Re: log1pl expm1l (Re: math/py-scipy import failure on netbsd-current

[Tobias Nygren <tnn%NetBSD.org@localhost>]

On Tue, 03 May 2022 08:16:42 +0900
Makoto Fujiwara <makoto%ki.nu@localhost> wrote:

> I had a patch on
>    math/R-igraph,

That is the workaround I had in mind.
Worth noting that computing this way loses up to 2 bits of
floating point precision when x is close to 0.
(But the standard allows for this.)

> +#define expm1l(x) expl(x) - 1

Please use extra parenthesis here.
Otherwise the macro might expand to unintended computation.

 +#define expm1l(x) (expl(x) - 1.0f)