matplotlib
diff --git a/‎lib/matplotlib/tests/test_rcparams.py
Copy file name to clipboardExpand all lines: lib/matplotlib/tests/test_rcparams.py
+36-20Lines changed: 36 additions & 20 deletions b/‎lib/matplotlib/tests/test_rcparams.py
Copy file name to clipboardExpand all lines: lib/matplotlib/tests/test_rcparams.py
+36-20Lines changed: 36 additions & 20 deletions
@@ -333,29 +333,45 @@ def generate_validator_testcases(valid):
                      ),
          'fail': (('aardvark', ValueError),
                   )
-         },
-         {'validator': _validate_linestyle,  # NB: case-insensitive
-         'success': (('-', '-'), ('solid', 'solid'),
-                     ('--', '--'), ('dashed', 'dashed'),
-                     ('-.', '-.'), ('dashdot', 'dashdot'),
-                     (':', ':'), ('dotted', 'dotted'),
-                     ('', ''), (' ', ' '),
-                     ('None', 'none'), ('none', 'none'),
-                     ('DoTtEd', 'dotted'),
-                     (['1.23', '4.56'], (None, [1.23, 4.56])),
-                     ([1.23, 456], (None, [1.23, 456.0])),
-                     ([1, 2, 3, 4], (None, [1.0, 2.0, 3.0, 4.0])),
-                     ),
-         'fail': (('aardvark', ValueError),  # not a valid string
-                  ((None, [1, 2]), ValueError),  # (offset, dashes) is not OK
-                  ((0, [1, 2]), ValueError),  # idem
-                  ((-1, [1, 2]), ValueError),  # idem
-                  ([1, 2, 3], ValueError),  # not a sequence of even length
-                  (1.23, ValueError)  # not a sequence
-                  )
          }
     )
 
+    # The behavior of _validate_linestyle depends on the version of Python.
+    # ASCII-compliant bytes arguments should pass on Python 2 because of the
+    # automatic conversion between bytes and strings. Python 3 does not
+    # perform such a conversion, so the same cases should raise an exception.
+    #
+    # Common cases:
+    ls_test = {'validator': _validate_linestyle,
+               'success': (('-', '-'), ('solid', 'solid'),
+                           ('--', '--'), ('dashed', 'dashed'),
+                           ('-.', '-.'), ('dashdot', 'dashdot'),
+                           (':', ':'), ('dotted', 'dotted'),
+                           ('', ''), (' ', ' '),
+                           ('None', 'none'), ('none', 'none'),
+                           ('DoTtEd', 'dotted'),  # case-insensitive
+                           (['1.23', '4.56'], (None, [1.23, 4.56])),
+                           ([1.23, 456], (None, [1.23, 456.0])),
+                           ([1, 2, 3, 4], (None, [1.0, 2.0, 3.0, 4.0])),
+                          ),
+               'fail': (('aardvark', ValueError),  # not a valid string
+                        ('dotted'.encode('utf-16'), ValueError),  # even on PY2
+                        ((None, [1, 2]), ValueError),  # (offset, dashes) != OK
+                        ((0, [1, 2]), ValueError),  # idem
+                        ((-1, [1, 2]), ValueError),  # idem
+                        ([1, 2, 3], ValueError),  # sequence with odd length
+                        (1.23, ValueError),  # not a sequence
+                       )
+                }
+    # Add some cases of bytes arguments that Python 2 can convert silently:
+    ls_bytes_args = (b'dotted', 'dotted'.encode('ascii'))
+    if six.PY3:
+        ls_test['fail'] += tuple((arg, ValueError) for arg in ls_bytes_args)
+    else:
+        ls_test['success'] += tuple((arg, 'dotted') for arg in ls_bytes_args)
+    # Update the validation test sequence.
+    validation_tests += (ls_test,)
+
     for validator_dict in validation_tests:
         validator = validator_dict['validator']
         if valid: