Issue #5148: Ignore 'U' in mode given to gzip.open() and gzip.GzipFile().

Lib/gzip.py

@@ -81,6 +81,10 @@ class GzipFile(io.BufferedIOBase):
 
         """
 
+        # Make sure we don't inadvertently enable universal newlines on the
+        # underlying file object - in read mode, this causes data corruption.
+        if mode:
+            mode = mode.replace('U', '')
         # guarantee the file is opened in binary mode on platforms
         # that care about that sort of thing
         if mode and 'b' not in mode:

Lib/test/test_gzip.py

@@ -53,6 +53,13 @@ class TestGzip(unittest.TestCase):
             d = f.read()
         self.assertEqual(d, data1*50)
 
+    def test_read_universal_newlines(self):
+        # Issue #5148: Reading breaks when mode contains 'U'.
+        self.test_write()
+        with gzip.GzipFile(self.filename, 'rU') as f:
+            d = f.read()
+        self.assertEqual(d, data1*50)
+
     def test_io_on_closed_object(self):
         # Test that I/O operations on closed GzipFile objects raise a
         # ValueError, just like the corresponding functions on file objects.

Misc/NEWS

@@ -122,6 +122,8 @@ Core and Builtins
 Library
 -------
 
+- Issue #5148: Ignore 'U' in mode given to gzip.open() and gzip.GzipFile().
+
 - Issue #16220: wsgiref now always calls close() on an iterable response.
   Patch by Brent Tubbs.