Stacy Smith wrote in a message to Mike Bilow:

 SS> EMFBI, but I've always called INT 28h to give up time slices
 SS> in the OS/2 DOS box.  I don't recall where I picked up this
 SS> information on giving up time slices in the OS/2 DOS boxes,
 SS> but it seems to have worked reliably for the past few years,
 SS> using Pulse as a measure.  Changing to the Windows method
 SS> would be a no-brainer, but I'm curious as to why two time
 SS> slice release calls under OS/2...

This is a pure DOS issue, unrelated to OS/2.  DOS itself calls Int 28h
every so often to let background tasks run, such as the PRINT TSR.  DOS
internally interprets Int 28h as a notification that things are in a safe
state for a TSR to pop up, for example.  Int 28h is intended to be called
from DOS, not from applications, since an application has no knowledge of
whether DOS is in a stable state or not.  In particular, DOS is not
re-entrant -- either natively or under OS/2 within a DOS session -- and it
therefore does all sorts of dangerous things such as switching static
stacks.  In general, you should not call Int 28h unless you know exactly
what you are doing.

On the other hand, Int 2Fh AX=1680h is a notification to OS/2 (or any other
DPMI server), not to DOS.  It should be used in preference to Int 28h if
your goal is to release the timeslice rather than to notify the TSR chain
about the internal condition of DOS.
 
-- Mike


---