Add ASAN/UBSAN build to CI (#397)

.clang-tidy

@@ -1 +1,4 @@
 Checks: '-*,cppcoreguidelines-*,-cppcoreguidelines-pro-type-union-access'
+CheckOptions:
+  - key: cppcoreguidelines-avoid-do-while.IgnoreMacros
+    value: true

.github/workflows/unit-test.yml

@@ -73,3 +73,24 @@ jobs:
         env:
           CIBW_TEST_REQUIRES: pytest pytest-xdist[psutil] pandas>=2.0.0 polars narwhals
           MACOSX_DEPLOYMENT_TARGET: "11.0"
+
+  test_sanitizers:
+    name: ASAN/UBSAN test
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@v4
+
+      # Used to host cibuildwheel
+      - uses: actions/setup-python@v3
+      - name: Install dependencies
+        run: python -m pip install nanobind pytest pandas polars narwhals pytest-xdist[psutil] pyarrow
+      - name: Build library
+        run: |
+          cmake -S . -B build -DPANTAB_USE_SANITIZERS=ON -DCMAKE_BUILD_TYPE=RelWithDebugInfo
+          cmake --build build
+      - name: Test library
+        env:
+          ASAN_OPTIONS: "detect_leaks=0"
+        run: |
+          cd build/src
+          LD_PRELOAD="$(gcc -print-file-name=libasan.so)" python -m pytest -m "not skip_asan" -s ../../tests/

CMakeLists.txt

@@ -7,7 +7,7 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON)
 
 if (MSVC)
 else()
-    add_compile_options(-Wall -Wextra -Werror)
+    add_compile_options(-Wall -Wextra)
 endif()
 
 find_package(Python COMPONENTS Interpreter Development.Module REQUIRED)

pyproject.toml

@@ -51,6 +51,9 @@ testpaths = ["tests"]
 addopts = [
     "--import-mode=importlib",
 ]
+markers = [
+  "skip_asan: tests that are known to fail with ASAN",
+]
 
 [tool.mypy]
 [[tool.mypy.overrides]]
@@ -66,6 +69,7 @@ known_first_party = "pantab"
 [tool.cibuildwheel]
 build = "cp39-*64 cp310-*64 cp311-*64 cp312-*64 cp313-*64"
 skip = "*musllinux*"
+manylinux-x86_64-image = "manylinux_2_28"
 
 test-command = "python -m pytest -n auto {project}/tests"
 test-requires = [

tests/test_reader.py

@@ -178,6 +178,7 @@ def test_reader_accepts_process_params(tmp_hyper):
     pt.frames_from_hyper(tmp_hyper, process_params=params)
 
 
+@pytest.mark.skip_asan
 def test_reader_invalid_process_params_raises(tmp_hyper):
     frame = pd.DataFrame(list(range(10)), columns=["nums"]).astype("int8")
     pt.frame_to_hyper(frame, tmp_hyper, table="test")

tests/test_writer.py

@@ -24,6 +24,7 @@ def test_bad_table_mode_raises(frame, tmp_hyper):
         pt.frames_to_hyper({"a": frame}, tmp_hyper, table_mode="x")
 
 
+@pytest.mark.skip_asan
 @pytest.mark.parametrize(
     "new_dtype,hyper_type_name",
     [(nw.Int64, "BIGINT"), (nw.Float64, "DOUBLE PRECISION")],
@@ -51,6 +52,7 @@ def test_append_mode_raises_column_dtype_mismatch(
         )
 
 
+@pytest.mark.skip_asan
 def test_append_mode_raises_ncolumns_mismatch(frame, tmp_hyper, table_name, compat):
     pt.frame_to_hyper(
         frame,
@@ -173,6 +175,7 @@ def test_string_type_to_existing_varchar(frame, tmp_hyper, compat):
     pt.frame_to_hyper(frame, tmp_hyper, table=table_name, table_mode="a")
 
 
+@pytest.mark.skip_asan
 def test_failed_write_doesnt_overwrite_file(
     frame, tmp_hyper, monkeypatch, table_mode, compat
 ):
@@ -266,6 +269,7 @@ def test_atomic_keyword_does_not_copy_or_move(
     mocked_move.assert_not_called()
 
 
+@pytest.mark.skip_asan
 def test_duplicate_columns_raises(tmp_hyper):
     frame = pd.DataFrame([[1, 1]], columns=[1, 1])
     msg = r"Duplicate column names found: \[1, 1\]"
@@ -276,6 +280,7 @@ def test_duplicate_columns_raises(tmp_hyper):
         pt.frames_to_hyper({"test": frame}, tmp_hyper)
 
 
+@pytest.mark.skip_asan
 def test_unsupported_dtype_raises(tmp_hyper):
     frame = pd.DataFrame([[pd.Timedelta("1D")]])
 
@@ -508,6 +513,7 @@ def test_writer_accepts_process_params(tmp_hyper):
     pt.frame_to_hyper(frame, tmp_hyper, table="test", process_params=params)
 
 
+@pytest.mark.skip_asan
 def test_writer_invalid_process_params_raises(tmp_hyper):
     frame = pd.DataFrame(list(range(10)), columns=["nums"]).astype("int8")
     params = {"not_a_real_parameter": "0"}