Re: SOCKET Connect() connects but shouldn't. -- RPG400-L


On 05/02/2006, at 7:31 AM, Scott Klement wrote:

First, byte-ordering. The bytes for network values must be in network
byte order. This happens to be big-endian but PCs and other toys use
little-endian thus your port number is probably being misinterpreted.

That'd make sense if he's running the code on a little-endian systemlike

a PC.  Is that the case?  If so, I completely missed that point...

That's certainly how I read his append late last night. Looking at itagain in the clear light of morning it seems I misinterpreted what hesaid. I think I got thrown by the "server on a PC" and "netstat fromDOS" comments and thought he was running the example code on a PCtoo--which makes even less sense to me this morning. The example codeis in RPG IV--how could it possibly be running on a PC?

OK so it seems he has a listener active on his PC and is running hisexample code on OS/400 and is trying to connect to the listener runningon the PC. In that case the byte order will be correct.

As to why connect() seems to work but doesn't actually connect? Well,presuming it is not really successfully connecting to an unexpectedsystem then it's very strange. connect() should return -1 for anunsuccessful connexion regardless of whether ICMP is disabled. Itshould at least time-out if the connexion is not made.

PING or TRACEROUTE will indicate whether ICMP Echo is disabled (atleast a reasonable guess) and if not will provide useful informationabout whether the target server is reachable and the network path bywhich it was reached.

A similar misinterpretation will occur with the IP address so you maybeattempting to connect to a completely different server which doeshappen
to have a listener on port 23585.
That's only true if he's hard-coding the binary form of the IP address.
The inet_addr() and gethostbyname() functions always return the IP
addresses in network byte order, so you'll just confuse things byflipping
it again.

It wasn't apparent from the example code how he was getting the valueof IP so I pointed out the same flaw could exist there. On a PC itcould. You're correct that applying htonl() on the values returned bygethostbyname() will confuse things but only if done on anon-big-endian system. In OS/400 C it wouldn't have any effect becauseboth htons() and htonl() are effectively no-ops. Further they do notexist as exposed functions so can't be called from RPG IV anyway.

Even though the example was in RPG (and this is an RPG forum) I tend tothink about sockets from a C perspective. That skewed view pretty muchinvalidates everything I said in my response because it doesn't applyto RPG IV sockets. I think I should stop answering appends late atnight and just wait until morning--by which time someone else will haveanswered anyway.


Regards,
Simon Coulter.
--------------------------------------------------------------------
   FlyByNight Software         AS/400 Technical Specialists

   http://www.flybynight.com.au/
   Phone: +61 3 9419 0175   Mobile: +61 0411 091 400        /"\
   Fax:   +61 3 9419 0175                                   \ /
                                                             X
                 ASCII Ribbon campaign against HTML E-Mail  / \
--------------------------------------------------------------------