Re: issue with tgammaf

["Keith Packard" <keithp@keithp.com>]

[-- Attachment #1.1: Type: text/plain, Size: 436 bytes --]

Paul Zimmermann <Paul.Zimmermann@inria.fr> writes:

>        Hi,
>
> in https://sourceware.org/pipermail/newlib/2020/017920.html I reported an
> issue with tgammaf(-0).
>
> With Newlib 3.3.0, there seems to be the same issue for negative subnormal
> values of the input, where tgammaf gives +inf instead of -inf.

We've been fixing a bunch of gamma problems; this is one that got missed
when the library is configured to report errno.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: 0001-libm-Make-tgamma-small-INFINITY.patch --]
[-- Type: text/x-diff, Size: 876 bytes --]

From 8cefdba40a4382de6b3f44a5d35880026727f1dd Mon Sep 17 00:00:00 2001
From: Keith Packard <keithp@keithp.com>
Date: Tue, 15 Sep 2020 10:10:07 -0700
Subject: [PATCH] libm: Make tgamma(-small) = -INFINITY

Need to copy the argument sign to the output for tgamma(finite)
overflow case.

Signed-off-by: Keith Packard <keithp@keithp.com>
---
 newlib/libm/math/k_standard.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/newlib/libm/math/k_standard.c b/newlib/libm/math/k_standard.c
index 906412ba7..beab51e16 100644
--- a/newlib/libm/math/k_standard.c
+++ b/newlib/libm/math/k_standard.c
@@ -331,7 +331,7 @@ static double zero = 0.0;	/* used as const */
 	    case 40:
 	    case 140:
 		/* gamma(finite) overflow */
-		retval = HUGE_VAL;
+		retval = copysign(HUGE_VAL, x);
 		errno = ERANGE;
 		break;
 	    case 41:
-- 
2.28.0


[-- Attachment #1.3: Type: text/plain, Size: 15 bytes --]


-- 
-keith

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]