Re: [PATCH] full-read, etc.: prefer signed types

[Paul Eggert]

On 2024-08-09 14:56, Bruno Haible wrote:
> But why also for full-read and full-write? These have a
> calling convention like fread() and fwrite() — a result < count
> is an error —, and fread() and fwrite() return size_t.

It's for the same reason I'm gradually switching to idx_t elsewhere. In 
well-controlled apps object sizes cannot exceed PTRDIFF_MAX so there's 
no generality loss in replacing size_t with idx_t. And there's a win in 
switching, because idx_t arithmetic is less error-prone.