Re: PG14: Avoid checking output-buffer-length for every encoded byte during pg_hex_encode

Em ter., 17 de ago. de 2021 às 00:43, Michael Paquier <michael(at)paquier(dot)xyz>
escreveu:

> On Mon, Aug 16, 2021 at 02:06:31PM -0300, Ranier Vilela wrote:
> > uint64 and size_t in 64 bits are equivalents.
> > uint64 and size_t in 32 bits can be weird, but anyway size_t at 32 bits
> can
> > handle 1GB.
>
> There is usually a reason why things are done as they are, so before
> suggesting changing something I would advise to read the threads that
> created those changes more carefully because they could be mentioned.
> In this case, we are talking about aef8948, and this part of its
> related thread:
> https://www.postgresql.org/message-id/20210105034739.GG7432@momjian.us

Because things have always been done a certain way doesn't mean it's right.
Using int to address strings is an mistake.
Even with an artificial limitation to only deal with 1GB, the correct one
to use would be size_t.

>
> > base64.c can be made in another patch.
>
> This file is a set of code paths used by authentication and SCRAM, it
> will never get as hot as the code paths that Hans has reported as
> these are one-time operations. Please note as well cfc40d3 for the
> reasons why things are handled this way. We absolutely have to use
> safe routines here.
>
I have no plans to touch base64.c

regards,
Ranier Vilela