Re: AT&T compatible C++ compiler (James Kanze US/ESC 60/3/141 #40763)
20 Jul 1996 21:27:53 -0400

          From comp.compilers

Related articles
AT&T compatible C++ compiler (Robert Mueller) (1996-07-18)
Re: AT&T compatible C++ compiler (1996-07-20)
| List of all articles for this month |

From: (James Kanze US/ESC 60/3/141 #40763)
Newsgroups: comp.compilers
Date: 20 Jul 1996 21:27:53 -0400
Organization: GABI Software, Sarl.
References: 96-07-113
Keywords: C++

Robert Mueller <> writes:

|> An OO-database vendor told me that the C++-code generated by their
|> database tools (e.g., by the precompiler) will only work with an AT&T
|> compatible C++-compiler. So what C++ compilers - available for Sun
|> (SunOS5.5) and HP - are AT&T compatible? I assume GNU is NOT (?). And I
|> furthermore assume that these AT&T compatible C++ Compilers are all
|> commercial.

Well, AT&T compatible is pretty vague. Presumably, they mean CFront,
but which version. If it is anything before version 2.1, and they
generate nested classes, you are probably out of luck.

I believe that the standard HP compiler is still a CFront derivative.
The current Sun CC is not, but earlier versions were, and Sun tries to
avoid breaking earlier code.

I suspect that if the generated code is in anyway clean, the only likely
problem with the current versions of g++ is lifetime of temporaries.

And the moderator adds:
|> [They probably mean that they use the ugly cfront C++ to C translator,
|> in some unspecified version. Ugh. -John]

I'd like to defend cfront. The compiler has not been maintained for
many years, and is old, but if you compare it with its contemporaries
(g++ 1.x, Zortech), it was light-years better. Even today, no other
compiler (that I've tried, at least) implements separate compilation of
templates. And "sorry, not implemented" is better than just generating
wrong code, which I've experienced with most other compilers.

I've a certain nostalgia for this compiler. It can't compete with modern
compilers (except for template instantiation), but having learned C++
with an early Zortech, and then using g++ 1.49, you cannot imagine the
wonder I felt when I finally started using Cfront. It was the first C++
compiler which was usable for anything other than just playing around.
James Kanze Tel.: (+33) 88 14 49 00 email:
GABI Software, Sarl., 8 rue des Francs-Bourgeois, F-67000 Strasbourg, France

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.