Path: utzoo!attcan!uunet!zaphod.mps.ohio-state.edu!usc!elroy.jpl.nasa.gov!decwrl!deccrl!news.crl.dec.com!decvax.dec.com!ima!iecc!compilers-sender
From: neil@wcc.oz.au (Neil Murray)
Newsgroups: comp.compilers
Subject: Linkers & Loaders, books & article references sought
Keywords: linker, bibliography, question
Message-ID: <9101240056.16069@munnari.oz.au>
Date: 23 Jan 91 05:20:45 GMT
Sender: compilers-sender@iecc.cambridge.ma.us
Reply-To: neil@wcc.oz.au (Neil Murray)
Organization: Compilers Central
Lines: 38
Approved: compilers@iecc.cambridge.ma.us


	On recent perusal of my bookshelf I noticed a goodly number of 
books on compiler theory but very little on linkers & loaders.  

	After a little searching I did find a piece on dynamic linking
contained in gnu sources "dld-3.2.1.tar.Z" on a local archive.  There
is a research report available from the archive-server@decwrl.dec.com
_Link-Time Code Modification_ report 89/17.

	The problem is both these reports deal with advanced techniques or 
extensions to a standard linker.  I'm after a text that covers linkers
from simple ones through overlay linkers to dynamic linkers.  I've seem
loaders as simple as an absolute load and run to the vax/vms image
activator but little on the internal data structures used and algorithms
employed?  What are the trade-offs in various techniques?  Do linkers'
and loaders' capabilities affect compiler writers in non-obvious ways?

	Speaking of capabilities what about capability architectures,
the requirements of the linker/loader must be quite different.

	Does anyone know of some decent books on the subject?

---
Neil Murray  (R&D Dept)		   Phone:  +61 3 764-1100  Fax: +61 3 764-1179
Webster Computer Corporation,	   Email:  neil@wcc.oz.au
PO Box 195,  Ferntree Gully	   UUCP:   {...,uunet}!munnari!wcc.oz.au!neil
Victoria, Australia, 3156.	   

[An excellent question.  Linking is an important topic that is very poorly
treated in the literature.  And yes, what you can ask the linker to do can
have significant effects on languages.  For example, the fact that traditional
C allows you to have many different uninitialized definitions of a variable
is definitely related to the fact that Unix linkers handle the situation
well.  (In many linkers "common" is a special case and there is a low limit
on the number of common areas allowed.) -John]
-- 
Send compilers articles to compilers@iecc.cambridge.ma.us or
{ima | spdcc | world}!iecc!compilers.  Meta-mail to compilers-request.