Allan Mertner wrote in a message to Mike Bilow:

 PW>> Now that could be a real problem.  If I have to write stuff in 
 PW>> C to get Speed up and running, I may as well write C in the 
 PW>> first place.

 MB> I don't know anything about Pascal, but writing a thunk DLL in 
 MB> C should be trivial.  It would amount to tedious typing at 
 MB> worst.  Using the proper _Far16 keywords or whatever was 
 MB> appropriate to your particular compiler, the compiler would do 
 MB> all of the work.

 AM> Sure, but it still required you to have a C compiler and know 
 AM> the basics of how to use it.  Why Speed-Pascal does not just 
 AM> support Far16 calls is beyond me. 

IBM used to provide a thunk toolkit or some such thing, intended to be used
from assembly language.  It's a distant memory, but it had a name like
THK2ASM or THNK2ASM, something like that.

The real complexity of a thunk arises during the link phase.  If SP uses
its own linker, then I could certainly see why they decided not to
implement direct thunking support.  Most compilers actually implement
thunks by generating cookbook code scripts, much the same way they might do
things like extended precision integer math.
 
-- Mike


---