Re: [PATCH v5] Replace {u}int_fast{16|32} with {u}int32_t

[Noah Goldstein <goldstein.w.n@gmail.com>]

On Wed, Apr 13, 2022 at 7:41 PM Paul Eggert <eggert@cs.ucla.edu> wrote:
>
> On 4/13/22 16:59, Noah Goldstein via Libc-alpha wrote:
> > -  return h & 0xffffffff;

Going to make this one uint32_t then as the name has meaning
and we want a uint32_t at the end.

> > +  return h;
>
> Please omit this change, as the "& 0xffffffff" is needed on
> (admittedly-hypothetical) hosts where unsigned int is wider than 32
> bits, which POSIX allows. Omitting this change shouldn't affect the
> generated machine code on typical platforms.
>
> > -           uint_fast16_t type;
> > +           uint16_t type;

Fixed in v6.
>
> Since 'type' is compared to qtypes[i] which is 'short int', I suggest
> using 'short int' here instead.
>
> Otherwise, looks good; thanks.
>
> A bigger cleanup would replace a lot of the 'uint32_t' stuff with
> 'unsigned', as a good deal of the code seems to be written under the
> mistaken assumption that int can be only 16 bits, but that's low priority.