Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!cmcl2!philabs!rdin!delftcc!sam
From: sam@delftcc.UUCP (Sam Kendall)
Newsgroups: net.lang.c,net.lang.c++,net.bugs
Subject: Problem with compiling automatically generating C
Message-ID: <127@delftcc.UUCP>
Date: Mon, 24-Feb-86 11:29:54 EST
Article-I.D.: delftcc.127
Posted: Mon Feb 24 11:29:54 1986
Date-Received: Sat, 1-Mar-86 01:07:05 EST
References: <347@watcgl.UUCP>
Organization: Delft Consulting Corp., New York
Lines: 26
Xref: watmath net.lang.c:8002 net.lang.c++:29 net.bugs:744
Summary: Expression tree space is too small on PCC-based C compilers

In article <347@watcgl.UUCP>, mherman@omnitor.UUCP writes:
> Second, [using C++] we ran into problems with the 4.2BSD C compiler with
> expression stack overflows (I'm not real sure it was the expression
> stack any more).  Anyway, one of our trusty Unix people modified cc to
> dynamically relocate the stack when the overflow occurred.

Our product (Bcc, a C checkout compiler), like C++, generates C code
which is compiled by the resident PCC-based compiler.  I think the
problem described above is the same one we run into: automatically
generated code tends to have large expressions, and large expressions
overflow the available storage for parse trees.  For the PCC, that
storage is a fixed-size array.

I urge UNIX vendors either to do what the "trusty Unix person" did, or
just to greatly increase ALL the table sizes in your PCC-based compiler,
the expression tree storage table in particular.  If you have virtual
memory, then making the table sizes very large should not hurt--for most
tables, the unused part will never be touched.  Don't assume merely
"reasonable" limits for table sizes, because automatically generated
programs will exceed The Limits of the Imagination.

----
Sam Kendall			     allegra \
Delft Consulting Corp.		seismo!cmcl2  ! delftcc!sam
+1 212 243-8700			       ihnp4 /
ARPA: delftcc!sam@nyu.ARPA