socketmodule.c connection handling incorect on windows

[garth42]

BPO	777597
Nosy	@tim-one, @devdanzin
Dependencies	bpo-965036: Fix for socketmodule.c connection handling incorect on windows #38930 - socketmodule.c connection handling incorec

^{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 2009-02-13.03:29:43.195>
created_at = <Date 2003-07-25.15:01:02.000>
labels = ['type-bug', 'library', 'OS-windows']
title = 'socketmodule.c connection handling incorect on windows'
updated_at = <Date 2009-02-13.03:29:43.193>
user = 'https://bugs.python.org/garth42'

bugs.python.org fields:

activity = <Date 2009-02-13.03:29:43.193>
actor = 'ajaksu2'
assignee = 'none'
closed = True
closed_date = <Date 2009-02-13.03:29:43.195>
closer = 'ajaksu2'
components = ['Library (Lib)', 'Windows']
creation = <Date 2003-07-25.15:01:02.000>
creator = 'garth42'
dependencies = ['965036']
files = []
hgrepos = []
issue_num = 777597
keywords = []
message_count = 6.0
messages = ['17330', '17331', '17332', '17333', '17334', '81864']
nosy_count = 5.0
nosy_names = ['tim.peters', 'nnorwitz', 'troels', 'garth42', 'ajaksu2']
pr_nums = []
priority = 'normal'
resolution = 'out of date'
stage = None
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue777597'
versions = ['Python 2.4']

[garth42]

The socketmodule.c code does not handle connection
refused correctly. This is due to a different of
operation on windows of select. The offending code is
in internal_connect in the MS_WINDOWS ifdef. The code
in should test exceptfds to check for connecttion
refused. If this is so itshould call
getsockopt(SOL_SOCKET, SO_ERROR,..) to get the error
status. (Source microsoft Platform SDK)
The suggested fix is shown below (untested)

#ifdef MS_WINDOWS

f (s->sock_timeout > 0.0) {
    if (res < 0 && WSAGetLastError() == WSAEWOULDBLOCK) {
	/* This is a mess.  Best solution: trust select */
	fd_set exfds;
	struct timeval tv;
	tv.tv_sec = (int)s->sock_timeout;
	tv.tv_usec = (int)((s->sock_timeout - tv.tv_sec) * 1e6);
	FD_ZERO(&exfds);
	FD_SET(s->sock_fd, &exfds);
	/* Platform SDK says so */
	res = select(s->sock_fd+1, NULL, NULL, &exfds, &tv);
	if (res > 0) { 
	    if( FD_ISSET( &exfds ) ) {
		/* Get the real reason */
	
getsockopt(s->sock_fd,SOL_SOCKET,SO_ERROR,(char*)&res,sizeof(res));
	    } else {
		/* God knows how we got here */
		res = 0;
	    }
	} else if( res == 0 ) {
	    res = WSAEWOULDBLOCK;
	} else
	{ 
	    /* Not sure we should return the erro from select? */
	    res =  WSAGetLastError();
	}
    }
} else if (res < 0)
    res = WSAGetLastError();

#else

[nnorwitz]

Logged In: YES
user_id=33168

Garth could you produce a patch against 2.3c2 with your
selected change and test it? It would help us a lot as we
are all very overloaded. Thanks.

[troels]

Logged In: YES
user_id=32863

I have turned Garth's code into a patch versus Python 2.3.4.
I don't believe the fix is correct and complete, but it
should serve as a starting point. Patch is in
http://python.org/sf/965036

[troels]

Logged In: YES
user_id=32863

http://python.org/sf/965036 has been updated with a fixed
and tested patch. Could somebody review and apply it? Thanks!

[tim-one]

Logged In: YES
user_id=31435

Unassigned myself (no particular competence here, and short
on time).

[devdanzin]

Fixed in r36739 according to bpo-965036.