gh-133885: Use locks instead of critical sections for _zstd (gh-134289)

Move from using critical sections to locks for the (de)compression methods.
Since the methods allow other threads to run, we should use a lock rather
than a critical section.

Author: emmatyping

Lib/test/test_zstd.py

@@ -2430,10 +2430,8 @@ def test_buffer_protocol(self):
             self.assertEqual(f.write(arr), LENGTH)
             self.assertEqual(f.tell(), LENGTH)
 
-@unittest.skip("it fails for now, see gh-133885")
 class FreeThreadingMethodTests(unittest.TestCase):
 
-    @unittest.skipUnless(Py_GIL_DISABLED, 'this test can only possibly fail with GIL disabled')
     @threading_helper.reap_threads
     @threading_helper.requires_working_threading()
     def test_compress_locking(self):
@@ -2470,7 +2468,6 @@ def run_method(method, input_data, output_data):
         actual = b''.join(output) + rest2
         self.assertEqual(expected, actual)
 
-    @unittest.skipUnless(Py_GIL_DISABLED, 'this test can only possibly fail with GIL disabled')
     @threading_helper.reap_threads
     @threading_helper.requires_working_threading()
     def test_decompress_locking(self):
@@ -2506,6 +2503,59 @@ def run_method(method, input_data, output_data):
         actual = b''.join(output)
         self.assertEqual(expected, actual)
 
+    @threading_helper.reap_threads
+    @threading_helper.requires_working_threading()
+    def test_compress_shared_dict(self):
+        num_threads = 8
+
+        def run_method(b):
+            level = threading.get_ident() % 4
+            # sync threads to increase chance of contention on
+            # capsule storing dictionary levels
+            b.wait()
+            ZstdCompressor(level=level,
+                           zstd_dict=TRAINED_DICT.as_digested_dict)
+            b.wait()
+            ZstdCompressor(level=level,
+                           zstd_dict=TRAINED_DICT.as_undigested_dict)
+            b.wait()
+            ZstdCompressor(level=level,
+                           zstd_dict=TRAINED_DICT.as_prefix)
+        threads = []
+
+        b = threading.Barrier(num_threads)
+        for i in range(num_threads):
+            thread = threading.Thread(target=run_method, args=(b,))
+
+            threads.append(thread)
+
+        with threading_helper.start_threads(threads):
+            pass
+
+    @threading_helper.reap_threads
+    @threading_helper.requires_working_threading()
+    def test_decompress_shared_dict(self):
+        num_threads = 8
+
+        def run_method(b):
+            # sync threads to increase chance of contention on
+            # decompression dictionary
+            b.wait()
+            ZstdDecompressor(zstd_dict=TRAINED_DICT.as_digested_dict)
+            b.wait()
+            ZstdDecompressor(zstd_dict=TRAINED_DICT.as_undigested_dict)
+            b.wait()
+            ZstdDecompressor(zstd_dict=TRAINED_DICT.as_prefix)
+        threads = []
+
+        b = threading.Barrier(num_threads)
+        for i in range(num_threads):
+            thread = threading.Thread(target=run_method, args=(b,))
+
+            threads.append(thread)
+
+        with threading_helper.start_threads(threads):
+            pass
 
 
 if __name__ == "__main__":