Tony Harminc wrote:
Paul Gilmartin wrote:
Why all this mickeymouse? Why doesn't ATTN simply work?
It's partly an ISPF thing; they broke the TSO attention paradigm many years
ago. Even the most basic use of ATTN doesn't work properly under ISPF; once
you've hit it, you can't return to the command in progress, as you can if
you started from a READY prompt.
It's completely up to the application. We support ATTN processing in our
ISPF-based products. For example, long-running FIND commands in (E)JES
can be easily stopped by pressing ATTN. (Might need to press it twice if
using a VTAM session manager. But, in that case, our application still
only sees it once.)
Another problem - and a more fundamental one - is that if the serious work
is being done in a system routine (SVC, PC, or system state code), attention
processing will usually be deferred.
Thankfully! This deferral is usually very short lived.
The real problem is that attention handling in TSO/E is strictly an
application issue. Most system services don't "play the game" and, even
if they wanted to, there isn't any kind of address-space-level
indication that ATTN was pressed (except that which you implement your
own). Sure, the exits get scheduled in an IRB. But how does the code in
the exit tell the application code running in the PRB to "break"? You
can set a flag in a shared data area. But, that's only good if the flag
is tested periodically by the mainline. What if the mainline is not
currently running your application's code? Or what if it's stuck waiting
for a resource of some kind? Any program that doesn't establish and
process its own attention exits has the potential to not recognize when
ATTN has been pressed. I'm afraid that's most programs. :-(
In (E)JES, we implemented a scheme where one ATTN "breaks" (by setting a
flag that is periodically tested) and a second ATTN forces an abend that
forces the TCB into recovery. Works pretty well. But sometimes people
get gratuitous dumps from impatient users.
--
Edward E Jaffe
Phoenix Software International, Inc
5200 W Century Blvd, Suite 800
Los Angeles, CA 90045
310-338-0400 x318
[EMAIL PROTECTED]
http://www.phoenixsoftware.com/
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
