Re: must use gcc builtin?

[Alan Barrett <apb%cequrux.com@localhost>]

On Tue, 21 Feb 2012, James K. Lowden wrote:
> Just for fun, I thought I'd try compiling Gary Capell's 
> "wily" editor. The code is from 2006, hardly ancient.  Its 
> library manipulates a va_list as pointer using assignment and 
> subtraction.

I believe that the definition of "undefined behavior" in section 
3.4.3 of the C99 standard applies:

     behavior, upon use of a nonportable or erroneous
     program construct or of erroneous data, for which this
     International Standard imposes no requirements".

I believe that using a va_list as if it were a pointer, and doing 
arithmetic on it, is something on which the Standard imposes 
no requirements, and that the resulting behaviour is therefore 
undefined.

Section 4 paragraph 2 is also pretty clear that "the omission of 
any explicit definition of behaviour" also counts as undefined 
behaviour.  So it's not necessary for the Standard to say "if you 
do <X> then the behaviour is undefined"; if the Standard merely 
does not talk about <X> at all, the the behaviour if you do <X> is 
undefined.

> I can't #undef __GNUC__.  Do we really want to require the use 
> of builtins?  I'd like a knob to compile idiomatic C, especially 
> *good* idiomatic C.

In my world, code that invokes undefined behaviour is not good 
idiomatic C.

> http://pubs.opengroup.org/onlinepubs/009604599/basedefs/stdarg.h.html
>
> does not define a type for va_list.  Assuming it's a pointer 
> might be nonportable, but it's not prohibited by the standard.

The standard hardly ever prohibits a program from doing something 
that's undefined.

--apb (Alan Barrett)