error responses from server are masked in smtplib when server closes connection

[bitdancer]

BPO	17498
Nosy	@bitdancer, @kushaldas
Files	issue17498.patch issue17498_v2.patch: New version of the patch which can be successfully applied to tip.

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2014-04-14.22:24:41.257>
created_at = <Date 2013-03-20.20:48:30.070>
labels = ['type-bug']
title = 'error responses from server are masked in smtplib when server closes connection'
updated_at = <Date 2014-04-14.22:24:41.255>
user = 'https://github.com/bitdancer'

bugs.python.org fields:

activity = <Date 2014-04-14.22:24:41.255>
actor = 'r.david.murray'
assignee = 'none'
closed = True
closed_date = <Date 2014-04-14.22:24:41.257>
closer = 'r.david.murray'
components = []
creation = <Date 2013-03-20.20:48:30.070>
creator = 'r.david.murray'
dependencies = []
files = ['29521', '34817']
hgrepos = []
issue_num = 17498
keywords = ['patch']
message_count = 6.0
messages = ['184793', '184810', '184829', '216089', '216235', '216237']
nosy_count = 3.0
nosy_names = ['r.david.murray', 'python-dev', 'kushal.das']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue17498'
versions = ['Python 3.4', 'Python 3.5']

[bitdancer]

If an SMTP server returns an error code and then closes the connection, smtplib will attempt to send an RSET command, which will then raise a socket error, which smtplib turns into a ServerDisconnectedError, thereby loosing the information about what error code the server actually returned. While (except for code 421, as noted in bpo-5713) a server doing this is technically out of spec with the RFC, it is a reality of the internet that it happens, and smtplib should handle it correctly.

Since the RSET is just to get the protocol back to a known starting state after an error, I believe the fix is to have an internal _rset method that wraps the RSET call in a try/except and just ignores any ServerDiscoonected error. The library would then call this internally instead of the public API method. If the library client continues on after the error code it gets back from calling an smtplib method that results in the server closing the connection, it will get the ServerDisconnected error the next time it tries to send a command to the server. Thus the behavior change introduced here is just to allow commands that end with server disconnects to return an error code instead of an (otherwise mysterious and quite possibly currently-mishandled) ServerDisconnected error. Since any existing code will already need to handle errors from commands sent to servers that *don't* disconnect, I believe that this is a safe fix from a backward compatibility standpoint.

[kushaldas]

Working on a patch for this.

[kushaldas]

The following patch adds a _rset method which catches any SMTPServerDisconnected exception, also includes the corresponding test and test_421_from_mail_cmd assert statement typo fix.

[kushaldas]

New version of the patch which can be successfully applied to tip.

[python-dev]

New changeset 3c441e9ccf87 by R David Murray in branch '3.4':
bpo-17498: Defer SMTPServerDisconnected errors until the next command.
http://hg.python.org/cpython/rev/3c441e9ccf87

New changeset 842014ab1c06 by R David Murray in branch 'default':
Merge bpo-17498: Defer SMTPServerDisconnected errors until the next command.
http://hg.python.org/cpython/rev/842014ab1c06

[bitdancer]

Thanks, Kushal. Sorry it took so long to get this committed :)