From:	SMTP%"eychaner@SUNCUB.BBSO.CALTECH.EDU"  7-FEB-1994 10:25:57.35
To:	EVERHART
CC:	
Subj:	Re: Help pls: ASTs and SMG

X-Newsgroups: comp.os.vms
Subject: Re: Help pls: ASTs and SMG
Message-Id: <2ivfha$20r@gap.cco.caltech.edu>
From: eychaner@SUNCUB.BBSO.CALTECH.EDU ("Channel Energy")
Date: 5 Feb 1994 06:49:14 GMT
Reply-To: eychaner@SUNCUB.BBSO.CALTECH.EDU
Organization: Big Bear Solar Observatory, Caltech
Nntp-Posting-Host: suncub.bbso.caltech.edu
Lines: 80
To: Info-VAX@CRVAX.SRI.COM
X-Gateway-Source-Info: USENET

palmer@qut.edu.au writes:
>Greetings all,
>	I have a small problem with SMG. I wish to put a clock in the corner
>of the screen of my application.

An interesting question, and one I have been thinking about recently
myself.  I'd appreciate any hints.  Here are a few of my thoughts on
the subject (I apologize for quoting so much text):

>Currently I am just running an ast routine via $setimr.  I look up which
>display has the cursor - update the clock and restore the cursor position
>to the 'known' position in that display and that works BUT.
>The SMG manual makes clear reference to 'not re-entrant' and 'not AST
>re-entrant'. I am concerned that depending what is currently being done in
>non AST code could get screwed up by the clock AST routine.
>The manual is not clear as to what one should do. Of course I can disable
>ASTs using SETAST - a big hammer.
>
>The mechanisms I am considering at the moment are -
>1) Disable AST's while any non-ast code is performing SMG functions -
>	looks one hell of a problem for read functions!

The solution for read functions is as follows.  Inside your AST routine,
issue a SMG$CANCEL_INPUT.  Then make your input routines requeue when a
return code of $_CANCEL or $_ABORT, like so:
(C code-fragment example for $READ_STRING, init_d and ret_d are string
descriptors.  Note: this routine uses a separate single-line window for
input, thus the $HOME_CURSOR rather than a $SET_CURSOR_ABS.)

/* Loop each time input is cancelled by AST */
do {
    /* Copy last return string to initial string */
    STR$COPY_DX (&init_d, &ret_d);
    SMG$HOME_CURSOR (&disp_id, 0);
    status = SMG$READ_STRING (&kb_id, &ret_d, &prompt_d, &maxlen, &mask, 0,
	    0, &length, 0, &disp_id, &init_d, 0, 0, 0);
    }
while (status == SS$_ABORT || status == SS$_CANCEL);

This works for broadcast trapping as well, eliminating a lot of hassle
with finding the cursor and repositioning it.  I find that using this,
I don't have to bother with finding the cursor and putting it back, at
least in broadcast-trapping routines.
Thanks to whoever pointed this out to me in the long-distant past.

>2) Use the Batch update feature and disable AST's while the output actually
>	occurs

Personally, I wouldn't worry TOO much about the output, especially if you
put the cursor back where you found it each time.  (Like I said, I never
do in my broadcast trapping routines, and it seems to work.)  How can you
be sure of disabling ASTs while the output occurs?  Disable ASTs, end
batching, and re-enable ASTs?  Is this guaranteed?

>3) Use a broadcast message to the terminal that has a special start and look
>	for that in the AST routine that displays broadcast messages. If it
>	is a time message then update the clock virtual display.

This is quite clever, but also quite slow, I suspect.  You might want to
just have a special broadcast message that tells the broadcast trap to
update the clock, rather than actually sending the time in a message.

>I can't be the only one to ponder this problem. Can someone please indicate
>how it can be done?

I really don't know, other than $CANCEL_INPUT for reads, and possibly
putting the cursor back.

You COULD use $CREATE_SUBPROCESS to associate a subprocess with a 1-line
time window, and then have your AST call $EXECUTE_COMMAND every time the
timer expires to execute a program that prints the time, which would then
appear in the window.
But that sounds REALLY dumb, and eats system resources besides.

Anyone else want to give it a shot?
-G.

Glenn Eychaner, C Programmer & VMS Manager, Big Bear Solar Obs., Caltech
eychaner@suncub.bbso.caltech.edu  (909)866-5791  40386 North Shore Lane
SUNCUB::EYCHANER, 5554::EYCHANER  (818)395-4014  Big Bear City, CA 92314