Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!mcgill-vision!bloom-beacon!ftp!poopsie!bootsie!olson
From: olson@bootsie.UUCP (Eric Olson)
Newsgroups: comp.sys.mac.programmer
Subject: Re: Lightspeed C programming questions??
Message-ID: <11@bootsie.UUCP>
Date: 8 Feb 90 22:15:52 GMT
References: <1990Feb7.174811.3550@aucs.uucp>
Reply-To: olson@bootsie.UUCP (Eric Olson)
Distribution: na
Organization: Lexington Software Design, Lexington, MA
Lines: 54

My mail to you bounced.  This was recently discussed in this newsgroup,
sorry to those who've seen it all before.

Your problem is a known misfeature of the glue for Think C (and MPW)
provided by Apple.  Basically, there is only _one_ actual clickLoop field,
the _last_ one set with SetClikLoop (this is because clickLoop procedures
need some glue, so the glue routine is the real global).  The solution
is to write your own glue routine, and stuff it into the clikLoop field
directly, pointing at your clikLoop Routine directly.  An example:

/*=============================================================================
 TEClickGlue

 THINK C's SetClikLoop function always points to a piece of code like
 this, but with jsr to the last address set with SetClikLoop.  This
 caused every TE Record to share a ClickLoop function (the last one
 set with SetClikLoop).  We stuff the TE clikLoop field directly in
 initialization, and point it here, to avoid this problem.
 =============================================================================*/
          
static Boolean        TEClickGlue()
{
      asm {
         move.l  d2,-(a7)      /* Must preserve d2             */
         clr.b   -(a7)         /* Space for a Boolean return   */
         jsr     TEClickLoop   /* Call the Pascal routine      */
         move.b  (a7)+,d0      /* Pop the result into d0       */
         move.l  (a7)+,d2      /* Restore d2                   */
         tst.b   d0            /* Set status flags for result  */
        }
             
        /********************
         *      We don't put a rts at the end because if compiled with
         *      profiling on, a stack frame is generated and it needs
	 *	to be UNLINKed by falling through.
         *
         *      If Profiling isn't on, no stack frame is generated and
         *      this function compiles with a simple rts at the end anyway.
         ********************/
}

Good Luck!

-Eric


-- 
Please note!  olson@bootsie.uucp will not work!  Use an address below:

Eric K. Olson                 Internet:   olson@endor.harvard.edu
Lexington Software Design     Usenet:     harvard!endor!olson
72A Lowell St.                Applelink:  olson@endor.harvard.edu@dasnet#
Lexington, MA  02173          Compuserve: >INTERNET:olson@endor.harvard.edu
(617) 863-9624                Bitnet:     OLSON@HARVARD