Unify ruff config and apply lint fixes across all sub-repos (#1703)

* Unify ruff config across sub-repos

Merge cuda_pathfinder's [tool.ruff] into the top-level ruff.toml and
remove the duplicate. The unified config is the union of all lint rules
(adding ARG, Q, N, C4, PIE, T20, RUF, PT, DTZ) with sane per-file
ignores for tests, examples, benchmarks, scripts, and CUDA naming
conventions in cuda_bindings.

Made-with: Cursor

* Add ruff to pixi dependencies

Made-with: Cursor

* Update ruff config with review feedback

- Remove combine-as-imports (keep as-imports separate from bare imports)
- Add C408 to global ignores (dict()/list()/tuple() calls are fine)
- Add RUF005 to global ignores (list concatenation is fine)
- Add E402, F403, F405 to __init__.py per-file-ignores
- Add E702, B028 to test per-file-ignores
- Add T201 for canary_probe_subprocess.py

Made-with: Cursor

* fix(ruff/cuda_core): I001 sort imports

- I001: unsorted imports — 48 fixed across 48 files

Made-with: Cursor

* fix(ruff/cuda_core): RUF100 remove stale noqa comments

- RUF100: unused-noqa — 80 fixed across 24 files

Made-with: Cursor

* fix(ruff/cuda_core): C4 simplify comprehensions

- C416: unnecessary-comprehension — use set()/list() directly
- C401: unnecessary-generator-set — use set comprehension
- C419: unnecessary-comprehension-in-call — use generator in all()/any()
- C400: unnecessary-generator-list — use list comprehension
- C414: unnecessary-double-cast-or-process — remove redundant list()
- C420: unnecessary-dict-comprehension — use dict.fromkeys()

40 fixed across 10 files

Made-with: Cursor

* fix(ruff/cuda_core): PIE misc cleanup

- PIE808: unnecessary-range-start — remove redundant range(0, n)
- PIE790: unnecessary-placeholder — remove pass after docstring
- PIE810: multiple-starts-ends-with — consolidate into tuple arg

3 fixed across 3 files

Made-with: Cursor

* fix(ruff/cuda_core): PT pytest-style fixes

- PT003: extraneous-scope-function — remove default scope="function"
- PT022: pytest-useless-yield-fixture — convert yield fixture to return
- PT001: pytest-fixture-incorrect-parentheses — remove empty parentheses
- PT013: pytest-incorrect-pytest-import — use pytest.param instead of from pytest import param

18 fixed across 4 files

Made-with: Cursor

* fix(ruff/cuda_core): RUF ruff-specific fixes

- RUF022: unsorted-dunder-all — sort __all__ entries (4 fixes across 4 files)
- RUF023: unsorted-dunder-slots — sort __slots__ entries (3 fixes across 1 file)
- RUF012: mutable-class-default — suppress with noqa for lookup tables (5 fixes in 1 file)

Made-with: Cursor

* fix(ruff/cuda_core): E pycodestyle fixes

- E711: none-comparison — use 'is not None' instead of '!= None' (1 fix in 1 file)

Made-with: Cursor

* fix(ruff/cuda_core): N naming convention fixes

- N806: non-lowercase-variable-in-function — rename CUDA_PATH to cuda_path (2 fixes in 1 file)
- N816: mixed-case-variable-in-global-scope — suppress for CUDA API naming (1 noqa in 1 file)
- N801: invalid-class-name — suppress for setuptools convention (1 noqa in 1 file)

Made-with: Cursor

* fix(ruff/cuda_core): ARG unused argument suppression

- ARG002: unused-method-argument — suppress for interface method (1 noqa in 1 file)

Made-with: Cursor

* fix(ruff/cuda_core): F pyflakes fixes

- F401: unused-import — suppress for re-exports and side-effect imports (4 noqa in 2 files)

Made-with: Cursor

* fix(ruff/cuda_core): B bugbear fixes

- B018: useless-expression — suppress for property access expected to raise (2 noqa in 1 file)

Made-with: Cursor

* fix(ruff/cuda_core): S bandit security suppressions

- S110: try-except-pass — suppress for best-effort cleanup (5 noqa in 2 files)
- S102: exec-builtin — suppress in test code (2 noqa in 2 files)
- S301: suspicious-pickle-usage — suppress in test code (3 noqa in 3 files)

Made-with: Cursor

* fix(ruff/cuda_bindings): I001 sort imports

- I001: unsorted imports — 42 fixed across NFILES files

Made-with: Cursor

* fix(ruff/cuda_bindings): RUF100 remove stale noqa comments

- RUF100: unused-noqa — 16 fixed across 9 files

Made-with: Cursor

* fix(ruff/cuda_bindings): C4 simplify comprehensions

- C416: unnecessary-comprehension — use set()/list() directly
- C401: unnecessary-generator-set — use set comprehension
- C419: unnecessary-comprehension-in-call — use generator in all()/any()
- C400: unnecessary-generator-list — use list comprehension
- C414: unnecessary-double-cast-or-process — remove redundant list()
- C420: unnecessary-dict-comprehension — use dict.fromkeys()

9 fixed across 3 files

Made-with: Cursor

* fix(ruff/cuda_bindings): PIE misc cleanup

- PIE808: unnecessary-range-start — remove redundant range(0, n)
- PIE790: unnecessary-placeholder — remove pass after docstring
- PIE810: multiple-starts-ends-with — consolidate into tuple arg

5 fixed across 4 files

Made-with: Cursor

* fix(ruff/cuda_bindings): PT pytest-style fixes

- PT003: extraneous-scope-function — remove default scope="function" (4 fixes)
- PT001: pytest-fixture-incorrect-parentheses — remove empty parentheses

4 fixed across 1 files

Made-with: Cursor

* fix(ruff/cuda_bindings): F pyflakes fixes

- F401: unused-import — suppress for intentional re-exports and side-effect imports (2 noqa in 2 files)

Made-with: Cursor

* fix(ruff/cuda_bindings): S bandit security suppressions

- S110: try-except-pass — suppress for best-effort cleanup (2 noqa in 1 file)
- S603: subprocess-without-shell-equals-true — suppress at call site (5 noqa in 4 files)
- S607: start-process-with-partial-path — suppress at call site (3 noqa in 2 files)

Made-with: Cursor

* fix(ruff/cuda_pathfinder): RUF100 remove stale noqa comments

- RUF100: unused-noqa — 12 fixed across 7 files

Made-with: Cursor

* fix(ruff/cuda_pathfinder): SIM simplify code

- SIM102: collapsible-if — merge nested if statements (1 fix in 1 file)
- SIM105: suppressible-exception — use contextlib.suppress instead of try/except/pass (1 fix in 1 file)

Made-with: Cursor

* fix(ruff/cuda_pathfinder): S bandit security suppressions

- S110: try-except-pass — suppress for best-effort cleanup (1 noqa in 1 file)

Made-with: Cursor

* fix(ruff/cuda_python_test_helpers): I001 sort imports

- I001: unsorted imports — across 1 files

Made-with: Cursor

* fix(ruff/cuda_python_test_helpers): RUF100 remove stale noqa comments

- RUF100: unused-noqa — 1 fixed across 1 files

Made-with: Cursor

* fix(ruff/cuda_python_test_helpers): RUF ruff-specific fixes

- RUF059: unused-unpacked-variable — prefix unused vars with underscore (2 fixes in 1 file)

Made-with: Cursor

* fix(ruff/toolshed): I001 sort imports

- I001: unsorted imports — across 2 files

Made-with: Cursor

* fix(ruff/toolshed): RUF100 remove stale noqa comments

- RUF100: unused-noqa — 2 fixed across 1 files

Made-with: Cursor

* fix(ruff/toolshed): DTZ datetime timezone fixes

- DTZ011: call-date-today — use timezone-aware datetime.now(tz=utc) (1 fix in 1 file)

Made-with: Cursor

* fix(ruff/toolshed): N naming convention fixes

- N806: non-lowercase-variable-in-function — rename TILE_SIZE to tile_size (1 fix in 1 file)

Made-with: Cursor

* fix(ruff/toolshed): S bandit security suppressions

- S603: subprocess-without-shell-equals-true — suppress at call site (1 noqa in 1 file)
- S607: start-process-with-partial-path — suppress at call site (1 noqa in 1 file)

Made-with: Cursor

* fix(ruff/ci): RUF100 remove stale noqa comments

- RUF100: unused-noqa — 1 fixed across 1 files

Made-with: Cursor

* fix(ruff/ci): RUF ruff-specific fixes

- RUF013: implicit-optional — annotate as T | None explicitly (2 fixes in 1 file)

Made-with: Cursor

* fix(ruff/ci): S bandit security suppressions

- S603: subprocess-without-shell-equals-true — suppress at call site (1 noqa in 1 file)

Made-with: Cursor

* fix(ruff/root): ARG unused argument suppression

- ARG001: unused-function-argument — suppress for pytest hook signature (1 noqa in 1 file)

Made-with: Cursor

* fixup: correct S noqa placement for multiline calls

Made-with: Cursor

* fix(ruff): apply ruff formatting

- 2 files reformatted (cuda_core/cuda/core/system/__init__.py, cuda_core/tests/test_utils.py)

Made-with: Cursor

* fix(ruff): add RUF012 to test/benchmark per-file-ignores

ctypes Structure._fields_ is a standard pattern, not a mutable default concern

Made-with: Cursor

* fix(core): use CUmemAllocationType for VMM allocation enums

Read CU_MEM_ALLOCATION_TYPE_* from CUmemAllocationType instead of CUmemAccess_flags so cuda.core imports correctly in test/docs CI jobs.

Made-with: Cursor

* fix(tests): use pytest.param in test_utils

Switch test parametrization calls to pytest.param so collection succeeds under current pytest and the pixi cuda_core suite runs end-to-end.

Made-with: Cursor

Author: cpcloud

ci/tools/merge_cuda_core_wheels.py

@@ -30,7 +30,7 @@
 from typing import List
 
 
-def run_command(cmd: List[str], cwd: Path = 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:
@@ -47,7 +47,7 @@ def run_command(cmd: List[str], cwd: Path = None, env: dict = os.environ) -> sub
     return result
 
 
-def print_wheel_directory_structure(wheel_path: Path, filter_prefix: str = "cuda/core/", label: str = None):
+def print_wheel_directory_structure(wheel_path: Path, filter_prefix: str = "cuda/core/", label: str | None = None):
     """Print the directory structure of a wheel file, similar to unzip -l output.
 
     Args:

conftest.py

@@ -6,7 +6,7 @@
 import pytest
 
 
-def pytest_collection_modifyitems(config, items):
+def pytest_collection_modifyitems(config, items):  # noqa: ARG001
     cuda_home = os.environ.get("CUDA_HOME")
     for item in items:
         nodeid = item.nodeid.replace("\\", "/")

cuda_bindings/benchmarks/conftest.py

@@ -3,6 +3,7 @@
 
 import numpy as np
 import pytest
+
 from cuda.bindings import driver as cuda
 from cuda.bindings import nvrtc
 from cuda.bindings import runtime as cudart
@@ -22,7 +23,7 @@ def ASSERT_DRV(err):
         raise RuntimeError(f"Unknown error type: {err}")
 
 
-@pytest.fixture(scope="function")
+@pytest.fixture
 def init_cuda():
     # Initialize
     (err,) = cuda.cuInit(0)
@@ -44,7 +45,7 @@ def init_cuda():
     ASSERT_DRV(err)
 
 
-@pytest.fixture(scope="function")
+@pytest.fixture
 def load_module():
     module = None
 

cuda_bindings/benchmarks/test_launch_latency.py

@@ -4,10 +4,10 @@
 import ctypes
 
 import pytest
-from cuda.bindings import driver as cuda
 from kernels import kernel_string
 
 from conftest import ASSERT_DRV
+from cuda.bindings import driver as cuda
 
 
 def launch(kernel, stream, args=(), arg_types=()):

cuda_bindings/benchmarks/test_pointer_attributes.py

@@ -4,9 +4,9 @@
 import random
 
 import pytest
-from cuda.bindings import driver as cuda
 
 from conftest import ASSERT_DRV
+from cuda.bindings import driver as cuda
 
 random.seed(0)
 

cuda_bindings/build_hooks.py

@@ -176,20 +176,20 @@ def _parse_headers(header_dict, include_path_list, parser_caching):
             CUDA_VERSION = parser.defs["macros"].get("CUDA_VERSION", "Unknown")
             print(f"Found CUDA_VERSION: {CUDA_VERSION}", flush=True)
 
-        found_types += {key for key in parser.defs["types"]}
-        found_types += {key for key in parser.defs["structs"]}
-        found_types += {key for key in parser.defs["unions"]}
-        found_types += {key for key in parser.defs["enums"]}
-        found_functions += {key for key in parser.defs["functions"]}
-        found_values += {key for key in parser.defs["values"]}
+        found_types += set(parser.defs["types"])
+        found_types += set(parser.defs["structs"])
+        found_types += set(parser.defs["unions"])
+        found_types += set(parser.defs["enums"])
+        found_functions += set(parser.defs["functions"])
+        found_values += set(parser.defs["values"])
 
         for key, value in parser.defs["structs"].items():
             struct_list[key] = _Struct(key, value["members"])
         for key, value in parser.defs["unions"].items():
             struct_list[key] = _Struct(key, value["members"])
 
         for key, value in struct_list.items():
-            if key.startswith("anon_union") or key.startswith("anon_struct"):
+            if key.startswith(("anon_union", "anon_struct")):
                 continue
 
             found_struct += [key]
@@ -408,7 +408,7 @@ def _cleanup_dst_files():
         )
 
     # Cythonize
-    cython_directives = dict(language_level=3, embedsignature=True, binding=True, freethreading_compatible=True)
+    cython_directives = {"language_level": 3, "embedsignature": True, "binding": True, "freethreading_compatible": True}
     if compile_for_coverage:
         cython_directives["linetrace"] = True
 

cuda_bindings/cuda/bindings/_test_helpers/arch_check.py

@@ -6,6 +6,7 @@
 from functools import cache
 
 import pytest
+
 from cuda.bindings import nvml
 
 

cuda_bindings/examples/0_Introduction/clock_nvrtc_test.py

@@ -6,6 +6,7 @@
 import numpy as np
 from common import common
 from common.helper_cuda import checkCudaErrors, findCudaDevice
+
 from cuda.bindings import driver as cuda
 
 clock_nvrtc = """\
@@ -67,7 +68,7 @@ def main():
     timer = np.empty(NUM_BLOCKS * 2, dtype="int64")
     hinput = np.empty(NUM_THREADS * 2, dtype="float32")
 
-    for i in range(0, NUM_THREADS * 2):
+    for i in range(NUM_THREADS * 2):
         hinput[i] = i
 
     devID = findCudaDevice()
@@ -105,7 +106,7 @@ def main():
 
     avgElapsedClocks = 0.0
 
-    for i in range(0, NUM_BLOCKS):
+    for i in range(NUM_BLOCKS):
         avgElapsedClocks += timer[i + NUM_BLOCKS] - timer[i]
 
     avgElapsedClocks = avgElapsedClocks / NUM_BLOCKS

cuda_bindings/examples/0_Introduction/simpleCubemapTexture_test.py

@@ -8,6 +8,7 @@
 import numpy as np
 from common import common
 from common.helper_cuda import checkCudaErrors, findCudaDevice
+
 from cuda.bindings import driver as cuda
 from cuda.bindings import runtime as cudart
 

cuda_bindings/examples/0_Introduction/simpleP2P_test.py

@@ -8,6 +8,7 @@
 import numpy as np
 from common import common
 from common.helper_cuda import checkCudaErrors
+
 from cuda.bindings import driver as cuda
 from cuda.bindings import runtime as cudart