On Thu, Sep 3, 2020 at 7:53 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Hm. I would not blame that on -fdelete-null-pointer-checks per se.
> Rather the problem is what we were touching on before: the dubious
> but standard-approved assumption that memcpy's arguments cannot be
> null.
Isn't it both, together? That is, it's the combination of that
assumption alongside -fdelete-null-pointer-checks's actual willingness
to propagate the assumption.
> I'd still leave -fdelete-null-pointer-checks
> enabled, because it can make valid and useful optimizations in
> other cases.
Is there any evidence that that's true? I wouldn't assume that the gcc
people exercised good judgement here.
--
Peter Geoghegan