This is already fixed in master branch (see attached patch)

Time for lwIP is short lately for all "active" members of the team, so if you could provide a patch fixing this (I haven't reproduced this, yet)...

This pseudocode would result in lwip_select() waking up for data received once (if less than 1000 bytes), but not second time data is received:

while( 1 )
{
   lwip_select( maxfdp1, &readset, NULL, NULL, NULL);
   lwip_read( s, pData, 1000 );
}


The reason is that when first data is read with lwip_read(), the loop in lwip_recv_tcp() is not exited resulting in calling
netconn_recv_tcp_pbuf_flags() twice which will call:
   ->netconn_recv_data_tcp()
      ->netconn_recv_data()
          ->API_EVENT(conn, NETCONN_EVT_RCVMINUS, len);

That results in decrementing the sock->rcvevent twice which, if only one receive event has occured, will then be -1. So next receive event will restore it to 0, but select() won't wake since it should be >0.

A fix is to change the exit condition for the loop in lwip_recv_tcp() in socket.c at line 1025 from

} while ((recv_left > 0) && !(flags & MSG_PEEK));

to

} while ((recv_left > 0) && !(flags & MSG_PEEK) && (sock->rcvevent > 0) );