RE: [PATCH v4 1/2] compiler.h: add const_true()

[David Laight <David.Laight@ACULAB.COM>]

From: Linus Torvalds
> Sent: 17 November 2024 18:00
> 
> On Sun, 17 Nov 2024 at 09:42, David Laight <David.Laight@aculab.com> wrote:
> >
> > #define const_true(x) __if_constexpr(x, x, 0)
> 
> No, let's not do this "double expansion" thing again.

It would be better that the proposed define :-)

> If we actually want to make things smarter, the trick to use is to
> know that only a constant _zero_ turns into a void pointer (aka NULL).
> 
> IOW, something like this:
> 
>    /*
>     * iff 'x' is a non-zero constant integer expression,
>     * then '!(x)' will be a zero constant integer expression,
>     * and casting that to 'void *' will result in a NULL
>     * pointer. Otherwise casting it to 'void *' will be just
>     * a regular 'void *'.
>     *
>     * The type of '0 ? NULL : (char *)' is 'char *'
>     * The type of '0 ? (void *) : (char *) is 'void *'
>     */
>     #define const_true(x) \
>         _Generic(0 ? (void *)((long)!(x)) : (char *)0, char *: 1, void *: 0)
> 
> should work, and doesn't do any double expansion of complex arguments.

I'm sure I have one place where I did want other than 1 or 0.
I do remember moving the '* 0' into the wrapper for __is_constexpr().

Now than min/max don't use __is_constexpr() I wonder if it still has
to be sane for pointers?
Supporting pointers just makes life hard - especially since (void *)1 isn't
constant.

I think everything can be built on a base if_const_zero(x, if_z, if_nz)
#define const_true(x) if_const_zero(!(x), 1, 0)
#define is_constexpr(x) if_const_zero((x) * 0), 1, 0)
which gives a bit more flexibility.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)