Use multiprocessing start_method "forkserver" (#4306)

* Use `multiprocessing` `start_method` `"forkserver"`

Alternative to PR #4305

* Add link to comment under PR #4105

* Unconditionally `pytest.skip("DEADLOCK")` for PyPy Windows

* Remove `SKIP_IF_DEADLOCK` entirely, for simplicity. Hopefully this PR will resolve the deadlocks for good.

* Add "In a nutshell" comment, in response to request by @EricCousineau-TRI

tests/conftest.py

@@ -7,6 +7,8 @@
 import contextlib
 import difflib
 import gc
+import multiprocessing
+import os
 import re
 import textwrap
 
@@ -15,6 +17,16 @@
 # Early diagnostic for failed imports
 import pybind11_tests
 
+if os.name != "nt":
+    # Full background: https://github.com/pybind/pybind11/issues/4105#issuecomment-1301004592
+    # In a nutshell: fork() after starting threads == flakiness in the form of deadlocks.
+    # It is actually a well-known pitfall, unfortunately without guard rails.
+    # "forkserver" is more performant than "spawn" (~9s vs ~13s for tests/test_gil_scoped.py,
+    # visit the issuecomment link above for details).
+    # Windows does not have fork() and the associated pitfall, therefore it is best left
+    # running with defaults.
+    multiprocessing.set_start_method("forkserver")
+
 _long_marker = re.compile(r"([0-9])L")
 _hexadecimal = re.compile(r"0x[0-9a-fA-F]+")
 

tests/test_gil_scoped.py

@@ -5,6 +5,7 @@
 
 import pytest
 
+import env
 from pybind11_tests import gil_scoped as m
 
 
@@ -144,7 +145,6 @@ def _intentional_deadlock():
 
 
 ALL_BASIC_TESTS_PLUS_INTENTIONAL_DEADLOCK = ALL_BASIC_TESTS + (_intentional_deadlock,)
-SKIP_IF_DEADLOCK = True  # See PR #4216
 
 
 def _run_in_process(target, *args, **kwargs):
@@ -181,7 +181,7 @@ def _run_in_process(target, *args, **kwargs):
         elif process.exitcode is None:
             assert t_delta > 0.9 * timeout
             msg = "DEADLOCK, most likely, exactly what this test is meant to detect."
-            if SKIP_IF_DEADLOCK:
+            if env.PYPY and env.WIN:
                 pytest.skip(msg)
             raise RuntimeError(msg)
         return process.exitcode