NetBSD Problem Report #58236

From www@netbsd.org  Wed May  8 16:14:24 2024
Return-Path: <www@netbsd.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits))
	(Client CN "mail.NetBSD.org", Issuer "mail.NetBSD.org CA" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 0AB6C1A9238
	for <gnats-bugs@gnats.NetBSD.org>; Wed,  8 May 2024 16:14:24 +0000 (UTC)
Message-Id: <20240508161422.9E7D61A923B@mollari.NetBSD.org>
Date: Wed,  8 May 2024 16:14:22 +0000 (UTC)
From: campbell+netbsd@mumble.net
Reply-To: campbell+netbsd@mumble.net
To: gnats-bugs@NetBSD.org
Subject: nexttoward(3) is broken on subnormals
X-Send-Pr-Version: www-1.0

>Number:         58236
>Category:       lib
>Synopsis:       nexttoward(3) is broken on subnormals
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    lib-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed May 08 16:15:01 +0000 2024
>Last-Modified:  Sat May 11 02:10:01 +0000 2024
>Originator:     Taylor R Campbell
>Release:        current, 10, 9, ...
>Organization:
The nexttowarbsd(3) Foundation
>Environment:
>Description:
nexttoward(k*n, INFINITY) should return (k + 1)*n, where n is the smallest positive subnormal and k is any integer below 2^p.  But instead it returns (k - 1)*n.

Similarly, nexttoward(k*n, -INFINITY) should return (k - 1)*n, but it returns (k + 1)*n.
>How-To-Repeat:
$ cat test.c
#include <math.h>
#include <stdio.h>

int
main(void)
{
	volatile double n = __DBL_DENORM_MIN__;
	volatile double d = INFINITY;
	printf("123*n goes up to %.0f*n\n", nextafter(123*n, d)/n);
	printf("123*n goes down to %.0f*n\n", nextafter(123*n, -d)/n);
	fflush(stdout);
	return ferror(stdout);
}
$ make test LDLIBS=-lm
cc -O2   -o test test.c -lm
$ ./test
123*n goes up to 124*n
123*n goes down to 122*n

>Fix:
Yes, please!

>Audit-Trail:
From: "Taylor R Campbell" <riastradh@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/58236 CVS commit: src/tests/lib/libm
Date: Wed, 8 May 2024 17:27:04 +0000

 Module Name:	src
 Committed By:	riastradh
 Date:		Wed May  8 17:27:04 UTC 2024

 Modified Files:
 	src/tests/lib/libm: t_next.c

 Log Message:
 tests/lib/libm/t_next: Expand substantially.

 This covers many more potential problem areas -- and includes a new
 xfail test for PR lib/58236: nexttoward(3) is broken on subnormals.


 To generate a diff of this commit:
 cvs rdiff -u -r1.3 -r1.4 src/tests/lib/libm/t_next.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: "Taylor R Campbell" <riastradh@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/58236 CVS commit: src
Date: Sat, 11 May 2024 02:07:54 +0000

 Module Name:	src
 Committed By:	riastradh
 Date:		Sat May 11 02:07:54 UTC 2024

 Modified Files:
 	src/lib/libm/src: s_nexttoward.c
 	src/tests/lib/libm: t_next.c

 Log Message:
 nexttoward(3): Fix high-word test on small positive subnormals.

 By this point in the logic, x can't be zero, so it's either positive
 or negative.

 The high word hx, however, can be zero, when x is a small positive
 subnormal.  This means x is a small positive subnormal, so if x > y
 we are computing nextDown, and if x < y we are computing nextUp.

 hx is a (signed 32-bit) integer, not a double floating-point number,
 so it's a little silly to compare hx > 0.0.  But that on its own
 isn't enough to trigger the bug because all signed 32-bit integers
 can be represented by double on all NetBSD architectures.

 PR lib/58236


 To generate a diff of this commit:
 cvs rdiff -u -r1.2 -r1.3 src/lib/libm/src/s_nexttoward.c
 cvs rdiff -u -r1.5 -r1.6 src/tests/lib/libm/t_next.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

Home	PR Database Search