Re: regex no longer needs to depend on malloc-gnu

[Paul Eggert]

Nathan Kennedy wrote:
> If saving that one test is worth the risk that MALLOC_0_IS_NONNULL will
> always be correctly be defined, e.g. that this will never always be true
> for all potential users:
>
> +#ifdef _LIBC
> +# define MALLOC_0_IS_NONNULL 1
>
> uClibc defines it too:
> libc-symbols.h:#define _LIBC    1
>
> Nathan

I thought you said that malloc (0) returns NULL in uClibc?  If so, then 
why would it be correct to define MALLOC_0_IS_NONNULL when uClibc is in use?

The current code sets MALLOC_0_IS_NONNULL to 1 when it's known that 
malloc (0) returns a nonnull value.  If there's something wrong with the 
test I suppose we can fix it, but that's the proper place to fix it.