🚀 go-pugleaf

I've seen this on both our 780 and 785 running VMS 4.2 thru 4.5.

From what I have been able to figure out, autobauding works like this:

 * If the line speed is not 9600 or 600 (the autobauding speeds), set
   it to 9600.
 * Read a character
 * If it's recognized in the 9600 baud table, use that speed (9600-1200
   baud)
 * If not, switch to the 600 table and read another character (600-50 baud)
 * If it's recognized, use that speed

Once VMS gets into the second baud rate table, it never gets out.  And the
table doesn't get reset when the line drops, so it's stuck there until
someone hits it at the right baud rate, the system gets rebooted, or
someone hits the BREAK key when they get connected to that line.

I can reproduce the problem by connecting to a line at 300 baud, hit
CR twice and disconnect.  Trying a connect at 9600 baud will get you
nothing, no matter how many CRs you type.  SHOW TERMINAL of that line
will show that the speed is set to 600 baud.

The fix is for VMS to set the line to a known state when a connection
is made.

We have a Develcon Dataswitch, so any terminal here can connect to any
port.  I finally ended up writing a FORTRAN program that checks the lines
to see if anybody is on, and if they're not, it sets the speed to 9600
(which corrects the autobauding problem).  I tried to educate our
users to hit BREAK if they got no response, but that didn't work.

Someday when I get some time I'll write an SPR on this.


Mike Iglesias
University of California, Irvine

Disclaimer:  Like I said above, it's been a while since I looked at the
code that does the autobauding, so there's bound to be something wrong
with some of the above statements.