[rescue] More Terminal & SGI Fun

[Jonathan C. Patschke]

     Okay, I admit it.  I like serial terminals.  My primary
applications these days are PINE (for mail), $EDITOR (either vim or
xemacs, depending on my mood), some SQL server console (mysql or psql,
depending upon the project), and the shell.  I can put serial terminals
in places where a full console wouldn't fit, and they're generally
trouble-free.

     So, I have a DEC VT420 terminal connected to an SGI Indy.  I've
been using it without problem at 9600baud, no flow control.  I typically
run the terminal in 48x132 mode, and I generally run a screen session.
Let me tell you, switching a 48x132 screen over 9600bps is an exercise
in patience.

     I got the weird idea that it might be less tedious to work at 38.4
kilobaud, which happens to be the maximum rate that both the terminal
and the Indy claim to support.  38.4 kilobaud without flow control just
doesn't work on a terminal that old.  The buffer overflows almost
instantly (I have no idea -why- IRIX is ignoring XOFF/XON) when working
with curses apps.

     My MMJ<->DIN8 cable was wired for data-leads only, so I modified it
as follows:

     MMJ     DIN8      Purpose
     1       7         DTR (vt420) / DSR (Indy)
     2       3         Transmit
     3       4         GND
     4       8         GND
     5       5         Receive
     6       1         DSR (vt420) / DTR (Indy)

     From Table C-1 (page 161) of /Installing and Using the VT420 Video
Terminal/ and from the "4D Compatible Pin Assignments (RS-232)" table of
man serial on IRIX 6.5.16, this seems to be correct, but I've histori-
cally had bad luck with getting RS-232 things to work.


  VT420 Table C-1 (abriged).         from "man serial":
     Pin  Mnemonic         .             ---------
      1   DTR              .            / 8  7  6 \
      2   TXD+             .           (  5  4  3  )
      3   TXD-             .            \  2   1  /
      4   RXD-             .             ---------
      5   RXD+             .4D Compatible Pin Assignments (RS-232)
      6   DSR              .
                           .      Pin  Name  Description
                           .       1   DTR   Data Terminal Ready
                           .       2   CTS   Clear to Send
                           .       3   TD    Transmit Data
                           .       4   SG    Signal Ground
                           .       5   RD    Receive Data
                           .       6   RTS   Request to Send
                           .       7   DCD   Data Carrier Detect
                           .       8   SG    Signal Ground


     When I tell init to open a getty on ttym2 (instead of ttyd2) and I
tell the VT420 to use the modem-control lines, I get a "Wait" indicator
that does not go away, as if the terminal isn't getting a signal on pin
6.  According to the serial man page, pin 1 on the DIN8 connector is
asserted by the serial driver as soon as /dev/ttym? is opened.

     When I tell the VT420 to not use the modem-control lines, but still
tell the Indy to open /dev/ttym2, the terminal acts as if there is no
flow control, as would be expected.  However, since I'm getting text
back from the Indy, either the terminal is asserting pin 1 anyway, or
the Indy is ignoring my request for flow control (yes, I did init Q and
killall getty).

     I don't have a serial protocol analyzer here to tell whom is at
fault.  I'm willing to blame the connector (although a quick check with
a multimeter showed no unusual shorts/crosses, and no cold joints), since
I'm pretty green at soldering, and DIN8 connectors are really...inter-
esting solder project for a newbie.  However, if I've made some fun-
damentally incorrect assumption here, I'd appreciate a hit with the
cluebat.  If the stuff here looks okay, it could be something as simple
as dyslexically flipping pins 1 and 2 or 8 and 6 on the DIN8.

     One thing that bothers me is that, on page 159 of the VT420 manual,
the text reads:

  The VT420 can process text (without ESC sequences) at approximately
  2000 characters/second.  This rate allows the VT420 to communicate at
  19,200 baud.  Some editing functions, such as inserting or deleting
  characters in lines, require additional time to process.

     Is it folly to think that I might be able to run the terminal at a
line speed of 38.4 kilobaud reliably?  Are later DEC terminals, such as
the VT525 more capable in this regard?  If DEC made no terminals that
could behave reliably at 38.4 kilobaud, who did?  Or, is it time to use
a SPARCstation ELC-turned-Xterm instead of a VT?


-- 
Jonathan Patschke
   > Can you SysAdmins tell me what might go on in a typical day?
   Hours of endless frustration punctuated by moments of sheer terror.
                                 --Saul Tannenbaum (in the Monastery)