Error when parsing a message with quoted name in the From field

adam-iris commented 4 years ago

When parsing an email whose From field is formatted like:

From: "Lastname, Firstname" <user@example.com>

authentication fails with a trace like:

Traceback (most recent call last):
  File "authheaders/test/test_authentication.py", line 124, in test_authenticate_headers
    res = authenticate_message(self.message8, "example.com", spf=False, dkim=False, dmarc=True, dnsfunc=self.dnsfunc)
  File "/code/conda_envs/fdsn_website/lib/python2.7/site-packages/authheaders/__init__.py", line 289, in authenticate_message
    dmarc_result = check_dmarc(msg, spf_result, dkim_result, dnsfunc=dnsfunc, psddmarc=psddmarc)
  File "/code/conda_envs/fdsn_website/lib/python2.7/site-packages/authheaders/__init__.py", line 220, in check_dmarc
    from_domain = get_domain_part(from_header)
  File "/code/conda_envs/fdsn_website/lib/python2.7/site-packages/authheaders/__init__.py", line 49, in get_domain_part
    return res[0].decode('ascii')
IndexError: list index out of range

I think the problem is at https://github.com/ValiMail/authentication-headers/blob/master/authheaders/__init__.py#L214

    from_headers = [x[1].split(b',') for x in headers if x[0].lower() == b"from"][0]

This yields an email address of Firstname" <user@example.com> for the example above.

I would propose a fix using email.utils.getaddresses:

    from_headers = [a[1] for a in getaddresses(x[1] for x in headers if x[0].lower() == b"from")]

This worked for me, and passes all the other unit tests.

If this fix seems reasonable, I can submit a PR for it.

kitterma commented 4 years ago

Yes. Please, as long as you have tested both python2.7 and a recent python3 version.

Scott K

On January 31, 2020 7:32:24 PM UTC, Adam Clark notifications@github.com wrote:

When parsing an email whose From field is formatted like:

From: "Lastname, Firstname" user@example.com

authentication fails with a trace like:
Traceback (most recent call last):
File "authheaders/test/test_authentication.py", line 124, in
test_authenticate_headers
res = authenticate_message(self.message8, "example.com", spf=False,
dkim=False, dmarc=True, dnsfunc=self.dnsfunc)
File
"/code/conda_envs/fdsn_website/lib/python2.7/site-packages/authheaders/__init__.py",
line 289, in authenticate_message
dmarc_result = check_dmarc(msg, spf_result, dkim_result,
dnsfunc=dnsfunc, psddmarc=psddmarc)
File
"/code/conda_envs/fdsn_website/lib/python2.7/site-packages/authheaders/__init__.py",
line 220, in check_dmarc
   from_domain = get_domain_part(from_header)
File
"/code/conda_envs/fdsn_website/lib/python2.7/site-packages/authheaders/__init__.py",
line 49, in get_domain_part
   return res[0].decode('ascii')
IndexError: list index out of range
I think the problem is at https://github.com/ValiMail/authentication-headers/blob/master/authheaders/__init__.py#L214
from_headers = [x[1].split(b',') for x in headers if x[0].lower() ==
b"from"][0]
This yields an email address of Firstname" <user@example.com> for the example above.

I would propose a fix using email.utils.getaddresses:
from_headers = [a[1] for a in getaddresses(x[1] for x in headers if
x[0].lower() == b"from")]
This worked for me, and passes all the other unit tests.

If this fix seems reasonable, I can submit a PR for it.

msapiro commented 4 years ago

I have the same issue, although I see it differently from @adam-iris . In my case, the From: is

From: "Lastname, Firstname" <user@example.com>\r\n

as above, and the relevant headers item is

[b'From', b' "Lastname, Firstname" <user@example.com>\r\n']

and the result from

[x  for x in headers if x[0].lower() == b"from"]

is

[[b'From', b' "Lastname, Firstname" <user@example.com>\r\n']]

Thus,

from_headers = [x[1].split(b',') for x in headers if x[0].lower() == b"from"][0]

yields

[b' "Lastname', b' Firstname" <user@example.com>\r\n']

which is of length 2 so code calls get_domain_part() on both values and get_domain_part(b' "Lastname') throws IndexError on

    res = re.findall(b'@([a-z0-9.]+)', address)
    return res[0].decode('ascii')

because address contains no @. Further, the suggested

from_headers = [a[1] for a in getaddresses(x[1] for x in headers if x[0].lower() == b"from")]

throws TypeError because the x[1]s are bytes and it wants string (at least for Python 3). This

from_headers = [a[1] for a in getaddresses(x[1].decode(errors='ignore').strip() for x in headers if x[0].lower() == b"from")]

works for me. The .decode(errors='ignore').strip converts to string ignoring possible non-utf-8 in the display name and removes the \r\n