Re: gamma() and lgamma() functions

On Wed, 4 Sept 2024 at 19:21, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> >> I'm not sure why you are doing the testing for special values (NaN etc.)
> >> yourself when the C library function already does it.  For example, if I
> >> remove the isnan(arg1) check in your dlgamma(), then it still behaves
> >> the same in all tests.
>
> > It's useful to do that so that we don't need to assume that every
> > platform conforms to the POSIX standard, and because it simplifies the
> > later overflow checks. This is consistent with the approach taken in
> > other functions, such as dexp(), dsin(), dcos(), etc.
>
> dexp() and those other functions date from the late stone age, before
> it was safe to assume that libm's behavior matched the POSIX specs.
> Today I think we can assume that, at least till proven differently.
> There's not necessarily anything wrong with what you've coded, but
> I don't buy this argument for it.
>

OK, thinking about this some more, I think we should reserve overflow
errors for genuine overflows, which I take to mean cases where the
exact mathematical result should be finite, but is too large to be
represented in a double.

In particular, this means that zero and negative integer inputs are
not genuine overflows, but should return NaN or +/-Inf, as described
in the POSIX spec.

Doing that, and assuming that tgamma() and lgamma() behave according
to spec, leads to the attached, somewhat simpler patch.

Regards,
Dean