diff -r dd67c8c53aea Lib/base64.py
--- a/Lib/base64.py	Mon Dec 07 02:31:11 2015 +0200
+++ b/Lib/base64.py	Wed Dec 23 10:10:11 2015 +0530
@@ -375,10 +375,15 @@
     """
     b = _bytes_from_decode_data(b)
     if adobe:
-        if not (b.startswith(_A85START) and b.endswith(_A85END)):
-            raise ValueError("Ascii85 encoded byte sequences must be bracketed "
-                             "by {!r} and {!r}".format(_A85START, _A85END))
-        b = b[2:-2] # Strip off start/end markers
+        if not b.endswith(_A85END):
+            raise ValueError(
+                "Ascii85 encoded byte sequences must end "
+                "with {!r}".format(_A85END)
+                )
+        if b.startswith(_A85START):
+            b = b[2:-2]  # Strip off start/end markers
+        else:
+            b = b[:-2]
     #
     # We have to go through this stepwise, so as to ignore spaces and handle
     # special short sequences
diff -r dd67c8c53aea Lib/test/test_base64.py
--- a/Lib/test/test_base64.py	Mon Dec 07 02:31:11 2015 +0200
+++ b/Lib/test/test_base64.py	Wed Dec 23 10:10:11 2015 +0530
@@ -568,8 +568,6 @@
                                       b"malformed", adobe=True)
         self.assertRaises(ValueError, base64.a85decode,
                                       b"<~still malformed", adobe=True)
-        self.assertRaises(ValueError, base64.a85decode,
-                                      b"also malformed~>", adobe=True)
 
         # With adobe=False (the default), Adobe framing markers are disallowed
         self.assertRaises(ValueError, base64.a85decode,