Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!pasteur!ucbvax!TAURUS.BITNET!leonid
From: leonid@TAURUS.BITNET
Newsgroups: comp.arch
Subject: Re: self-modifying code
Message-ID: <670@taurus.BITNET>
Date: 3 Mar 88 15:21:45 GMT
References: <486@taux01.UUCP> <902@athos.rutgers.edu>
Sender: daemon@ucbvax.BERKELEY.EDU
Reply-To: virgo!leonid (Leonid Rosenboim)
Organization: Tel-Aviv Univesity Math and CS school, Israel
Lines: 13

May I add to this discussion that on most UNIX system with Virtual Memory
(e.g. BSD 4.2/3) there is alot of self-modifying code. Except for the
debugger case, the idea is implemented by copying some code onto the
stack, modifying it there and executing. Some examples I can think of
are signal handling, Floating-Point hardware dependent code on SUN.
Since dynamic loading is not yet impelemented on BSD systems I dont
really know how they're gonna do it.

In a word, VM UNIX systems tend to keep the text segment readonly
for the sace of text sharing, swap space saving etc, with the few
exceptions where that stack segment is used to keep temporary copies
of executable code that must be modified. Debugging is an exception.

Leonid