[comp.emacs] Gnuemacs 18.45 on AT&T 3B15

USGCORB@LEHIIBM1.BITNET (Stephen Corbesero) (07/13/87)

After much frustration in simply getting the software, I am now faced with
'Fatal Error (10).bus error'.  Does anyone have any experience with
GnuEmacs and/or the 3B15 (running System V.2.1) to offer me some helpful
household hints?

Thanks in advance.

Karl.Kleinpaste@cbstr1.UUCP (07/15/87)

USGCORB@LEHIIBM1.BITNET writes:
   After much frustration in simply getting the software, I am now faced with
   'Fatal Error (10).bus error'.  Does anyone have any experience with
   GnuEmacs and/or the 3B15 (running System V.2.1) to offer me some helpful
   household hints?

I have a number of 3B15s here, and I have been using GNU Emacs on all
of them for quite some time, i.e., since the vicinity of 18.26 or
thereabouts, when general 3B support was added by me and others from
whom I borrowed much.  This machine (cbstr1) is a 3B15.

I have no troubles at all installing GNU Emacs; it's a totally vanilla
operation, just as the dist-18.45/INSTALL file should tell you.  That
much said, you can go back and check the obvious.  Don't be insulted
by the following list; most of them I have done myself...
    [1] Did you specify m-att3b.h and s-usg5-2-2.h in src/config.h?
	(s-usg5-2.h should work equally well.)  (It's a bad morning
	when you mis-compile on a 3B with m-vax.h. :-)
    [2] Is your process size limit (aka MAXMEM) suitably large for the
	beast you are creating?  Our 3B15s are set for a 12Mb MAXMEM.
	That's more than you need, but get all you can.  MAXMEM is in
	2Kb pages, so ours is 6144.  You will need minimum 2Mb virtual
	(MAXMEM of 1024) to run well, if at all.
    [3] Make sure that your ulimit is not screwing you.  In all
	probability, your ulimit is 2048, which is in 512b blocks, that
	is, 1Mb max file size.  Pitiably small.  Change the CDLIMIT in
	<sys/param.h> and recompile your kernel if you have sources,
	or use the script etc/ulimit.hack, a bit of grotesqueness I
	created one morning when total frustration took over on a
	binary-only 3B2 I have here.
    [4] A few people have been tempted to turn C_OPTIMIZE_SWITCH back
	on.  Don't do it; /lib/optim for the 3B{2,5,15} [all
	essentially identical processors from the users' view] is
	braindamaged about certain constructs, and overoptimizes.
	Leave C_OPTIMIZE_SWITCH in m-att3b.h disabled.

Other than that, there's darn little to say.  I use it all the time,
and keep sufficiently up-to-date that I re-install a new version every
month or so.  No problems for me.

Cheers,
Karl