If someboby would look into TCPIP beta newsgroup...

QNX Newsgroups (Archives) qdn.public.qnx4
1

I have a problem…

2

Just to remind: Subj.
Or tcpip beta dead? – It’s about tcprt 4.25 too.

Vladimir Timonin wrote:

I have a problem…

Looks like serious one…
Or am I doing something wrong there?

Regards
Vladimir

3

Sorry, the TCP/IP 5.0 Patch A beta has been over for a while. The
release archive has not been posted due to delays in modifying of the
web site for myqnx to handle QNX 4 products. There is no change to the
binary components between the last beta archive and the release archive.
The beta.tcpip was intended to be closed for the time being. Athough I don’t
see any message in the newsgroup to state this I will arrange for
something to be stated.

To answer your question, As far as I can tell, this is the expected
behavior. If you call connect() on a non-blocking socket and then
select(), the socket can be set ready for both read() and write() in
the event of an error. When you call read() or write(), the error
will be delivered. Is this what you mean?

Thanks

Dave

4

Dave Brown wrote:

Sorry, the TCP/IP 5.0 Patch A beta has been over for a while. The
release archive has not been posted due to delays in modifying of the
web site for myqnx to handle QNX 4 products. There is no change to the
binary components between the last beta archive and the release archive.
The beta.tcpip was intended to be closed for the time being. Athough I
don’t see any message in the newsgroup to state this I will arrange for
something to be stated.

To answer your question, As far as I can tell, this is the expected
behavior. If you call connect() on a non-blocking socket and then
select(), the socket can be set ready for both read() and write() in
the event of an error. When you call read() or write(), the error
will be delivered. Is this what you mean?

Thanks

Dave

To go little bit further:
after connect() on a socket I’m waiting for the write permission,
after getting write permission I’m trying socket with recv(…, MSG_OOB) -
and it gives me no error; moreover write() appears without error (just can
EWOULDBLOCK after several of them) and only after last SYN/ICMP select()
shows me read permission and read() returns 0 with errno (if I remember
properly, 257 - socket not connected).

If would be nice if TCP/IP stack
after first SYN/ICMP give write permission, error, and will not send other
SYNs;
or give write permission and error after last SYN/ICMP.

I got this problem with 4.25 and just tried with 5.0.

Regards
Vladimir