Path: utzoo!attcan!uunet!mcsun!ukc!slxsys!ibmpcug!robobar!ronald
From: ronald@robobar.co.uk (Ronald S H Khoo)
Newsgroups: comp.unix.xenix
Subject: alloca for 286 xenix (WAS Re: GNU diff for XENIX)
Summary: try C alloca
Message-ID: <1990Mar5.123643.18510@robobar.co.uk>
Date: 5 Mar 90 12:36:43 GMT
References: <1990Feb15.233047.1867@metro.ucc.su.oz.au> <10306@hoptoad.uucp> <952@chinacat.Lonestar.ORG> <190@rdb1.UUCP> <955@chinacat.Lonestar.ORG>
Reply-To: ronald@ibmpcug.CO.UK (Ronald S H Khoo)
Organization: Robobar Ltd., Perivale, Middx., ENGLAND.
Lines: 25

In article <955@chinacat.Lonestar.ORG> chip@chinacat.Lonestar.ORG (Chip Rosenthal) writes:
> Given that the default stack
> size is 0x1000 bytes, and gnu programs tend to alloca() like crazy, it
> isn't surprising that you can't get very far.
Arnold Robbins will want you to say _Some_ GNU programs :-)
But you're right. :-(
 
> I haven't looked at the source close enough to see how much gnu diff beats
> on alloca(), but you might want to fixhdr(C) the executable to make the
> stack bigger.

I suspect that it might be more useful to  use Doug Gwyn's C alloca emulation
instead, as distributed with some GNU programs, including gawk (GNU Awk).
This uses malloc to emulate alloca and should not be too much of a hit for
grep.  I'm surprised that any fixed stack implementation would dare put a
'real' alloca in the ?libc.a at all.

Disclaimer: I don't have 286 xenix running on anything anymore, so I have
no idea how relevant this may be.  I certainly don't remember having alloca
when I _did_ use 286 xenix.

-- 
Eunet: Ronald.Khoo@robobar.Co.Uk   Phone: +44 1 991 1142    Fax: +44 1 998 8343
Paper: Robobar Ltd. 22 Wadsworth Road, Perivale, Middx., UB6 7JD ENGLAND.
$Header: /usr/ronald/.signature,v 1.2 90/01/26 15:17:15 ronald Exp $ :-)