TST: Fix test_nowrite_df with recent Pandas where df.values.ctypes.data seems lazy 🤔

Author: kernc

backtesting/test/_test.py

@@ -588,17 +588,18 @@ def test_nowrite_df(self):
         # Test we don't write into passed data df by default.
         # Important for copy-on-write in Backtest.optimize()
         df = EURUSD.astype(float)
-        values = df.values.ctypes.data
-        assert values == df.values.ctypes.data
+        _ = df.values.ctypes.data  # dummy call for the value to stabilize!
+        mem_addr = df.values.ctypes.data
+        assert mem_addr == df.values.ctypes.data
 
         class S(SmaCross):
             def init(self):
                 super().init()
-                assert values == self.data.df.values.ctypes.data
+                assert mem_addr == self.data.df.values.ctypes.data, (mem_addr, self.data.df.values.ctypes.data)
 
         bt = Backtest(df, S)
         _ = bt.run()
-        assert values == bt._data.values.ctypes.data
+        assert mem_addr == bt._data.values.ctypes.data
 
     def test_multiprocessing_windows_spawn(self):
         df = GOOG.iloc[:100]