Maximum process size on 2.3

[Steve Nuchia]

>From article <524 at wa3wbu.UUCP>, by john at wa3wbu.UUCP (John Gayman):
>     I understand that Microport 2.3 (and earlier) has a maximum process
> size of 1MB. I tried running pathalias on the world and this exceeds
> this limit. So I run that maps on North America only for now. Is this

Isn't it a bitch?

> something thats changable in the linkkit ?  I see Microport has a special
> kernel available upon request. Whats the story ?

It isn't changable in the linkkit, but they will send you the
modified linkkit at no charge, at least if you are on support.
It works fine, I've got mine set for 2 1/2 megs, uname -a says:
    system5 nuchat 2 230.13-I AT

BUT:  pathalias tries to build some kind of data structure, a hash
	table I think, that it rescales dynamically as the input
	comes in.  When it tries to exceed the mallocable size
	(I have a hack that lets it malloc > 65,000 bytes) it
	core dumps.

So I'm still running a subset of the maps, actually the same subset
I was running before.  I'm leaving the large process hack in place
because it lifts the limit on a ihave/sendme optimizer I wrote and
one of my users is working on some games that really want to be on
a vax ...

This is my "mk.paths" script, in case anyone is interrested.
Note that I manage to squeeze in all of the US and the entire
domainized world, which may and may not be better than all
of north america.  Not also the hacks that strip the comments
from the retained copies of the maps and the one that expires
deceased map tables.
-----
find maps -mtime +90 -print -exec rm \{\} \;
cd nmaps
find /files/news/comp/mail/maps -type f -newer ../paths -exec unshar \{\} \;
rm *.hdr
for x in *
do
	(echo 'g/^#/d'; echo w; echo q) | ed - $x
done
mv * ../maps
cd ..
pathalias -f path.input /usr/lib/news/map.entry maps/d.* maps/u.usa.* | pathproc > npaths
rm opaths; mv paths opaths; mv npaths paths
mv analysis oanalysis
cut -f2 paths | cut -f1-2 -d! | sort | uniq -c > analysis

-- 
Steve Nuchia	    | [...] but the machine would probably be allowed no mercy.
uunet!nuchat!steve  | In other words then, if a machine is expected to be
(713) 334 6720	    | infallible, it cannot be intelligent.  - Alan Turing, 1947