TST: Disable testing that hangs Azure CI

ci/run_tests.sh

@@ -24,7 +24,7 @@ if [[ $(uname) == "Linux" && -z $DISPLAY ]]; then
     XVFB="xvfb-run "
 fi
 
-PYTEST_CMD="${XVFB}pytest -r fEs -m \"$PATTERN\" -n $PYTEST_WORKERS --dist=loadfile $TEST_ARGS $COVERAGE $PYTEST_TARGET"
+PYTEST_CMD="${XVFB}pytest -v -r fEs -m \"$PATTERN\" -n $PYTEST_WORKERS --dist=loadfile $TEST_ARGS $COVERAGE $PYTEST_TARGET"
 
 if [[ $(uname) != "Linux"  && $(uname) != "Darwin" ]]; then
     PYTEST_CMD="$PYTEST_CMD --ignore=pandas/tests/plotting/"

pandas/tests/io/excel/conftest.py

@@ -1,5 +1,6 @@
 import pytest
 
+from pandas.compat import is_platform_windows
 import pandas.util._test_decorators as td
 
 import pandas._testing as tm
@@ -43,7 +44,8 @@ def read_ext(request):
     return request.param
 
 
-@pytest.fixture(autouse=True)
+# Checking for file leaks can hang on Windows CI
+@pytest.fixture(autouse=not is_platform_windows())
 def check_for_file_leaks():
     """
     Fixture to run around every test to ensure that we are not leaking files.

pandas/tests/io/excel/test_readers.py

@@ -150,9 +150,7 @@ def parser(self, *args, **kwargs):
             expected = expected_defaults[read_ext[1:]]
         assert result == expected
 
-    def test_usecols_int(self, read_ext, df_ref):
-        df_ref = df_ref.reindex(columns=["A", "B", "C"])
-
+    def test_usecols_int(self, read_ext):
         # usecols as int
         msg = "Passing an integer for `usecols`"
         with pytest.raises(ValueError, match=msg):
@@ -1311,7 +1309,7 @@ def cd_and_set_engine(self, engine, datapath, monkeypatch):
         monkeypatch.chdir(datapath("io", "data", "excel"))
         monkeypatch.setattr(pd, "ExcelFile", func)
 
-    def test_engine_used(self, read_ext, engine, monkeypatch):
+    def test_engine_used(self, read_ext, engine):
         expected_defaults = {
             "xlsx": "openpyxl",
             "xlsm": "openpyxl",

pandas/tests/io/parser/conftest.py

@@ -12,6 +12,9 @@
 )
 import pandas._testing as tm
 
+from pandas.io.parsers import readers
+from pandas.io.parsers.arrow_parser_wrapper import ArrowParserWrapper
+
 
 class BaseParser:
     engine: str | None = None
@@ -101,18 +104,29 @@ def csv1(datapath):
 
 
 @pytest.fixture(params=_all_parsers, ids=_all_parser_ids)
-def all_parsers(request):
+def all_parsers(request, monkeypatch):
     """
     Fixture all of the CSV parsers.
     """
     parser = request.param()
     if parser.engine == "pyarrow":
         pytest.importorskip("pyarrow", VERSIONS["pyarrow"])
         # Try setting num cpus to 1 to avoid hangs?
+        # Disable threads in CI environment to avoid timeouts
         import pyarrow
 
         pyarrow.set_cpu_count(1)
-    return parser
+
+        class NoThreadArrowParserWrapper(ArrowParserWrapper):
+            def _get_pyarrow_options(self):
+                super()._get_pyarrow_options()
+                self.read_options["use_threads"] = False
+
+        with monkeypatch.context() as m:
+            m.setattr(readers, "ArrowParserWrapper", NoThreadArrowParserWrapper)
+            yield parser
+    else:
+        yield parser
 
 
 @pytest.fixture(params=_c_parsers_only, ids=_c_parser_ids)

pandas/tests/io/test_sql.py

@@ -498,6 +498,7 @@ def sqlite_buildin_iris(sqlite_buildin, iris_path):
 all_connectable_iris = sqlalchemy_connectable_iris + ["sqlite_buildin_iris"]
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", all_connectable)
 @pytest.mark.parametrize("method", [None, "multi"])
 def test_to_sql(conn, method, test_frame1, request):
@@ -508,6 +509,7 @@ def test_to_sql(conn, method, test_frame1, request):
     assert count_rows(conn, "test_frame") == len(test_frame1)
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", all_connectable)
 @pytest.mark.parametrize("mode, num_row_coef", [("replace", 1), ("append", 2)])
 def test_to_sql_exist(conn, mode, num_row_coef, test_frame1, request):
@@ -519,6 +521,7 @@ def test_to_sql_exist(conn, mode, num_row_coef, test_frame1, request):
     assert count_rows(conn, "test_frame") == num_row_coef * len(test_frame1)
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", all_connectable)
 def test_to_sql_exist_fail(conn, test_frame1, request):
     conn = request.getfixturevalue(conn)
@@ -531,6 +534,7 @@ def test_to_sql_exist_fail(conn, test_frame1, request):
         pandasSQL.to_sql(test_frame1, "test_frame", if_exists="fail")
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", all_connectable_iris)
 def test_read_iris(conn, request):
     conn = request.getfixturevalue(conn)
@@ -539,6 +543,7 @@ def test_read_iris(conn, request):
     check_iris_frame(iris_frame)
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", sqlalchemy_connectable)
 def test_to_sql_callable(conn, test_frame1, request):
     conn = request.getfixturevalue(conn)
@@ -557,6 +562,7 @@ def sample(pd_table, conn, keys, data_iter):
     assert count_rows(conn, "test_frame") == len(test_frame1)
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", mysql_connectable)
 def test_default_type_conversion(conn, request):
     conn = request.getfixturevalue(conn)
@@ -575,6 +581,7 @@ def test_default_type_conversion(conn, request):
     assert issubclass(df.BoolColWithNull.dtype.type, np.floating)
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", mysql_connectable)
 def test_read_procedure(conn, request):
     conn = request.getfixturevalue(conn)
@@ -611,6 +618,7 @@ def test_read_procedure(conn, request):
     tm.assert_frame_equal(df, res2)
 
 
+@pytest.mark.db
 @pytest.mark.parametrize("conn", postgresql_connectable)
 def test_copy_from_callable_insertion_method(conn, request):
     # GH 8953

pandas/tests/window/test_numba.py

@@ -1,6 +1,12 @@
+import os
+
 import numpy as np
 import pytest
 
+from pandas.compat import (
+    is_platform_mac,
+    is_platform_windows,
+)
 from pandas.errors import NumbaUtilError
 import pandas.util._test_decorators as td
 
@@ -13,6 +19,15 @@
 import pandas._testing as tm
 from pandas.core.util.numba_ import NUMBA_FUNC_CACHE
 
+# TODO: Mark these as pytest.mark.single GH 44584
+pytestmark = pytest.mark.skipif(
+    os.environ.get("PANDAS_CI", "0") == "1"
+    and (is_platform_windows() or is_platform_mac()),
+    reason="On Azure CI, Windows can fail with "
+    "'Windows fatal exception: stack overflow' "
+    "and MacOS can timeout",
+)
+
 
 @pytest.fixture(params=["single", "table"])
 def method(request):

pandas/tests/window/test_online.py

@@ -1,6 +1,12 @@
+import os
+
 import numpy as np
 import pytest
 
+from pandas.compat import (
+    is_platform_mac,
+    is_platform_windows,
+)
 import pandas.util._test_decorators as td
 
 from pandas import (
@@ -9,6 +15,15 @@
 )
 import pandas._testing as tm
 
+# TODO: Mark these as pytest.mark.single GH 44584
+pytestmark = pytest.mark.skipif(
+    os.environ.get("PANDAS_CI", "0") == "1"
+    and (is_platform_windows() or is_platform_mac()),
+    reason="On Azure CI, Windows can fail with "
+    "'Windows fatal exception: stack overflow' "
+    "and MacOS can timeout",
+)
+
 
 @td.skip_if_no("numba")
 @pytest.mark.filterwarnings("ignore:\n")

pandas/util/_test_decorators.py

@@ -266,7 +266,8 @@ def file_leak_context():
     ContextManager analogue to check_file_leaks.
     """
     psutil = safe_import("psutil")
-    if not psutil:
+    if not psutil or is_platform_windows():
+        # Checking for file leaks can hang on Windows CI
         yield
     else:
         proc = psutil.Process()