NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/comp/lang/cplus/19992

Text File | 1993-01-28 | 1.7 KB | 45 lines

Newsgroups: comp.lang.c++
Path: sparky!uunet!charon.amdahl.com!pacbell.com!att-out!cbnewsh!gls
From: [e-mail] (Col. G. L. Sicherman)
Subject: H-P C++ 3.0
Organization: AT&T
Date: Tue, 26 Jan 1993 20:23:52 GMT
Message-ID: <[e-mail]>
Sender: [e-mail] (Col. G. L. Sicherman)
Lines: 34

I've been using Hewlett-Packard C++ 3.0 lately.  It's pleasantly strict
about many things that version 2 was lax about.  On the other hand,
generating the -I options for template linking is a nuisance, and we've
been bitten by a couple of unrelated bugs:

1. The standard components generously let you use version 2 of the
   List class by #including <List_old.h>.  You still have to convert
   your declarations to angle brackets, but that's progress.  Anyway,
   the old routines that return a copy of an element used to deal with
   "no such element" in this fashion:

    // if (no such element)
    // return 0;

   In other words, somebody programmed them to return 0, was perplexed
   when the compiler rejected it, and reluctantly commented the statement
   out.

   In version 3, some of these statements have been uncommented.  The
   result is that you can't compile List<String> at all.  You can't
   compile a list of any type with two or more constructors, unless
   one of them takes an int.

2. CC -S sometimes generates bogus assembly code.  We're getting
   DS (define symbol) directives with an argument of /dev/null.
   Of course, /dev/null is not a legal name for a symbol.  If we
   compile to C first and then to assembly language, there's no
   problem.

We've solved problem 1 by restoring the screw marks, and problem 2 by
compiling to C first.  Does anybody have any better ideas?
--
Col. G. L. Sicherman
[e-mail]