fix: the SIGTERM handler is now opt-in. #1310

CHANGES.rst

@@ -20,9 +20,16 @@ development at the same time, such as 4.5.x and 5.0.
 Unreleased
 ----------
 
+- A new setting, :ref:`config_run_sigterm`, controls whether a SIGTERM signal
+  handler is used.  In 6.3, the signal handler was always installed, to capture
+  data at unusual process ends.  Unfortunately, this introduced other problems
+  (see `issue 1310`_).  Now the signal handler is only used if you opt-in by
+  setting ``[run] sigterm = true``.
+
 - On Python 3.11, the ``[toml]`` extra no longer installs tomli, instead using
   tomllib from the standard library.  Thanks `Shantanu <pull 1359_>`_.
 
+.. _issue 1310: https://github.com/nedbat/coveragepy/issues/1310
 .. _pull 1359: https://github.com/nedbat/coveragepy/pull/1359
 
 

coverage/config.py

@@ -190,6 +190,7 @@ def __init__(self):
         self.relative_files = False
         self.run_include = None
         self.run_omit = None
+        self.sigterm = False
         self.source = None
         self.source_pkgs = []
         self.timid = False
@@ -364,6 +365,7 @@ def copy(self):
         ('relative_files', 'run:relative_files', 'boolean'),
         ('run_include', 'run:include', 'list'),
         ('run_omit', 'run:omit', 'list'),
+        ('sigterm', 'run:sigterm', 'boolean'),
         ('source', 'run:source', 'list'),
         ('source_pkgs', 'run:source_pkgs', 'list'),
         ('timid', 'run:timid', 'boolean'),

coverage/control.py

@@ -536,12 +536,13 @@ def _init_for_start(self):
 
         # Register our clean-up handlers.
         atexit.register(self._atexit)
-        is_main = (threading.current_thread() == threading.main_thread())
-        if is_main and not env.WINDOWS:
-            # The Python docs seem to imply that SIGTERM works uniformly even
-            # on Windows, but that's not my experience, and this agrees:
-            # https://stackoverflow.com/questions/35772001/x/35792192#35792192
-            self._old_sigterm = signal.signal(signal.SIGTERM, self._on_sigterm)
+        if self.config.sigterm:
+            is_main = (threading.current_thread() == threading.main_thread())
+            if is_main and not env.WINDOWS:
+                # The Python docs seem to imply that SIGTERM works uniformly even
+                # on Windows, but that's not my experience, and this agrees:
+                # https://stackoverflow.com/questions/35772001/x/35792192#35792192
+                self._old_sigterm = signal.signal(signal.SIGTERM, self._on_sigterm)
 
     def _init_data(self, suffix):
         """Create a data file if we don't have one yet."""

doc/config.rst

@@ -268,6 +268,24 @@ need to know the source origin.
 .. versionadded:: 5.0
 
 
+.. _config_run_sigterm:
+
+[run] sigterm
+.............
+
+(boolean, default False) if true, register a SIGTERM signal handler to capture
+data when the process ends due to a SIGTERM signal.  This includes
+:meth:`Process.terminate <python:multiprocessing.Process.terminate>`, and other
+ways to terminate a process.  This can help when collecting data in usual
+situations, but can also introduce problems (see `issue 1310`_).
+
+Only on Linux and Mac.
+
+.. _issue 1310: https://github.com/nedbat/coveragepy/issues/1310
+
+.. versionadded:: 6.4 (in 6.3 this was always enabled)
+
+
 .. _config_run_source:
 
 [run] source

tests/test_concurrency.py

@@ -696,7 +696,8 @@ def random_load():                              # pragma: nested
 class SigtermTest(CoverageTest):
     """Tests of our handling of SIGTERM."""
 
-    def test_sigterm_saves_data(self):
+    @pytest.mark.parametrize("sigterm", [False, True])
+    def test_sigterm_saves_data(self, sigterm):
         # A terminated process should save its coverage data.
         self.make_file("clobbered.py", """\
             import multiprocessing
@@ -724,7 +725,8 @@ def subproc(x):
             [run]
             parallel = True
             concurrency = multiprocessing
-            """)
+            """ + ("sigterm = true" if sigterm else "")
+            )
         out = self.run_command("coverage run clobbered.py")
         # Under the Python tracer on Linux, we get the "Trace function changed"
         # message. Does that matter?
@@ -735,7 +737,11 @@ def subproc(x):
         assert out == "START\nNOT THREE\nEND\n"
         self.run_command("coverage combine")
         out = self.run_command("coverage report -m")
-        assert self.squeezed_lines(out)[2] == "clobbered.py 17 1 94% 6"
+        if sigterm:
+            expected = "clobbered.py 17 1 94% 6"
+        else:
+            expected = "clobbered.py 17 5 71% 5-10"
+        assert self.squeezed_lines(out)[2] == expected
 
     def test_sigterm_still_runs(self):
         # A terminated process still runs its own SIGTERM handler.
@@ -766,6 +772,7 @@ def on_sigterm(signum, frame):
             [run]
             parallel = True
             concurrency = multiprocessing
+            sigterm = True
             """)
         out = self.run_command("coverage run handler.py")
         assert out == "START\nSIGTERM\nEND\n"