Re: Using Non-Blocking Sockets

[Scott Klement <rpg400-l@xxxxxxxxxxxxxxxx>]

Jeff,

Each socket has a read buffer and a write buffer.  When you do a send() 
or write() to a socket, it will be placed in the buffer if there's 
enough space, and your program will not need to wait for the data to 
actually traverse the network.

In blocking mode, if there is not enough space in the write buffer, the 
program will stop and wait until space is available, and the send/write 
call will only return control to your program after space has been made 
available in the buffer.

In non-blocking mode, if there's space in the buffer, it will work just 
like blocking mode, but if there isn't space in the buffer it will 
return only as many bytes as could be successfully placed in the buffer 
(i.e. the return value from send() or write() will be smaller than the 
mount you asked to write) and errno will be set to EWOULDBLOCK to let 
you know why the data couldn't be written.

If you get EWOULDBLOCK, you can use the select() API to wait until 
there's space again by waiting until the buffer is writeable.

I'm not familiar with your application, so I can't tell you if this 
would have an "adverse affect", but I would recommend using non-blocking 
all of the time, and using select() when necessary to wait for stuff to 
be ready.  This makes your programs MUCH more robust in my experience.



On 8/20/2015 10:57 AM, Jeff Young wrote:
> I have read the tutorial the Scott Klement has regarding using Non-blocked
> sockets to test for timeouts on a recv function.
> I have a client socket program that connects to a server, sends data and
> then waits for a reply on the same socket that it sent the data on.
> If I set the socket to Non-Blocked at the start of the program, will it
> have any adverse effect on my send of data or will it only effect the
> receive?
>
> TIA
>
>
> Jeff Young
> Sr. Programmer Analyst