Input output parity & stop bits (was: Re: autowrap)

[Guy Harris]

> Correct me if I am wrong, but I thought clearing both the EVEN and ODD
> bits in the flags member of the sgtty (or whatever) structure provides
> no parity I/O.

It doesn't - at least it doesn't on 4.[23]BSD nor on any SunOS release with
which I'm familiar.  In those systems, it requests 7 bits and even parity, but
tells the system not to discard characters that have parity errors.  This is
not the same as what is being requested, which is to send and expect 8 bits
and neither send nor expect a parity bit.  The latter mode is useful, for
example, with terminals that have a "meta" key that turns the 8th bit on for
characters transmitted when that key is held down, or for terminals such as the
VT220 that support an 8-bit character set.

In 4.2BSD, you can get 8 bit, no parity I/O only by turning RAW or LITOUT mode
on.  These are not generally useful for everyday use; RAW turns off minor
features such as line editing and interrupt characters, while LITOUT turns off
equally minor features such as mapping '\n' to CR-LF.  In 4.3BSD, and in SunOS
3.2 and later (and in other systems that have picked up the 4.3BSD tty driver),
you can get 8 bit, no parity I/O by turning PASS8 mode on.  This puts the
hardware into 8 bit, no parity mode, *and* tells the tty driver not to strip
input to 7 bits.  Unfortunately, it doesn't tell the tty driver not to strip
*output* to 7 bits; to get that, you have to turn RAW or LITOUT mode on, with
the problems already mentioned.

In System III, System V, SunOS 4.0 and later, and other systems that have a tty
driver with the S3/S5 interface, you can get 8 bit, no parity I/O by setting
the character size to CS8, turning off PARENB, and turning off ISTRIP.  In this
mode, the tty driver will strip neither input nor output to 7 bits (definitely
true for SunOS 4.0, probably true for S5, possibly true for S3; the S3 and S5
drivers use 8-bit bytes internally to indicate delays, so there may be some
problems - I think the S3 driver didn't handle these problems, but the S5
driver, or at least the S5R2 and later driver, does).

In SunOS 4.0, which also supports the V7/BSD-style "ioctl"s, turning PASS8 mode
on sets the character size to CS8 and turns off PARENB and ISTRIP, so it gives
you a true 8-bit data path in both directions.

> On a related note, I log in from home quite a bit & use xmodem to
> transfer data. I set my local machine to 8 data, no parity (which
> works !!) and 1 stop bit. All is well till I try an upload, at which point
> the system chokes. If I reset my local end to 8 N 2 all is well, i.e. I
> assume the UNIX box is running 8 N 2.

If the UNIX box is running a V7-style (which includes 4.xBSD) driver, the only
way it can run with 2 stop bits is when it's running at 110 baud.  If it's
running an S5-style driver, you can control the stop bits and the baud rate
independently.