Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!rpi!zaphod.mps.ohio-state.edu!usc!snorkelwacker!bloom-beacon!eru!luth!sunic!mcsun!ukc!mucs!cns!umaida!jf
From: jf@ap.co.umist.ac.uk (John Forrest)
Newsgroups: comp.windows.x
Subject: X on Apollos using Inlibs
Message-ID: <1990Jul13.113425.9816@cns.umist.ac.uk>
Date: 13 Jul 90 11:34:25 GMT
Sender: usenet@cns.umist.ac.uk (Usenet news user)
Reply-To: jf@ap.co.umist.ac.uk (John Forrest)
Organization: UMIST Computation, Manchester, UK.
Lines: 41

If you don't use Apollo/Domain computers you can stop reading here.

We put this message up a while back, but got no real reply. Essentially
has anyone managed to get the MIT R4 distribution to work on OS10 Apollo's
using global inlibs? We do it with local inlibs fairly successfully - the
essential changes to the distribution are related to some extra shell
scripts in utils which are used instead of cc and ar (they call cc and ar, 
but not always in a simple manner). Various lines for load options are also
changed, but not by much. The current setup inlibs all library modules that
appear in files called .shared in the appopriate source directories. This
allows us some form of control. We currently inlib Xlib, Xt, Xaw, Xmu and
the extensions library [ We run an "Athena" environment without motif or 
XView, but the same principles apply ]. The problem with this is, however,
that the initial data portion of these libraries appears in virtually
every process (X or not), and  add's up to ~ 100K of redundant data per 
process. Compared to the plus side that simple X programs such as xload take
only 5K of code space, this seems worth it. However, a better approach,
if possible, would be to use global inlibs - where the initial tables could
be made read-only, and thus the total memory usage would be a lot less.
However, my experience with global inlibs is limited - I have got one library
to work (a replacement for the supplied BIND resolve library), and found I
could have only these sorts of variables:

Read Only Initialised Static Variables
RW Uninitialised Static Variables
RW Uninitialised Global Variables
RO Initialised Global Variables, but with the Data Replaced by 0's!

Actually, I'm not completely sure about the latter. It might be I've done
something wrong. The point is that I did manage to adapt the resolve library
I ported, but only by making changes to the source - given the size of that
library, too many really. Has anybody tried the equivalent with the X
libraries?
Even if all the ClassInitialize procedures are called at initial boot time,
it ought to be possible? (shouldn't it?). I'd like to think someone has had
some luck on this matter.

John Forrest
Dept of Computation
UMIST
Manchester UK.