Fix cacos real-part inaccuracy for result real part near 0 (bug 15023).

1 files changed, 31 insertions, 0 deletions

diff --git a/ChangeLog b/ChangeLog
index e7cfb64..890e3d4 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,34 @@
+2013-01-17  Joseph Myers  <joseph@codesourcery.com>
+
+	[BZ #15023]
+	* include/complex.h: Condition contents on [!_COMPLEX_H].
+	(__kernel_casinhf): New prototype.
+	(__kernel_casinh): Likewise.
+	(__kernel_casinhl): Likewise.
+	* math/Makefile (libm_calls): Add k_casinh.
+	* math/k_casinh.c: New file.
+	* math/k_casinhf.c: Likewise.
+	* math/k_casinhl.c: Likewise.
+	* math/s_cacos.c (__cacos): Implement using __kernel_casinh for
+	finite nonzero arguments.
+	* math/s_cacosf.c (__cacosf): Implement using __kernel_casinhf for
+	finite nonzero arguments.
+	* math/s_cacosl.c (__cacosl): Implement using __kernel_casinhl for
+	finite nonzero arguments.
+	* math/s_casinh.c: Do not include <float.h>.
+	(__casinh): Move code for finite nonzero arguments to k_casinh.c.
+	* math/s_casinhf.c: Do not include <float.h>.
+	(__casinhf): Move code for finite nonzero arguments to
+	k_casinhf.c.
+	* math/s_casinhl.c: Do not include <float.h>.
+	[LDBL_MANT_DIG == 106] (LDBL_EPSILON): Do not undefine and
+	redefine.
+	(__casinhl): Move code for finite nonzero arguments to
+	k_casinhl.c.
+	* math/libm-test.inc (cacos_test): Add more tests.
+	* sysdeps/i386/fpu/libm-test-ulps: Update.
+	* sysdeps/x86_64/fpu/libm-test-ulps: Likewise.
+
 2013-01-17  Pino Toscano  <toscano.pino@tiscali.it>
 
 	* sysdeps/unix/sysv/linux/malloc-sysdep.h (HAVE_MREMAP): New define.