Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cornell!uw-beaver!Teknowledge.COM!unix!hplabs!hpfcso!mjs
From: mjs@hpfcso.HP.COM (Marc Sabatella)
Newsgroups: comp.sys.hp
Subject: Re: Linking into an exec file
Message-ID: <7370034@hpfcso.HP.COM>
Date: 16 Nov 89 16:56:51 GMT
References: <6281@tank.uchicago.edu>
Organization: Hewlett-Packard, Fort Collins, CO, USA
Lines: 30

>I am trying to link a normal .o file with an already
>compiled program (a Lisp interpreter) that has hooks
>for calling "foreign procedures" located in its image.
>
>I have tried to use "ld" but it complains that the exec
>file is not of the correct type.  I know about ld's"-A" flag
>but can't seem to get it to work.

An executable file has no relocation information, thus it is (normally) useless
to ld(1).  How did you get the "hooks" into your Lisp interpreter?  The
solution to your problem depends a great deal on that.

An aside on the -A option to ld(1): normally it is used to create an object
file which can be loaded into your program dynamically - ie, you malloc space
for it, invoke "ld -A <current program> -R <address returned by malloc> ..."
from within your program, then fread(3) or read(2) the resultant object file's
text and data segments into memory (and allocate space for BSS as well).  You
can then look up symbol addresses in the resultant file's symbol table and
call the loaded file's functions indirectly.  The exact details of this differ
between the Series 300 and 800.

There are somewhat clever ways of using -A statically, and again, depending on
the details of your "hooks", you may be able to get it to work.  By the way,
there is a known bug in the -A option which makes it incompatible with -e, -h,
and -u on the Series 300.  This has been fixed for 7.0

--------------
Marc Sabatella
HP Colorado Language Lab
marc%hpfcrt@hplabs.hp.com