🚀 go-pugleaf

David Brownell (db@argon.Eng.Sun.COM) wrote:
: On the other hand, one of the things I'd really like to see changed is
: how much stuff a C++ compiler does "for me" automagically.  It's a rare

Agreed. What the C++ compiler does is a constant source of bugs.

: So, a pair of proposals that I'd be happier to see accepted:

:    (1)	It is an error for a class to have virtual functions and a
: 	nonvirtual destructor.

I'd vote for it being a warning. Then compilers can be graded on their
user friendliness.

: 	
:    (2a)	C++ compilers don't automagically generate ANYTHING for classes
: 	any more:  constructors, destructors, assignment and address-of
: 	operators, etc.  If such routines are used but not declared and
: 	defined, that is an error.

Agree 100%. Also, any compiler that implements a warning for a class which
only declares one or two of the (ctor, dtor, copy ctor) set would get my vote.

The C++ FAQ has other suggestions for a C++ lint to do.

Also, drop all this "struct is a class with different visibility". Our
programmers here have spent years learning what a C struct is. Don't change
it. Just make the restriction that any struct which defines a method
must be converted into a class. It makes life simple: a C++ struct can
only define data, and the moment it declares a method, then it
must be a class.

I'd also vote for any compiler which produced a warning for an inline method
definition

class K {
	void f() { // no !
	};
..
};

And another warning the compiler implementers out there might consider
is that if a constructor calls new, and the matching destructor does
not, then I want to see a warning for that destructor.

But hey, all this talk of improving the details of the compiler
implementation fails to address the real problem: The ARM has been
publically available for nearly three years now, and I have yet to see
a compiler which implements the ARM ( less Exceptions and Templates
obviously). Ok I will be able to use the full functionality of the ARM
when (if ?) the moment it arrives, but I can't even get a compiler
for a reasonable language subset.

I'm sure the reason for this is the complexity of the language. 
Me, I'd vote for Typesafe ANSI C with Objects thrown in. 
No inheritance, no virtual functions, no polymorphism. The rest is hard
for the compilers to implement, hard for the programmers to understand,
and hard to build reliable systems that work using those features.

Regards
--
David C Binderman     Dept US/ESI, Bau 60, SEL-Alcatel AG, Lorenz Strasse 10,
D-7000 Stuttgart 40, Germany      dcb@us-es.sel.de   Tel: +49 711 821 3841