AT&T 386 UNIX Vr3.1

[William E. Davidsen Jr]

In article <441 at uport.UUCP> keith at uport.UUCP (Keith Hankin) writes:
| In article <145 at carpet.WLK.COM> bill at carpet.WLK.COM (Bill Kennedy) writes:

| >That covers the cosmetics, why do I think that AT&T is better?  Their
| >compiler compiles C-Tree (an incredibly portable package) and the binary
| >that comes out works.  Microport compiles it but breaks when sscanf is
| >called.  AT&T's compiler doesn't write code that the assembler can't assemble.
| 
| Microport's C compiler is THE SAME COMPILER that you get from AT&T.  Any
| problems with Microport's compiler should be the same with AT&T.
| Perhaps you got a very new version of the AT&T C compiler, one which we
| do not yet have to ship.

I'm going to add some information on the 386UNIX C compiler. I have a
benchmark suite which I have run on about 70 machines, without serious
problems. In trying to run it under 386ix I noted that about 30% of the
programs wouldn't compile, most of which caused the compiler to emit
code the assembler couldn't use ("there is no register 25" messages) and
a few "core dumped" cases.

I have seen this on the 386ix, Microport, and Plexus versions. I have
not had a chance to try it on AT&T or Bell Tech, but I assume there is a
problem in the version used by all as a staring point. I reported the
bug to INteractive and Microport, and have no idea if it's fixed.

The Xenix compiler compiled everything, ran everything, and was faster
for the things which ran on both. This was the basis of my personal
decision to run Xenix, and is not a statement that Xenix would be best
in all applications.
-- 
	bill davidsen		(wedu at ge-crd.arpa)
  {uunet | philabs | seismo}!steinmetz!crdos1!davidsen
"Stupidity, like virtue, is its own reward" -me