Not big deal, just curious.

Im playing with UDP right now, and I just realized that recvfrom will most of the time return WSAECONNRESET when theres no data coming or going from anywhere..Second msdn:

"On a UDP-datagram socket this error indicates a previous send operation resulted in an ICMP Port Unreachable message."

Than I realized this error is due sendto (if I remove it, the only error is wouldblock)...thats so weird. Why recvfrom would want to care about sendto status or whatever?? Should I treat this just like wouldblock (ignoring it that is)?

Do I get this JUST because Im testing on a single machine with loopback address (since UDP is unreliable and very "raw", how can he knows the message is unreachable, I hope its a local thing, like it didnt even managed to "go for a trip", and thus the error)

The send operation locally completes without problems, but the network later returns an ICMP packet indicating that the destination is invalid. Subsequent socket operations will report that error state.

How you deal with this message is up to you, depending on what you're doing. The message does indicate that your intended destination is not receiving your packets, so usually you'd want to take that into consideration.

If you're getting this error on the loopback it suggests that you haven't correctly bound the listening socket prior to calling sendto.

NVM, after looking around a bit this is a spurious message in this case. Just ignore it.

SOCKET_ERROR really means theres an error on the socket huh? I wasnt seeing things by that perspective (it looks so obvious now)...I think I need to change my per function error codes to a socket status thing.

This is an "optimization" done by Microsoft.

When you sendto() to a host/port combination that isn't actually expecting a UDP datagram, that host will likely send back a ICMP error message.

In the regular socket API, there isn't a good way of actually receiving this error, so Microsoft wired that message back up to the call to recvfrom(). If there is a received, un-delivered ICMP error of this kind, it will be returned by recvfrom(). If you want to know which particular remote host/port didn't respond, check the address returned by recvfrom().

This is actually useful. You can use this information to detect that a remote machine is not running a server (if this is on the client,) or that a remote client has shut down / crashed (if this is on the server.) Or you can just use it to detect that you're passing the wrong parameters to sendto() and are not sending the datagrams to the address/port you think you are :-)

The main thing to realize about this error is that it doesn't mean the socket is dead -- you can loop right back and call recvfrom() on it again, to dequeue whatever UDP datagrams are pending on the socket.

So is this actually a lower level thing than UDP? Since UDP doesnt do anything I wasnt expecting something like that.

Its really a lot useful..To handle "disconnection" this is a lot faster than waiting for a timeout.

Yes, it's faster than a timeout, but it's not guaranteed. Some hosts don't send the ICMP; some networks filter them; some machines may quit by being turned off (not being able to respond at all) etc. So you need the timeout, and can perhaps accelerate detection of some disconnects with this message.

Eh, nevermind again. I just re-read that place that said it was an erroneous error, and the guy was just speaking in a very unclear fashion. My original statement is re-instated. -.-

So is this actually a lower level thing than UDP? Since UDP doesnt do anything I wasnt expecting something like that.

Its really a lot useful..To handle "disconnection" this is a lot faster than waiting for a timeout.

Slightly lower-level. ICMP is part of the TCP/IP suite that handles control messages such as connection failure or pings. It's at the same level as a 'raw' socket (network layer), which UDP (transport layer) is a minimalist wrapper for.

The go-to book for this stuff is TCP/IP Illustrated, Vol. 1. You can get it used for about $10-20.