Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!portal!cup.portal.com!Radagast
From: Radagast@cup.portal.com (sullivan - segall)
Newsgroups: comp.sys.amiga.emulations
Subject: Re: Emulator Mechanics (sorry long post)
Message-ID: <39935@cup.portal.com>
Date: 8 Mar 91 00:27:32 GMT
References: <4992@mindlink.UUCP>
  <1991Mar6.010141.5905@mintaka.lcs.mit.edu> <1303@macuni.mqcc.mq.oz>
  <1304@macuni.mqcc.mq.oz>
Distribution: na
Organization: The Portal System (TM)
Lines: 36

Even without self-modifying code, it can be very difficult to
separate code from data.  I will often use functions that 
have inline data (eg:

	call Print
	db "Hello World!",0
	mov  ax,4c00h
        int  21h

or:

	call Case
	db 2
	dw OFFSET default_sub
	db 'A'
	dw OFFSET case_a_sub
	db 'B'
	dw OFFSET case_b_sub
  	jmp loop

)

   This sort of problem is impossible to resolve without an interpreter
running at the same time as the cross-compiler.  By following the path
of the program, you might possibly find most of the code and data
segments, but it is quite easy to write a branch which will only be
taken one out of a million times.  If the compiler assumes that the
conditional branch may or may not be taken in the future (ie: ignoring
the current processor state) it might very well find itself modifying
data rather than code.  The branch may have been written conditionally
to take advantage of the relative addressing inherent in those 
branches so that the code is relocatable on the fly.  I think
you'd stand a better chance with a resourcer, and a cross compiler
allowing you to control the intermediate code, and correct any
obvious flaws, than with a one step process. 
-kls