bpo-34022: Stop forcing of hash-based invalidation with SOURCE_DATE_EPOCH (GH-9607)


Unconditional forcing of ``CHECKED_HASH`` invalidation was introduced in
3.7.0 in bpo-29708.  The change is bad, as it unconditionally overrides
*invalidation_mode*, even if it was passed as an explicit argument to
``py_compile.compile()`` or ``compileall``.  An environment variable
should *never* override an explicit argument to a library function.
That change leads to multiple test failures if the ``SOURCE_DATE_EPOCH``
environment variable is set.

This changes ``py_compile.compile()`` to only look at
``SOURCE_DATE_EPOCH`` if no explicit *invalidation_mode* was specified.
I also made various relevant tests run with explicit control over the
value of ``SOURCE_DATE_EPOCH``.

While looking at this, I noticed that ``zipimport`` does not work
with hash-based .pycs _at all_, though I left the fixes for
subsequent commits.
(cherry picked from commit a6b3ec5b6d4f6387820fccc570eea08b9615620d)

Co-authored-by: Elvis Pranskevichus <elvis@magic.io>
diff --git a/Lib/test/test_compileall.py b/Lib/test/test_compileall.py
index 2995e08..2e25523 100644
--- a/Lib/test/test_compileall.py
+++ b/Lib/test/test_compileall.py
@@ -22,7 +22,11 @@
 from test import support
 from test.support import script_helper
 
-class CompileallTests(unittest.TestCase):
+from .test_py_compile import without_source_date_epoch
+from .test_py_compile import SourceDateEpochTestMeta
+
+
+class CompileallTestsBase:
 
     def setUp(self):
         self.directory = tempfile.mkdtemp()
@@ -46,7 +50,7 @@
         with open(self.bad_source_path, 'w') as file:
             file.write('x (\n')
 
-    def data(self):
+    def timestamp_metadata(self):
         with open(self.bc_path, 'rb') as file:
             data = file.read(12)
         mtime = int(os.stat(self.source_path).st_mtime)
@@ -57,16 +61,18 @@
     def recreation_check(self, metadata):
         """Check that compileall recreates bytecode when the new metadata is
         used."""
+        if os.environ.get('SOURCE_DATE_EPOCH'):
+            raise unittest.SkipTest('SOURCE_DATE_EPOCH is set')
         py_compile.compile(self.source_path)
-        self.assertEqual(*self.data())
+        self.assertEqual(*self.timestamp_metadata())
         with open(self.bc_path, 'rb') as file:
             bc = file.read()[len(metadata):]
         with open(self.bc_path, 'wb') as file:
             file.write(metadata)
             file.write(bc)
-        self.assertNotEqual(*self.data())
+        self.assertNotEqual(*self.timestamp_metadata())
         compileall.compile_dir(self.directory, force=False, quiet=True)
-        self.assertTrue(*self.data())
+        self.assertTrue(*self.timestamp_metadata())
 
     def test_mtime(self):
         # Test a change in mtime leads to a new .pyc.
@@ -189,6 +195,21 @@
         compileall.compile_dir(self.directory, quiet=True, workers=5)
         self.assertTrue(compile_file_mock.called)
 
+
+class CompileallTestsWithSourceEpoch(CompileallTestsBase,
+                                     unittest.TestCase,
+                                     metaclass=SourceDateEpochTestMeta,
+                                     source_date_epoch=True):
+    pass
+
+
+class CompileallTestsWithoutSourceEpoch(CompileallTestsBase,
+                                        unittest.TestCase,
+                                        metaclass=SourceDateEpochTestMeta,
+                                        source_date_epoch=False):
+    pass
+
+
 class EncodingTest(unittest.TestCase):
     """Issue 6716: compileall should escape source code when printing errors
     to stdout."""
@@ -212,7 +233,7 @@
             sys.stdout = orig_stdout
 
 
-class CommandLineTests(unittest.TestCase):
+class CommandLineTestsBase:
     """Test compileall's CLI."""
 
     @classmethod
@@ -285,6 +306,7 @@
         self.assertNotCompiled(self.initfn)
         self.assertNotCompiled(self.barfn)
 
+    @without_source_date_epoch  # timestamp invalidation test
     def test_no_args_respects_force_flag(self):
         self._skip_if_sys_path_not_writable()
         bazfn = script_helper.make_script(self.directory, 'baz', '')
@@ -353,6 +375,7 @@
         self.assertTrue(os.path.exists(self.pkgdir_cachedir))
         self.assertFalse(os.path.exists(cachecachedir))
 
+    @without_source_date_epoch  # timestamp invalidation test
     def test_force(self):
         self.assertRunOK('-q', self.pkgdir)
         pycpath = importlib.util.cache_from_source(self.barfn)
@@ -556,5 +579,20 @@
             self.assertEqual(compile_dir.call_args[-1]['workers'], None)
 
 
+class CommmandLineTestsWithSourceEpoch(CommandLineTestsBase,
+                                       unittest.TestCase,
+                                       metaclass=SourceDateEpochTestMeta,
+                                       source_date_epoch=True):
+    pass
+
+
+class CommmandLineTestsNoSourceEpoch(CommandLineTestsBase,
+                                     unittest.TestCase,
+                                     metaclass=SourceDateEpochTestMeta,
+                                     source_date_epoch=False):
+    pass
+
+
+
 if __name__ == "__main__":
     unittest.main()
diff --git a/Lib/test/test_importlib/source/test_file_loader.py b/Lib/test/test_importlib/source/test_file_loader.py
index cc80f26..c916d7c 100644
--- a/Lib/test/test_importlib/source/test_file_loader.py
+++ b/Lib/test/test_importlib/source/test_file_loader.py
@@ -19,6 +19,9 @@
 
 from test.support import make_legacy_pyc, unload
 
+from test.test_py_compile import without_source_date_epoch
+from test.test_py_compile import SourceDateEpochTestMeta
+
 
 class SimpleTest(abc.LoaderTests):
 
@@ -359,6 +362,17 @@
                     abc=importlib_abc, util=importlib_util)
 
 
+class SourceDateEpochTestMeta(SourceDateEpochTestMeta,
+                              type(Source_SimpleTest)):
+    pass
+
+
+class SourceDateEpoch_SimpleTest(Source_SimpleTest,
+                                 metaclass=SourceDateEpochTestMeta,
+                                 source_date_epoch=True):
+    pass
+
+
 class BadBytecodeTest:
 
     def import_(self, file, module_name):
@@ -617,6 +631,7 @@
 
     # [bad timestamp]
     @util.writes_bytecode_files
+    @without_source_date_epoch
     def test_old_timestamp(self):
         # When the timestamp is older than the source, bytecode should be
         # regenerated.
diff --git a/Lib/test/test_py_compile.py b/Lib/test/test_py_compile.py
index 8fc0b33..f86abe2 100644
--- a/Lib/test/test_py_compile.py
+++ b/Lib/test/test_py_compile.py
@@ -1,3 +1,4 @@
+import functools
 import importlib.util
 import os
 import py_compile
@@ -10,7 +11,44 @@
 from test import support
 
 
-class PyCompileTests(unittest.TestCase):
+def without_source_date_epoch(fxn):
+    """Runs function with SOURCE_DATE_EPOCH unset."""
+    @functools.wraps(fxn)
+    def wrapper(*args, **kwargs):
+        with support.EnvironmentVarGuard() as env:
+            env.unset('SOURCE_DATE_EPOCH')
+            return fxn(*args, **kwargs)
+    return wrapper
+
+
+def with_source_date_epoch(fxn):
+    """Runs function with SOURCE_DATE_EPOCH set."""
+    @functools.wraps(fxn)
+    def wrapper(*args, **kwargs):
+        with support.EnvironmentVarGuard() as env:
+            env['SOURCE_DATE_EPOCH'] = '123456789'
+            return fxn(*args, **kwargs)
+    return wrapper
+
+
+# Run tests with SOURCE_DATE_EPOCH set or unset explicitly.
+class SourceDateEpochTestMeta(type(unittest.TestCase)):
+    def __new__(mcls, name, bases, dct, *, source_date_epoch):
+        cls = super().__new__(mcls, name, bases, dct)
+
+        for attr in dir(cls):
+            if attr.startswith('test_'):
+                meth = getattr(cls, attr)
+                if source_date_epoch:
+                    wrapper = with_source_date_epoch(meth)
+                else:
+                    wrapper = without_source_date_epoch(meth)
+                setattr(cls, attr, wrapper)
+
+        return cls
+
+
+class PyCompileTestsBase:
 
     def setUp(self):
         self.directory = tempfile.mkdtemp()
@@ -99,16 +137,18 @@
             importlib.util.cache_from_source(bad_coding)))
 
     def test_source_date_epoch(self):
-        testtime = 123456789
-        with support.EnvironmentVarGuard() as env:
-            env["SOURCE_DATE_EPOCH"] = str(testtime)
-            py_compile.compile(self.source_path, self.pyc_path)
+        py_compile.compile(self.source_path, self.pyc_path)
         self.assertTrue(os.path.exists(self.pyc_path))
         self.assertFalse(os.path.exists(self.cache_path))
         with open(self.pyc_path, 'rb') as fp:
             flags = importlib._bootstrap_external._classify_pyc(
                 fp.read(), 'test', {})
-        self.assertEqual(flags, 0b11)
+        if os.environ.get('SOURCE_DATE_EPOCH'):
+            expected_flags = 0b11
+        else:
+            expected_flags = 0b00
+
+        self.assertEqual(flags, expected_flags)
 
     @unittest.skipIf(sys.flags.optimize > 0, 'test does not work with -O')
     def test_double_dot_no_clobber(self):
@@ -153,5 +193,19 @@
         self.assertEqual(flags, 0b1)
 
 
+class PyCompileTestsWithSourceEpoch(PyCompileTestsBase,
+                                    unittest.TestCase,
+                                    metaclass=SourceDateEpochTestMeta,
+                                    source_date_epoch=True):
+    pass
+
+
+class PyCompileTestsWithoutSourceEpoch(PyCompileTestsBase,
+                                       unittest.TestCase,
+                                       metaclass=SourceDateEpochTestMeta,
+                                       source_date_epoch=False):
+    pass
+
+
 if __name__ == "__main__":
     unittest.main()