kaz@helios.crest.nt.com (Kaz Kylheku) writes:
>No. C is perceived as being portable. Those who perceive C++ as portable
>are naive or mistaken.

>In article <3459AC95.1D75@dynamite.com.au>,
>Alan E & Carmel J Brain  <aebrain@dynamite.com.au> wrote:
>>Now C++ on the other hand, written using CodeWarrior 10 on a Mac, ported
>>to CodeWarrior 10 on an IBM... or even MVC++ 4 vs MVC++ 5... or worse
>>still CodeWarrior 9 on a Mac to CodeWarrior 10 on a Mac to MVC++ 5 on an
>>IBM... In 15,000 LOCs of C++ how many would you reasonably expect to
>>have to be changed? (yes, these were actual examples too)  

Let me provide a data point.
I am using a modern UNIX system, based on an architecture that has been
around for 10 years or so.  We have the very latest C++ compiler from
the vendor.  We have another C++ compiler.

I picked up the September 1997 issue of the C/C++ Users Journal,
and downloaded all of the source code for that issue.
How many of the C++ programs could I compile?

	NONE OF THEM.

I got one of them working after about half an hour of hacking; the
others defeated me.

Porting from one version of CodeWarrior to another is no real test of
portability; no test at all really.  C++ compiler differences are
*much* bigger than machine differences.

Porting from one version of MVC++ to another is no real test of
portability; no test at all really.  C++ compiler differences are
*much* bigger than machine differences.

Porting between CodeWarrior and MVC++ *is* a test of portability,
but it still is a rather pathetic test.

If I had to port 15 kSLOC of C++ from a Mac or a PC to this system,
I would expect major changes to every translation unit and minor
changes to every class.  I would not be in the least suprised if
a week's work led to total failure.  I am aware of several C++
projects which have had to abandon initial plans to release the code
on more than one machine; there was one commercial development
started here that ran into that very problem last year, and another
up the road the year before.

Let's just take a few of the things I'm seeing in C++ these days that
are NOT supported by all compilers:
	namespace
	mutable
	explicit
	static data members initialised in a class definition
	template<>	(that is, a template with no parameters).

>Even though it's easy to introduce machine dependencies into C or C++ code, I
>would expect that most of your C++ headaches would be related to a lack of
>standardization, or the different pace of adaptation of the draft features by
>various vendors. (That alone is, to me, reason enough to avoid C++ if I can).

That is precisely the problem.

>Then there are uses of non-portable functions that are not in the standard
>library. E.g. it's impossible to port an XWindow application to Microsoft
>Windows without rewriting portions of it, but it's not really the fault of the
>underlying language.

(A) Let us all thank and praise Ousterhout for Tk.  (And there's an Ada/Tk
    interface.  Portable GUIs in Ada.)

(B) There's an outfit called Mainsoft (www.mainsoft.com) selling something
    called MainWin, which they claim lets you run NT code under UNIX.
    I know _nothing_ about the company or their products other than what
    I've seen in their ad.  Anyone know how well it works with the Ada
    Win32 binding?

-- 
John �neas Byron O'Keefe; 1921/02/04-1997/09/27; TLG,TLTA,BBTNOTL.
Richard A. O'Keefe; RMIT Comp.Sci; http://www.cs.rmit.edu.au/%7Eok