From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <9801201504.AA13767@clio.rice.edu>
Subject: Re: MCLSSAA2 : hooking interrupt 9
To: Anthony DOT Appleyard AT umist DOT ac DOT uk
Date: Tue, 20 Jan 1998 09:04:18 -0600 (CST)
Cc: djgpp-workers AT delorie DOT com
In-Reply-To: <741A416BE6@fs2.mt.umist.ac.uk> from "Anthony.Appleyard" at Jan 20, 98 08:23:45 am
Content-Type: text
Precedence: bulk

> When the hook routine calls the user function, should it save everything so
> that the user function can be a C/C++ function? Or should it save nothing and
> the user function must be in assembler and itself save and restore everything
> that it uses?

Anything which passes arguments on the stack, or uses automatic arrays
will potentially not work.  GCC assumes that SS = DS, so that you can 
reference addresses interchangably with [EBX] or [EBP].  Simple functions
don't trip over this, complicated ones will.

This is one of the things the wrapper routines do (set up a local SS=DS
stack).  But the extra overhead isn't needed for fast and simple routine.

If we pusha/popa (with seg regs too) and set up DS/ES properly, I don't
see a problem with the routine being in C (or C++) as long as someone
keeps it simple.  A gcc -S and examination of the generated assembler
would be prudent; if you avoid calling subroutines you are probably OK.