MNT Configures boundscheck=True for one of the CI instances (scikit-learn#21779)

thomasjpfan · web-flow · commit a611fed55fd3 · 2021-11-25T14:43:30.000+01:00
diff --git a/azure-pipelines.yml b/azure-pipelines.yml
@@ -200,6 +200,7 @@ jobs:
         SCIPY_VERSION: 'min'
         MATPLOTLIB_VERSION: 'min'
         THREADPOOLCTL_VERSION: '2.2.0'
+        SKLEARN_ENABLE_DEBUG_CYTHON_DIRECTIVES: '1'
       # Linux environment to test the latest available dependencies and MKL.
       # It runs tests requiring lightgbm, pandas and PyAMG.
       pylatest_pip_openblas_pandas:
diff --git a/doc/computing/parallelism.rst b/doc/computing/parallelism.rst
@@ -214,3 +214,9 @@ These environment variables should be set before importing scikit-learn.
     When this environment variable is set to a non zero value, the tests
     that need network access are skipped. When this environment variable is
     not set then network tests are skipped.
+
+:SKLEARN_ENABLE_DEBUG_CYTHON_DIRECTIVES:
+
+    When this environment variable is set to a non zero value, the `Cython`
+    derivative, `boundscheck` is set to `True`. This is useful for finding
+    segfaults.
diff --git a/sklearn/_build_utils/__init__.py b/sklearn/_build_utils/__init__.py
@@ -70,6 +70,11 @@ def cythonize_extensions(top_path, config):
             # CPU particularly in CI (cf loky#114)
             n_jobs = joblib.cpu_count()
 
+    # Additional checks for Cython
+    cython_enable_debug_directives = (
+        os.environ.get("SKLEARN_ENABLE_DEBUG_CYTHON_DIRECTIVES", "0") != "0"
+    )
+
     config.ext_modules = cythonize(
         config.ext_modules,
         nthreads=n_jobs,
@@ -78,7 +83,7 @@ def cythonize_extensions(top_path, config):
         },
         compiler_directives={
             "language_level": 3,
-            "boundscheck": False,
+            "boundscheck": cython_enable_debug_directives,
             "wraparound": False,
             "initializedcheck": False,
             "nonecheck": False,
