Move to RFC 3986 for normalizing URLs, allowing ~ in URL path components

Matches python behavior since version 3.7 and also browser behavior

Author: kovidgoyal

mechanize/_urllib2_fork.py

@@ -138,7 +138,7 @@ def request_host(request):
     b"ABCDEFGHIJKLMNOPQRSTUVWXYZ" b"abcdefghijklmnopqrstuvwxyz" b"0123456789._!-"
 ))
 USERINFO_CHARS = ZONE_ID_CHARS | set(bytearray(b"$&'()*+,;=:"))
-PATH_CHARS = USERINFO_CHARS | set(bytearray(b'@/'))
+PATH_CHARS = USERINFO_CHARS | set(bytearray(b'@/~'))
 QUERY_CHARS = FRAGMENT_CHARS = PATH_CHARS | {ord(b"?")}
 
 

test/test_urllib2.py

@@ -1871,11 +1871,11 @@ def test_add_data(self):
         self.assertEqual("POST", self.get.get_method())
 
     def test_get_full_url(self):
-        self.assertEqual("http://www.python.org/%7Ejeremy/",
+        self.assertEqual("http://www.python.org/~jeremy/",
                          self.get.get_full_url())
 
     def test_selector(self):
-        self.assertEqual("/%7Ejeremy/", self.get.get_selector())
+        self.assertEqual("/~jeremy/", self.get.get_selector())
         req = Request("http://www.python.org/")
         self.assertEqual("/", req.get_selector())