Merge pull request #949 from suhr25/fix-logging-handler-leak

fix: prevent logging handler accumulation in LoggingHelper

Author: jonbrenas

malariagen_data/util.py

@@ -899,6 +899,12 @@ def __init__(
             handler = logging.FileHandler(out)
         self._handler = handler
 
+        # Remove any pre-existing handlers from the singleton logger to prevent
+        # accumulation (and FileHandler FD leaks) on repeated instantiation.
+        for existing_handler in logger.handlers[:]:
+            logger.removeHandler(existing_handler)
+            existing_handler.close()
+
         # configure handler
         if handler is not None:
             if debug:

tests/anoph/test_base.py

@@ -1,3 +1,6 @@
+import io
+import logging
+
 import numpy as np
 import pandas as pd
 import pytest
@@ -8,6 +11,7 @@
 from malariagen_data import ag3 as _ag3
 from malariagen_data import adir1 as _adir1
 from malariagen_data.anoph.base import AnophelesBase
+from malariagen_data.util import LoggingHelper
 
 
 @pytest.fixture
@@ -258,6 +262,32 @@ def test_lookup_study(fixture, api):
         api.lookup_study("foobar")
 
 
+def test_logging_helper_no_handler_accumulation():
+    # Regression test: repeated LoggingHelper construction on the same logger
+    # name must not accumulate handlers (StreamHandler leak, FileHandler FD leak).
+    logger_name = "test_logging_helper_no_handler_accumulation"
+    for _ in range(10):
+        LoggingHelper(name=logger_name, out=io.StringIO())
+    logger = logging.getLogger(logger_name)
+    assert (
+        len(logger.handlers) <= 1
+    ), f"Handler leak: {len(logger.handlers)} handlers after 10 instantiations"
+
+
+def test_logging_helper_no_duplicate_output():
+    # Regression test: a message emitted after N instantiations must appear
+    # exactly once in the output stream.
+    logger_name = "test_logging_helper_no_duplicate_output"
+    out = io.StringIO()
+    for _ in range(5):
+        helper = LoggingHelper(name=logger_name, out=out)
+    helper.info("sentinel")
+    output = out.getvalue()
+    assert (
+        output.count("sentinel") == 1
+    ), f"Duplicate log output: 'sentinel' appeared {output.count('sentinel')} times"
+
+
 def _strip_terms_of_use_from_manifest(manifest_path):
     """Rewrite a manifest TSV file without terms-of-use columns."""
     df = pd.read_csv(manifest_path, sep="\t")