Currently the `msg` argument of `smtplib.SMTP.sendmail` accept a `str` in Py3k if every characters in this `str` is in ASCII range, or a `bytes`.

This is confusing for new comer because:

1. When you send your mail using only ASCII characters, everything is fine (no matter you use bytes or str).
2. When sometimes you included non-ASCII characters in your email, the traceback is hard to understand.

Here's an example of such traceback:

Traceback (most recent call last):
  File "./manage.py", line 113, in <module>
    manager.run()
  File "/data/web/cgi-bin/venv/lib/python3.3/site-packages/flask_script/__init__.py", line 405, in run
    result = self.handle(sys.argv[0], sys.argv[1:])
  File "/data/web/cgi-bin/venv/lib/python3.3/site-packages/flask_script/__init__.py", line 384, in handle
    return handle(app, *positional_args, **kwargs)
  File "/data/web/cgi-bin/venv/lib/python3.3/site-packages/flask_script/commands.py", line 145, in handle
    return self.run(*args, **kwargs)
  File "./manage.py", line 108, in run
    conn.send(msg)
  File "/data/web/cgi-bin/venv/lib/python3.3/site-packages/flask_mail.py", line 168, in send
    message.as_string())
  File "/data/web/cgi-bin/python-3.3.3/lib/python3.3/smtplib.py", line 747, in sendmail
    msg = _fix_eols(msg).encode('ascii')
UnicodeEncodeError: 'ascii' codec can't encode character '\u9f99' in position 646: ordinal not in range(128)

Here's my proposal:

--- smtplib.py.orig     2014-02-01 21:26:47.000000000 -0500
+++ smtplib.py  2014-02-01 21:37:51.000000000 -0500
@@ -744,7 +744,12 @@
         esmtp_opts = []
         print(msg)
         if isinstance(msg, str):
-            msg = _fix_eols(msg).encode('ascii')
+            try:
+                msg = _fix_eols(msg).encode('ascii')
+            except UnicodeEncodeError:
+                raise SMTPException(
+                    "msg may be a string containing characters in the "
+                    "ASCII range, or a byte string.")
         if self.does_esmtp:
             # Hmmm? what's this? -ddm
             # self.esmtp_features['7bit']=""

Assuming that the inline patch in msg209943 is acceptable I'll volunteer to change the test code unless the originator wants the job.

I have no plan to work further, Mark can take the code and improve anyway you want.

I don't believe that there is much sense in implementing the proposed change.  If the stdlib was changed every time a UnicodeEncodeError was reported just to show a slightly clearer exception message, I dread to think how much extra code would be added in total.  Committing a change such as this takes core dev time which IMHO opinion could be put to better use.