Subject: Re: Keyboard problems.
To: None <port-sun3@NetBSD.ORG>
From: der Mouse <mouse@Collatz.McRCIM.McGill.EDU>
List: port-sun3
Date: 08/10/1995 11:19:44
> I occasionaly get a bell (^G) that doesn't want to stop.  At this
> point, X seems to die because messages are suddenly displayed on the
> console, in stead of the Xconsole window, messing up the X-display.
> Xdm then takes charge and restarts the server etc. (keyboard still
> beeping), and after a while I can login again, and then make the new
> xterm go beep, which finaly silences the keyboard.

I've seen (er, heard) this happen too, under SunOS.  Usually on a
SPARC, but the keyboards are pretty much the same.

I've also got a Sun-3/260 running SunOS 3.5, with a type-3 keyboard.
Occasionally, usually when I'm typing fast, I'll hear a brief aborted
keyclick sound (I don't normally have keyclick on) and I'll lose about
three keyboard events; the effect is to lose about two keystrokes, and
the first keystroke lost is always an e.

On other occasions, for no visible reason, it will get the down event
and lose the up event for some key, which (since autorepeat is handled
by the X server) leads to "stuck autorepeat" behavior.  This too seems
to happen most when two keys are pressed in very quick succession; a
particular offender is ESC-p at the command line, calling back a
previous command.  (The p starts autorepeating.)

I have not noticed any correlation with much of anything else.  Since
the keyboard communication is at only 1200 baud, and it seems to
correlate with typing very fast for short bursts, I wonder if it's just
overrunning the serial line?

> I don't recall any such problems on the ttya/ttyb ports, but then
> again they have only been in use for a few days now, after getting
> the latest binary snapshot. There are, however, always a few
> characters missing from the output when I run ls, or in the fortune
> or login prompt.  For a given baudrate, these are always the same
> characters when I repeat the command.

You mentioned you use bash, which I believe does command-line editing.
This sounds a lot like a bug I recall hearing about somewhere, maybe
even on a NetBSD list, which caused a character to get dropped if the
tty modes were changed while output was in progress.

					der Mouse