Race condition in asyncore poll

[steveazope]

BPO	629097
Nosy	@gvanrossum, @akuchling

^{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 = 'https://github.com/akuchling'
closed_at = <Date 2004-07-10.17:15:53.000>
created_at = <Date 2002-10-26.14:48:09.000>
labels = ['library']
title = 'Race condition in asyncore poll'
updated_at = <Date 2004-07-10.17:15:53.000>
user = 'https://bugs.python.org/steveazope'

bugs.python.org fields:

activity = <Date 2004-07-10.17:15:53.000>
actor = 'akuchling'
assignee = 'akuchling'
closed = True
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2002-10-26.14:48:09.000>
creator = 'stevea_zope'
dependencies = []
files = []
hgrepos = []
issue_num = 629097
keywords = []
message_count = 4.0
messages = ['12952', '12953', '12954', '12955']
nosy_count = 4.0
nosy_names = ['gvanrossum', 'akuchling', 'stevea_zope', 'klimkin']
pr_nums = []
priority = 'normal'
resolution = 'wont fix'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue629097'
versions = ['Python 2.3']

[steveazope]

In the following post to the Zope 3 developers' list, I
describe a race condition in the poll method of asyncore.

http://lists.zope.org/pipermail/zope3-dev/2002-October/003091.html

The problem is this:

There is a global dict socket_map. In the poll method,
socket_map is processed into appropriate arguments for
a select.select call. However, if a socket that is
represented socket_map is closed during the time
between the processing of socket_map and the
select.select call, that call will fail with a Bad File
Descriptor (EBADF) error.

One solution is to patch asyncore to catch EBADF errors
raised by select.select, and at that point see if the
file descriptors in the current version of socket_map
are the same as in the processed data used for the
select.select call. If they are the same, re-raise the
error, otherwise, ignore the error.

In another email to the Zope 3 developers' list, Jeremy
Hylton queries whether there are any other similar
problems in asyncore.

http://lists.zope.org/pipermail/zope3-dev/2002-October/003093.html

[gvanrossum]

Logged In: YES
user_id=6380

According to Jeremy, this is more a matter of "don't do that".

The right solution is to make sure that sockets are only
closed by the main thread (the thread running asyncore.loop()).

I wonder if we should just close this bug report?

[klimkin]

Logged In: YES
user_id=410460

Yup, closing file in separate thread is program's (not
asyncore) error.

[akuchling]

Logged In: YES
user_id=11375

Closing this bug.