Fix for #876637 - Random stack corruption from socketmodule

[troels]

BPO	970288
Nosy	@loewis, @devdanzin
Dependencies	bpo-876637: Random stack corruption from socketmodule.c
Files	socketmodule_timeout_using_poll.patch

^{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 2009-02-14.13:57:08.389>
created_at = <Date 2004-06-10.11:13:59.000>
labels = ['extension-modules', 'type-bug']
title = 'Fix for python/cpython#39819 - Random stack corruption from socketmodule'
updated_at = <Date 2009-02-14.13:57:08.377>
user = 'https://bugs.python.org/troels'

bugs.python.org fields:

activity = <Date 2009-02-14.13:57:08.377>
actor = 'ajaksu2'
assignee = 'none'
closed = True
closed_date = <Date 2009-02-14.13:57:08.389>
closer = 'ajaksu2'
components = ['Extension Modules']
creation = <Date 2004-06-10.11:13:59.000>
creator = 'troels'
dependencies = ['876637']
files = ['6027']
hgrepos = []
issue_num = 970288
keywords = ['patch']
message_count = 5.0
messages = ['46164', '46165', '46166', '46167', '82055']
nosy_count = 5.0
nosy_names = ['loewis', 'nnorwitz', 'troels', 'ajaksu2', 'vila']
pr_nums = []
priority = 'normal'
resolution = 'out of date'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue970288'
versions = ['Python 2.4']

[troels]

Here's a patch to convert socketmodule to use poll()
for the timeout handling.

ifdefs and includes have been copied from selectmodule,
so I believe they should be correct for all platforms.

Compiled and tested on Linux (RH73).

[loewis]

Logged In: YES
user_id=21627

While the patch is good as it stands, it is also incomplete:
If poll(2) is not available, it might still overrun the
FD_SETSIZE. I think there should be another return value
from internal_select (-1), indicating an internal error. A
Python RuntimeError should be set in this case. Callers
should check for -1, and return immediately.

Are you willing to extend the patch in that direction?

[troels]

Logged In: YES
user_id=32863

You're right, however I think that it is a very minor
problem compared to the current situation. All major UNIX
platforms support poll(2) in my experience. The only
important Python platform without poll is Windows, and it
doesn't have the FD_SETSIZE problem.

Thus I think the patch should be applied as it stands, but
the original bug ticket can be left open for another
enterprising individual to make the changes you suggest.

I'm afraid that work commitments prevent me from being that
individual for the foreseeable future...

[nnorwitz]

Logged In: YES
user_id=33168

A fix for the bug (sd over FD_SETSIZE) has been checked in.
Note select is called in Modules/_ssl.c too.

[devdanzin]

Fixed in revision 42253.