[En-Nut-Discussion] killing threads waiting for a connection in NutTcpAccept

[Harald Kipp]

José Vallet wrote:
> Harald Kipp wrote:
>> José Vallet wrote:
>>
>>> I guess that for now we will have to live with a dirty solution to 
>>> unblock NutTcpAccept.
>> It may also work to close the stream and/or socket handle from another
>> thread, if these handles are globally available.
> 
> Actually closing the socket is what we tried first, but unfortunately it 
> didn't work. When the socket is closed NutTcpAccept continues blocking 
> the thread.
> 
> The stream is created right after NutTcpAccept, so we cannot close it 
> before.

José,

You're right, my hasty answer was blindfold.

In fact the socket is in TCPS_LISTEN state. When closing this socket, it
is simply destroyed, leaving the listening thread in a blocked state.
I'd consider this a bug.

NutTcpAbortSocket(sock, ECONNRESET);

should help here. Although its description states, that it should not be
called by application code, it is cleaner than using the internals of
the TCPSOCKET structure.

Harald