On Sun, Aug 19, 2018 at 01:15:58AM -0400, Tom Lane wrote:
> Nico Williams <nico@cryptonector.com> writes:
> > On Sat, Aug 18, 2018 at 04:34:50PM -0400, Tom Lane wrote:
> >> So now I'm about ready to propose that we just *always* use
> >> snprintf.c, and forget all of the related configure probing.
>
> > You'd also get to ensure that all uses from *die() are
> > async-signal-safe.
>
> [ raised eyebrow... ] That seems like more than I care to promise
> here. But even if snprintf itself were unconditionally safe,
> there's plenty of other stuff in that code path that isn't.
One step at a time, no? And there's the other benefits.
