve_ctk_root_via_canary_none_when_unrecognized(mocker):
     mocker.patch(
         f"{_MODULE}._resolve_system_loaded_abs_path_in_subprocess",
-        return_value=os.path.join(os.sep, "weird", "path", "libcudart.so.13"),
+        return_value=str(Path(os.sep, "weird", "path", "libcudart.so.13")),
     )
     assert resolve_ctk_root_via_canary("cudart") is None
 

cuda_pathfinder/tests/test_driver_lib_loading.py

@@ -9,6 +9,7 @@
 """
 
 import os
+from pathlib import Path
 
 import pytest
 from child_load_nvidia_dynamic_lib_helper import (
@@ -157,7 +158,7 @@ def raise_child_process_failed():
         abs_path = payload.abs_path
         assert abs_path is not None
         info_summary_append(f"abs_path={quote_for_shell(abs_path)}")
-        assert os.path.isfile(abs_path)
+        assert Path(abs_path).is_file()
 
 
 def test_real_query_driver_cuda_version(info_summary_append):

cuda_pathfinder/tests/test_find_bitcode_lib.py

@@ -66,7 +66,7 @@ def _located_bitcode_lib_asserts(located_bitcode_lib):
     assert isinstance(located_bitcode_lib.filename, str)
     assert isinstance(located_bitcode_lib.found_via, str)
     assert located_bitcode_lib.found_via in ("site-packages", "conda", "CUDA_PATH")
-    assert os.path.isfile(located_bitcode_lib.abs_path)
+    assert Path(located_bitcode_lib.abs_path).is_file()
 
 
 @pytest.mark.usefixtures("clear_find_bitcode_lib_cache")
@@ -83,10 +83,10 @@ def test_locate_bitcode_lib(info_summary_append, libname):
 
     info_summary_append(f"{lib_path=!r}")
     _located_bitcode_lib_asserts(located_lib)
-    assert os.path.isfile(lib_path)
+    assert Path(lib_path).is_file()
     assert lib_path == located_lib.abs_path
     expected_filename = located_lib.filename
-    assert os.path.basename(lib_path) == expected_filename
+    assert Path(lib_path).name == expected_filename
 
 
 @pytest.mark.usefixtures("clear_find_bitcode_lib_cache")
@@ -156,7 +156,7 @@ def test_find_bitcode_lib_not_found_error_includes_cuda_home_directory_listing(m
         find_bitcode_lib("device")
 
     message = str(exc_info.value)
-    expected_missing_file = os.path.join(str(lib_dir), _bitcode_lib_filename("device"))
+    expected_missing_file = lib_dir / _bitcode_lib_filename("device")
     assert f"No such file: {expected_missing_file}" in message
     assert f'listdir("{lib_dir}"):' in message
     assert "README.txt" in message

cuda_pathfinder/tests/test_find_nvidia_headers.py

@@ -138,12 +138,12 @@ def test_locate_non_ctk_headers(info_summary_append, libname):
     info_summary_append(f"{hdr_dir=!r}")
     if hdr_dir:
         _located_hdr_dir_asserts(located_hdr_dir)
-        assert os.path.isdir(hdr_dir)
-        assert os.path.isfile(os.path.join(hdr_dir, SUPPORTED_HEADERS_NON_CTK[libname]))
+        hdr_dir_path = Path(hdr_dir)
+        assert hdr_dir_path.is_dir()
+        assert (hdr_dir_path / SUPPORTED_HEADERS_NON_CTK[libname]).is_file()
     if have_distribution_for(libname):
         assert hdr_dir is not None
-        hdr_dir_parts = hdr_dir.split(os.path.sep)
-        assert "site-packages" in hdr_dir_parts
+        assert "site-packages" in Path(hdr_dir).parts
     elif STRICTNESS == "all_must_work":
         assert hdr_dir is not None
         if conda_prefix := os.environ.get("CONDA_PREFIX"):
@@ -152,6 +152,8 @@ def test_locate_non_ctk_headers(info_summary_append, libname):
             inst_dirs = SUPPORTED_INSTALL_DIRS_NON_CTK.get(libname)
             if inst_dirs is not None:
                 for inst_dir in inst_dirs:
+                    # Absolute glob pattern: Path.glob needs a separate base dir,
+                    # and the wildcard is not pinned to the last component.
                     globbed = glob.glob(inst_dir)
                     if hdr_dir in globbed:
                         break
@@ -172,9 +174,10 @@ def test_locate_ctk_headers(info_summary_append, libname):
     info_summary_append(f"{hdr_dir=!r}")
     if hdr_dir:
         _located_hdr_dir_asserts(located_hdr_dir)
-        assert os.path.isdir(hdr_dir)
+        hdr_dir_path = Path(hdr_dir)
+        assert hdr_dir_path.is_dir()
         h_filename = SUPPORTED_HEADERS_CTK[libname]
-        assert os.path.isfile(os.path.join(hdr_dir, h_filename))
+        assert (hdr_dir_path / h_filename).is_file()
     if STRICTNESS == "all_must_work":
         if libname == "cudla":
             skip_if_missing_libnvcudla_so(libname, timeout=30)

cuda_pathfinder/tests/test_find_static_lib.py

@@ -52,7 +52,7 @@ def _located_static_lib_asserts(located_static_lib):
     assert isinstance(located_static_lib.filename, str)
     assert isinstance(located_static_lib.found_via, str)
     assert located_static_lib.found_via in ("site-packages", "conda", "CUDA_PATH")
-    assert os.path.isfile(located_static_lib.abs_path)
+    assert Path(located_static_lib.abs_path).is_file()
 
 
 @pytest.mark.usefixtures("clear_find_static_lib_cache")
@@ -69,10 +69,10 @@ def test_locate_static_lib(info_summary_append, libname):
 
     info_summary_append(f"abs_path={quote_for_shell(lib_path)}")
     _located_static_lib_asserts(located_lib)
-    assert os.path.isfile(lib_path)
+    assert Path(lib_path).is_file()
     assert lib_path == located_lib.abs_path
     expected_filename = located_lib.filename
-    assert os.path.basename(lib_path) == expected_filename
+    assert Path(lib_path).name == expected_filename
 
 
 @pytest.mark.usefixtures("clear_find_static_lib_cache")
@@ -81,7 +81,7 @@ def test_locate_static_lib_search_order(monkeypatch, tmp_path):
     conda_rel_path = CUDADEVRT_INFO["conda_rel_paths"][0]
 
     site_pkg_rel = CUDADEVRT_INFO["site_packages_dirs"][0]
-    site_packages_lib_dir = tmp_path / "site-packages" / Path(site_pkg_rel.replace("/", os.sep))
+    site_packages_lib_dir = tmp_path / "site-packages" / Path(site_pkg_rel)
     site_packages_path = _make_static_lib_file(site_packages_lib_dir, filename)
 
     conda_prefix = tmp_path / "conda-prefix"
@@ -202,7 +202,7 @@ def test_find_static_lib_not_found_error_includes_cuda_home_directory_listing(mo
         find_static_lib("cudadevrt")
 
     message = str(exc_info.value)
-    expected_missing_file = os.path.join(str(lib_dir), filename)
+    expected_missing_file = lib_dir / filename
     assert f"No such file: {expected_missing_file}" in message
     assert f'listdir("{lib_dir}"):' in message
     assert "README.txt" in message

cuda_pathfinder/tests/test_load_nvidia_dynamic_lib.py

@@ -3,6 +3,7 @@
 
 import os
 import platform
+from pathlib import Path
 
 import pytest
 from child_load_nvidia_dynamic_lib_helper import (
@@ -159,4 +160,4 @@ def raise_child_process_failed():
         abs_path = payload.abs_path
         assert abs_path is not None
         info_summary_append(f"abs_path={quote_for_shell(abs_path)}")
-        assert os.path.isfile(abs_path)  # double-check the abs_path
+        assert Path(abs_path).is_file()  # double-check the abs_path

cuda_pathfinder/tests/test_utils_find_sub_dirs.py

@@ -1,7 +1,7 @@
 # SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-import os
+from pathlib import Path
 
 import pytest
 
@@ -77,7 +77,7 @@ def test_empty_parent_paths():
 def test_empty_sub_dirs(test_tree):
     parent_paths = test_tree["parent_paths"]
     result = find_sub_dirs(parent_paths, ())
-    expected = [p for p in parent_paths if os.path.isdir(p)]
+    expected = [p for p in parent_paths if Path(p).is_dir()]
     assert sorted(result) == sorted(expected)