From: clewis@ferret.ocunix.on.ca (Chris Lewis)
Newsgroups: comp.lang.postscript,comp.text,comp.sources.bugs,comp.sys.apollo,comp.sys.hp
Subject: Psroff problems with Apollo - solved!!!!
Date: Thu, 9 Jan 92 01:16:24 EST
Organization: Elegant Communications Inc., Ottawa, Canada

[Sorry for the wide cross-posting.  This stuff seems a bit
esoteric, but it will be familiar with some psroff users.
Especially the Apollo ones.  But this is the best avenue
for spreading the information.]

Configuring psroff to work with Apollos has seemed to be a bit of hit
and miss affair - a configuration working on one machine wouldn't work
with another.  All boiling down to problems generating width
tables.  Has been a real support nightmare for me.

Well, the problem with Apollo SR10.3 has been solved.  SR10.3
ships with a version of troff that can't even read its own width
tables properly (like DEC RISC/Ultrix...).  Because of this, the
program psroff uses to analyse SR10.3's COFF-headered objects used as
width tables is fooled, and HEADERSIZE would get set improperly.

An improper HEADERSIZE leads to very poor character widths
(some characters overlapping) and numerous "floating exception"
messages coming from troff/psroff.

These problems caused many people to abandon psroff on Apollos,
or disable the use of the width tables by removing ".fp"'s
(font loads) in the troff header files used by psroff.
Definately suboptimal ... they're using the widths compiled into
troff which are for a WANG C/A/T phototypesetter, which only
roughly approximate Postscript widths, and are rather dramatically
different from HP Laserjet.  Without this correct, they might as
well be using thack - thack's a lot smaller and easier to install.
But thack doesn't have width tables...

For SR10.3, the correct HEADERSIZE value is 304.  Even though
"dumpft" says 336.  I think that 336 is the right setting for
SR10.2.  I would like to thank Keith Cantrell for having the
patience to help me debug this over the phone.

For those people who have had trouble, try setting HEADERSIZE to
304, rebuilding and try "make test".  You'll be very pleased with
the result.  If you still have trouble, please contact me.