Re: [musl] [PATCH 1/1] reallocarray casting overflow check explicitally

[Rich Felker <dalias@libc.org>]

On Tue, May 11, 2021 at 09:33:06AM +0200, Szabolcs Nagy wrote:
> * David CARLIER <devnexen@gmail.com> [2021-05-10 20:48:26 +0100]:
> > From 04e3caa580cfe501ad00d85d63040de329962823 Mon Sep 17 00:00:00 2001
> > From: David Carlier <devnexen@gmail.com>
> > Date: Mon, 10 May 2021 20:45:12 +0100
> > Subject: [PATCH] reallocarray overflow check cast to size_t
> > 
> > ---
> >  src/malloc/reallocarray.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/src/malloc/reallocarray.c b/src/malloc/reallocarray.c
> > index 4a6ebe46..2556c917 100644
> > --- a/src/malloc/reallocarray.c
> > +++ b/src/malloc/reallocarray.c
> > @@ -4,7 +4,7 @@
> >  
> >  void *reallocarray(void *ptr, size_t m, size_t n)
> >  {
> > -	if (n && m > -1 / n) {
> > +	if (n && m > (size_t)-1 / n) {
> 
> you need to explain this change more i think.
> 
> the usual arithmetic conversion rules of c already convert both
> operands of / to size_t, unless size_t has lower conversion rank
> than signed int, but such a system would have problems because
> size_t was always promoted to int so you would need a lot more
> casts in musl (and other software) to avoid signed int overflow
> and negative arithmetics.

I think the idea was just to make it explicit, but generally casting
is discouraged in musl. The corresponding code in calloc has a cast
but it's a historical artifact and contrary to the coding style used
throughout most of the source.

Rich