Bug 24528

	I tried to adapt a Linux application for model railroads
	(Digital Direct for Linux). I got no acceptable acknowledge from
	the NMRA-programmer application. AFAIS this is due to the
	following defect in the handling of Modem status signals:
	The input signal is on RI. After triggering the corresponding
	status bit further reads via TIOCMGET don't reflect the state
	of the line until output occurs.

Fix: 

IMHO TIOCMGET should update 
	the internal copy of MSR, at least after reading. Thus the next
	TIOCMGET gets the right info.

	I saw that sio is heavily affected by current development.
	So I just fix it at home, and give you just the information.

	Further comments on the 16550A driver. I feel that there are
	reasons to relieve unnecessary restrictions on the baud rates.
	The hardware is able to produce at least 3 more useful Baud rates:
	1. (and least) NMRA timing needs 58us pulses resulting in ~17000Bd,
	2. The bus of my NMRA-DCC is on 62500 Bd,
	3. (probably the only important) MIDI is on 31250Bd.
	You could just accept any Bd rate not exceeding some upper limit
	for this device.
How-To-Repeat: 
	Use kermit and 'show modem'. RI goes to off only after output    
	(in connect state e.g.).