Excerpted from message dated 11-24-95, David Noon to Murray Lesser:

DN>On Wednesday, 95/11/22, Murray Lesser wrote to David Noon about "Pl/i
  >Compile Options" as follows:

Hi David--

    We are now out of phase.  I posted a correction to my message of
11/22 headed "PL/I Query" on 11/25.  Among other things, I pointed out
that I had asked the wrong question.  I also included an excerpt from a
test source code.  My local BBS apparently wasn't getting a Fidonet feed
for a couple of days, so if you didn't receive the 11/25 post, let me
know.

DN>I have found some bugs in the %INCLUDE files shipped with PL/I, both
  >compiler and toolkit. The latter had fewer, so that is what I use.

    The most egregious bug in OS2.CV2 (PLITK CSD#1) is that it declares
INCL_NOPMAPI and then immediately tests to see if it has a value.  Since
it now can't, OS2.CV2 then loads in PM.CV2, which proceeds to load in
all the rest of the PM %INCLUDE files.  This is very discouraging to
those of us who write character-mode apps, so I continue to use the PL/I
compiler CPY %INCLUDE files.

DN>I suppose I should offer to write another article for the newsletter;
  >perhaps something a little more technical this time. Perhaps
  >something on using PL/I's ability to define new statements in order
  >to encapsulate the API. This is a topic that has intrigued me for a
  >couple of years.

    If I understand what you are saying, this would be a great idea.
Even if I don't understand, it probably still is a great idea .  I
find that putting the API prototypes into the program itself, rather
than including OS2.CPY, shortens the compile time immensely, as well as
reducing the number of "warning" messages (not declared, assumed
character) one gets.

    FYI, I have assumed Christie has vanished because the last several
posts I have made to teampli{at}vnet.ibm.com have been answered by Carolyn.
Also, Shauntel Christensen is listed as Editor for Issue 4 of The PL/I
Connection.  I do find Carolyn a very pleasant person to exchange e-mail
messages with, so you may well find her to be an acceptable substitute
for Christie.  Try it, you may like it.

DN>Try declaring the keyboard file as UNBUF and with RECSIZE(1), then
  >open it as device STDIN: with TYPE(FIXED). This should allow you to
  >read keystrokes as they occur. Something like...

    As pointed out in the 11/25 post, I really wanted an equivalent to
_getch, rather than to CHARIN.  My problem is to display "Press any key
to continue" and have it work, cleaning out the logical type-ahead
buffer, without requiring a .  My app should block on keyboard input
while running a second thread that beeps every second, until the
keystroke restarts thread #1 to DETACH the second thread.  I am almost
ready to try writing my own function using the KBDCharin API call.  I am
not sure whether I want to try it by calling the appropriate OS2
%INCLUDE files (now that I know how to make them work), or by putting
the function prototypes into the program itself.  In any case, I will
soon learn how the PL/I compiler handles thunking.

    Thanks for the tag line.

          --Murray

___
 * MR/2 2.25 #120 * Fidonet is almost like having a social life
                                                                           
                                  
---