Use external psleak module for memleak tests (#2698)

...and get rid of internal psutil.test.MemleakTestCase framework.

Author: giampaolo

MANIFEST.in

@@ -137,8 +137,6 @@ include psutil/arch/windows/services.c
 include psutil/arch/windows/socks.c
 include psutil/arch/windows/sys.c
 include psutil/arch/windows/wmi.c
-include psutil/test/__init__.py
-include psutil/test/memleak.py
 include pyproject.toml
 include scripts/battery.py
 include scripts/cpu_distribution.py

Makefile

@@ -76,6 +76,7 @@ install-sysdeps:
 install-pydeps-test:  ## Install python deps necessary to run unit tests.
 	$(MAKE) install-pip
 	PIP_BREAK_SYSTEM_PACKAGES=1 $(PYTHON) -m pip install $(PIP_INSTALL_ARGS) `$(PYTHON) -c "import setup; print(' '.join(setup.TEST_DEPS))"`
+	PIP_BREAK_SYSTEM_PACKAGES=1 $(PYTHON) -m pip install git+https://github.com/giampaolo/psleak.git
 
 install-pydeps-dev:  ## Install python deps meant for local development.
 	$(MAKE) install-git-hooks
@@ -136,7 +137,7 @@ test-platform:  ## Run specific platform tests only.
 	$(RUN_TEST) tests/test_`$(PYTHON) -c 'import psutil; print([x.lower() for x in ("LINUX", "BSD", "OSX", "SUNOS", "WINDOWS", "AIX") if getattr(psutil, x)][0])'`.py $(ARGS)
 
 test-memleaks:  ## Memory leak tests.
-	$(RUN_TEST) tests/test_memleaks.py $(ARGS)
+	PYTHONMALLOC=malloc $(RUN_TEST) -k test_memleaks.py $(ARGS)
 
 test-sudo:  ## Run tests requiring root privileges.
 	# Use unittest runner because pytest may not be installed as root.
@@ -246,7 +247,7 @@ ci-test-cibuildwheel:  ## Run CI tests for the built wheels.
 	mkdir -p .tests
 	cp -r tests .tests/
 	cd .tests/ && PYTHONPATH=$$(pwd) $(PYTHON_ENV_VARS) $(PYTHON) -m pytest -k "not test_memleaks.py"
-	cd .tests/ && PYTHONPATH=$$(pwd) $(PYTHON_ENV_VARS) $(PYTHON) -m pytest -k "test_memleaks.py"
+	cd .tests/ && PYTHONPATH=$$(pwd) $(PYTHON_ENV_VARS) PYTHONMALLOC=malloc $(PYTHON) -m pytest -k "test_memleaks.py"
 
 ci-check-dist:  ## Run all sanity checks re. to the package distribution.
 	$(PYTHON) -m pip install -U setuptools virtualenv twine check-manifest validate-pyproject[all] abi3audit

README.rst

@@ -464,17 +464,6 @@ Heap info
     pheap(heap_used=5177792, mmap_used=819200)
     >>> psutil.heap_trim()
 
-Detecting memory leaks in C functions
--------------------------------------
-
-.. code-block:: python
-
-    from psutil.test import MemoryLeakTestCase
-
-    class TestLeaks(MemoryLeakTestCase):
-        def test_fun(self):
-            self.execute(some_function)
-
 Windows services
 ----------------
 

docs/index.rst

@@ -2299,80 +2299,6 @@ Example code:
    'status': 'stopped',
    'username': 'NT AUTHORITY\\LocalService'}
 
-Testing utilities
-=================
-
-The ``psutil.test`` subpackage includes a helper class to assist in writing
-memory-leak detection tests.
-
-.. class:: psutil.test.MemoryLeakTestCase
-
-  A testing framework for detecting memory leaks in functions, typically those
-  implemented in C that forget to ``free()`` heap memory, call ``Py_DECREF`` on
-  Python objects, and so on. It works by comparing the process's memory usage
-  before and after repeatedly calling the target function.
-
-  Detecting memory leaks reliably is inherently difficult (and probably
-  impossible) because of how the OS manages memory, garbage collection, and
-  caching. Memory usage may even decrease between runs. So this is not meant to
-  be bullet proof. To reduce false positives, when an increase in memory is
-  detected (mem > 0), the test is retried up to 5 times, increasing the
-  number of function calls each time. If memory continues to grow, the test is
-  considered a failure.
-  The test currently monitors RSS, VMS, and `USS <https://gmpy.dev/blog/2016/real-process-memory-and-environ-in-python>`__ memory.
-  On supported platforms, it also monitors **heap metrics** (``heap_used``, ``mmap_used`` from
-  :func:`heap_info`).
-
-  In addition it also ensures that the target function does not leak
-  file descriptors (UNIX) or handles (Windows).
-
-  .. versionadded:: 7.2.0
-
-  .. warning::
-    This class is experimental, meaning its API or internal algorithm may
-    change in the future.
-
-  Usage example::
-
-    from psutil.test import MemoryLeakTestCase
-
-    class TestLeaks(MemoryLeakTestCase):
-        def test_fun(self):
-            self.execute(some_function)
-
-  Class attributes and methods:
-
-  .. attribute:: times
-    :value: 200
-
-    Number of times to call the tested function in each iteration.
-
-  .. attribute:: retries
-    :value: 5
-
-    Maximum number of retries if memory growth is detected.
-
-  .. attribute:: warmup_times
-    :value: 10
-
-    Number of warm-up calls before measurements begin.
-
-  .. attribute:: tolerance
-    :value: 0
-
-    Allowed memory difference (in bytes) before considering it a leak.
-
-  .. attribute:: verbosity
-    :value: 1
-
-    0 = no messages; 1 = print diagnostics when memory increases during the
-    test run.
-
-  .. method:: execute(fun, *, times=None, warmup_times=None, retries=None, tolerance=None)
-
-    Run a full leak test on a callable. If specified, the optional arguments
-    override the class attributes with the same name.
-
 Constants
 =========