Re: [lwip-users] HTTP Auth, LWIP connection management and FIN packets

[Sergio R. Caprile]

There will be a FIN, you are instructing the server to close the
connection, you are sending "Connection: close" in the HTTP/1.1 header.
If the connection is lost, you will receive a callback.
If you can't wait, start a timer...

Nevertheless, if you still want to close on client side, you can do so
as you are doing now, but if the RST breaks something (it shouldn't),
you can start a timer, stopped on server closure or error, and firing
closure on expiration.

On 03/03/2016 01:14 a.m., Sergio R. Caprile wrote:
> The piggybacked SYN in frame 4 is considered part of the stream and so
> the sequence number in frame 5 is one more than what you would expect.
> The RST flag is also a correct response because you close the connection
> before the stack can process the FIN (what it would do after you return
> ERR_OK, and will call you again to inform you of the closure)
> RFC1192 4.4.2.13
> https://tools.ietf.org/html/rfc1122#page-87
> 
> I'm not sure if the FIN can be considered part of the stream, but I've
> found references to that, and lwIP seems to be doing so.
> 
> If this is true, the test I suggested should end with the expected FIN
> ACK on the client side.
> 


-- 
Sergio R. Caprile, Human Being, Bs.As., Argentina
        http://www.scaprile.ldir.com.ar/