Fw: [PATCH 0/3] libm: Clean up gamma functions

[C Howland]

> ------------------------------
> *From:* Newlib <newlib-bounces@sourceware.org> on behalf of Keith Packard
> via Newlib <newlib@sourceware.org>
> *Sent:* Wednesday, August 26, 2020 1:03 PM
> *To:* newlib@sourceware.org <newlib@sourceware.org>
> *Subject:* [PATCH 0/3] libm: Clean up gamma functions
>
> EXTERNAL EMAIL - This email originated from outside of CACI. Do not click
> any links or attachments unless you recognize and trust the sender.
>
>
>
>
>
> This  series  cleans   up  some  API  oddities   and  inaccuracies  in
> errno/exception reporting  from the  various gamma  related functions.
> With this series applied, newlib now  matches the POSIX spec (and glibc)
> for  tgamma/lgamma  functions. I've  left  the  'gamma' functions  BSD
> compatible (tgamma instead of lgamma) to avoid changing the API.
>
> [PATCH 1/3] libm: Fix sign value returned from __ieee754_lgamma*_r(-0)
>
>         Following the C spec, gamma(-0) should be -INFINITY rather than
>         +INFINITY, so the *signgamp value in lgamma should be -1
>         rather than +1 for this case.
>
What C spec?  Neither C99 nor C11 say so that I see for lgamma().  POSIX
makes the direct statement "If x is a non-positive integer, a pole error
shall occur and lgamma(), lgammaf(), and lgammal() shall return +HUGE_VAL,
+HUGE_VALF, and +HUGE_VALL, respectively."  0, regardless of sign, falls
under that.
https://pubs.opengroup.org/onlinepubs/9699919799/functions/lgamma.html
Additionally, the Linux lgamma man page, using GLIBC, says the same thing
as quoted from POSIX (not word for word, but the identical result).
(POSIX does require the +-0 +-INFINITY for tgamma().  Did these get crossed
up?)
Craig

>
> [PATCH 2/3] libm: Remove __ieee754_gamma_r variants
>
>         Back in 2002, a patch changed the various gamma functions so
>         that the returned abs(gamma) instead of the ln of gamma. This
>         left the _r variants in place, with their extra parameter to
>         return the sign of the resulting value. This doesn't make any
>         sense to me and looks like it was just a mistake. I've changed
>         the __ieee754_gamma functions to apply the sign value locally
>         and eliminated the gamma_r versions as those are no longer useful.
>
> [PATCH 3/3] libm: Adjust errno/exception values for gamma/lgamma
>
>         There's a pretty detailed discussion in the commit message
>         here, the short version is that tgamma and lgamma are
>         specified with different errno/exception behaviour for
>         one class of inputs (negative integers).
>