cannot send email with only bcc

[mmerickel]

message = Message(subject='test', body='this is a test', bcc=['me@example.com'])
mailer.send(message)
transaction.commit()

Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/runpy.py", line 162, in _run_module_as_main
    "__main__", fname, loader, pkg_name)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/runpy.py", line 72, in _run_code
    exec code in run_globals
  File "test.py", line 83, in <module>
    main(sys.argv)
  File "test.py", line 79, in main
    mailer.send(message)
  File "lib/python2.7/site-packages/pyramid_mailer/mailer.py", line 178, in send
    return self.direct_delivery.send(*self._message_args(message))
  File "lib/python2.7/site-packages/pyramid_mailer/mailer.py", line 222, in _message_args
    message.to_message())
  File "lib/python2.7/site-packages/pyramid_mailer/message.py", line 82, in to_message
    self.validate()
  File "lib/python2.7/site-packages/pyramid_mailer/message.py", line 132, in validate
    raise InvalidMessage, "No recipients have been added"
pyramid_mailer.exceptions.InvalidMessage: No recipients have been added

[rpatterson]

See #10 for possibly related history. I'm wondering, as @mcdonc did, if that pull request introduced this bug.

[mmerickel]

Note that Message(cc=..) has a similar problem. All scenarios require the recipients argument for some reason.

[mcdonc]

The tests for cc-only and bcc-only worked under Python 2.6 but failed under Python 3.2 on rpatterson's master. Instead of fixing it, I just caused pyramid_mailer to raise a more descriptive error message saying:

            raise InvalidMessage("Must have at least one direct recipient "
                                 "even if cc or bcc set")

We can revisit this later, I just couldn't figure out how to make it work under 3.2.

[rpatterson]

Odd, I re-ran tox before pulling and merging and the tests passed under 3.2 for me. Investigating now.

[rpatterson]

When I run tox against 95ff3e7, your first commit on top of my master, I see the failures in all envs. When I run tox against d2c6adf, my last change I see only an unrelated failure under jython (which I missed before, sorry). So I'm not seeing a 3.2 specific failure.

But reading your comment and commit logs more carefully, it seems like you want pyramid_mailer to enforce a prohibition against cc/bcc only messages so I guess I just shouldn't have gone off and implemented what mmerickel requested without discussion. Oops, sorry. :-)

[mmerickel]

@rpatterson, he doesn't want to enforce against cc/bcc only messages. That was just his clear error message until he could revisit the issue. Keep fighting. :-)

[rpatterson]

@mmerickel You keep fighting! And here's how... :-)

Checkout d2c6adf and see if the 3.2 tests pass for you.

[rpatterson]

@mmerickel Are you sure he doesn't want to? Have you looked at 95ff3e7? @mcdonc Care to clarify whether you want to allow messages with cc/bcc but no recipients?

[mcdonc]

No I don't want to enforce against it. I just couldn't figure out the failure I see.

When I comment out the check I added in message.py:

        ## if (self.cc or self.bcc) and not self.recipients:
        ##     raise InvalidMessage("Must have at least one direct recipient "
        ##                          "even if cc or bcc set")

And add this test to tests.TestMessage

    def test_cc_without_recipients_2(self):

        from pyramid_mailer.message import Message

        msg = Message(subject="testing",
                      sender="sender@example.com",
                      body="testing",
                      cc=["tosomeoneelse@example.com"])
        response = msg.get_response()
        self.assertTrue("Cc: tosomeoneelse@example.com" in str(response))

This is the exception I get on my system under Python 3.2:

======================================================================
ERROR: test_cc_without_recipients_2 (pyramid_mailer.tests.TestMessage)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/chrism/projects/pyramid_mailer/pyramid_mailer/tests.py", line 173, in test_cc_without_recipients
    self.assertTrue("Cc: tosomeoneelse@example.com" in str(response))
  File "/home/chrism/projects/pyramid_mailer/pyramid_mailer/response.py", line 239, in __str__
    return self.to_message().as_string()
  File "/home/chrism/opt/Python-3.2/lib/python3.2/email/message.py", line 167, in as_string
    g.flatten(self, unixfrom=unixfrom)
  File "/home/chrism/opt/Python-3.2/lib/python3.2/email/generator.py", line 88, in flatten
    self._write(msg)
  File "/home/chrism/opt/Python-3.2/lib/python3.2/email/generator.py", line 141, in _write
    self._write_headers(msg)
  File "/home/chrism/opt/Python-3.2/lib/python3.2/email/generator.py", line 176, in _write_headers
    self.write(header.encode(linesep=self._NL)+self._NL)
  File "/home/chrism/opt/Python-3.2/lib/python3.2/email/header.py", line 317, in encode
    formatter.feed(lines[0], charset)
IndexError: list index out of range

It doesn't happen on Python 2.6 or 2.7. I noticed while PDB stepping through it that the "To" header is the empty string. I suspect that may be why but I didn't chase it down and punted by disallowing it.

[rpatterson]

@mcdonc Ok, I'll apply that patch and investigate. Thanks!

[mcdonc]

By the way I made new releases of pyramid_mailer and repoze.sendmail last night, thanks for the work Ross!

[rpatterson]

@mcdonc Yeah, I don't get that error or any failure under 3.2 when I revert your related changes and then add the test you provided above.

@mmerickel Can you try running tox against rpatterson/pyramid_mailer@fb2a013?

[mcdonc]

I'm on Python 3.2 (not 3.2.1 or 3.2.2), maybe that's the difference.

[rpatterson]

Ok, I'm testing 3.2 now

[rpatterson]

Hmm, I couldn't build build 3.2 with ssl support, I think because of http://bugs.python.org/issue12012. Without ssl, the imports in the tests are broken. So I manually applied the patch that fixes that bug:

http://hg.python.org/cpython/raw-rev/20beec22764f

Then I was able to reproduce the issue and commit a workaround that skips headers that would be empty strings which seems appropriate anyways.

[mcdonc]

This was fixed with the merge of the above-mentioned pull request. Thanks Ross!