Update.

1997-04-02 16:28  Ulrich Drepper  <drepper@cygnus.com>

	* grp/fgetgrent.c: Don't use fixed buffer length.  Allow dynamic
	adjustment if the underlying function say it is too small.
	* pwd/fgetpwent.c: Likewise.
	* shadow/fgetspent.c: Likewise.
	* shadow/sgetspent.c: Likewise.

	* grp/fgetgrent_r.c (__fgetgrent_r): If `parse_result' returns -1
	this means there was not enough room for the gr_mem array.

	* nss/nss_files/files-XXX.c: Leave loop and return error if
	`parse_result' returns -1.

	* nss/nss_files/files-parse.c (parse_line): Return -1 if parsing
	of trailing list fails because of low memory.
	(parse_list): Rewrite.  The logic was completly incorrect since
	the checks for overflowing the array were not executed when
	needed.

	* shadow/sgetspent_r.c: Only a return value > 0 from parse_line
	signals success.

	* nss/getXXbyYY.c: Make sure NULL is returned if we ran out of memory.
	* nss/getXXent.c: Likewise.

	* stdlib/strtod.c (SET_MANTISSA): Make sure that at least one bit
	is set in mantissa.
	* stdlib/strtof.c: Likewise.
	* stdlib/strtold.c: Likewise.
	* wcsmbs/wcstof.c: Likewise.
	* wcsmbs/wcstold.c: Likewise.

	* sysdeps/libm-ieee754/s_nan.c: Use strtod is parameter is not empty
	string.
	* sysdeps/libm-ieee754/s_nanf.c: Likewise.
	* sysdeps/libm-ieee754/s_nanld.c: Likewise.

1997-04-02 13:56  Ulrich Drepper  <drepper@cygnus.com>

	* manual/errno.texi: Add description of strerror_r.

1997-04-01 19:07  H.J. Lu (hjl@gnu.ai.mit.edu)

	* manual/errno.texi: Add ENOMEDIUM and EMEDIUMTYPE.

1997-04-02 13:55  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/gnu/errlist.awk: The Texinfo file is named errno.texi.

1997-03-30 10:59  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/m68k/fpu/s_cexp.c: New file.
	* sysdeps/m68k/fpu/s_cexpf.c: New file.
	* sysdeps/m68k/fpu/s_cexpl.c: New file.
	* sysdeps/m68k/fpu/s_nearbyint.c: New file.
	* sysdeps/m68k/fpu/s_nearbyintf.c: New file.
	* sysdeps/m68k/fpu/s_nearbyintl.c: New file.
	* sysdeps/m68k/fpu/s_trunc.c: New file.
	* sysdeps/m68k/fpu/s_truncf.c: New file.
	* sysdeps/m68k/fpu/s_truncl.c: New file.

	* sysdeps/m68k/fpu/s_modf.c: Rewritten.
	* sysdeps/m68k/fpu/s_modff.c: Adapted.
	* sysdeps/m68k/fpu/s_modfl.c: Likewise.

	* sysdeps/m68k/fpu/e_atan2.c: Use __signbit inline.
	* sysdeps/m68k/fpu/e_pow.c: Likewise.  Write 0.0/0.0 for NaN.  Fix
	typo in test for `y = 0.5'.

	* sysdeps/m68k/fpu/__math.h: Define __trunc, __signbit,
	__nearbyint.  Remove __modf.

	* sysdeps/m68k/Makefile (CFLAGS-s_copysignl.c) [$(subdir)=math]:
	Define for gcc bug workaround.

	* math/libm-test.c (log2_test): Add epsilon for long double to
	"log2 (e)" and "log2 (16)".
	(modf_test): Fix test name.
	(sqrt_test): Add missing FUNC.

1997-03-29 16:24  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sunrpc/Makefile (CFLAGS-xrquota.c): Fix typo in variable name.

1997-03-28 20:56  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* stdlib/stdlib.h: Declare __clearenv.

1997-03-28 18:03  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/libm-ieee754/s_asinhl.c: Fix Ulrich Dreppers mangling of
	the last change.

1997-03-28 17:55  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* configure.in: Call AC_CANONICAL_BUILD before referencing
	$build.  Fix test for BUILD_CC.

1997-04-02 01:29  Philip Blundell  <pjb27@cam.ac.uk>

	* sysdeps/libm-ieee754/s_log2f.c (__log2f): Fix typo.

1997-03-31 18:53  Philip Blundell  <pjb27@cam.ac.uk>

	* sysdeps/unix/sysv/linux/net/route.h: Add new IPv6 definitions.

1997-04-01 14:08  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/i386/memset.c: Fix typo.
	* sysdeps/stub/fchmod.c: Fix typo.
	* sysdeps/stub/fchdir.c: Fix typo.
	Reported by Fila Kolodny <fila@ibi.com>.

1 files changed, 12 insertions, 18 deletions

diff --git a/sysdeps/libm-ieee754/s_nan.c b/sysdeps/libm-ieee754/s_nan.c
index 0d065d5c72..35d92646c0 100644
--- a/sysdeps/libm-ieee754/s_nan.c
+++ b/sysdeps/libm-ieee754/s_nan.c
@@ -18,31 +18,25 @@
    write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
    Boston, MA 02111-1307, USA.  */
 
-/* XXX The ISO C 9X standard mentions a `n-char-sequence' which is provided
-   as the argument to this function but I have no clue what this means.
-   Perhaps it is a description of the bits set in the mantissa.  */
 #include <math.h>
-#ifdef HANDLE_TAGP
-# include <stdlib.h>
-# include <string.h>
-#else
-# include <ieee754.h>
-#endif
+#include <stdlib.h>
+#include <string.h>
+#include <ieee754.h>
 
 
 double
 __nan (const char *tagp)
 {
-#ifdef HANDLE_TAGP
-  /* If we ever should have use of the TAGP parameter we will use the
-     strtod function to analyze it.  */
-  char buf[6 + strlen (tagp)];
-  sprintf (buf, "NAN(%s)", tagp);
-  return strtod (buf, NULL);
-#else
   static const union ieee754_double nan_value =
-  { ieee: { mantissa1: 0x1, mantissa0: 0x0, exponent: 0x7ff, negative: 0 } };
+  { ieee: { mantissa1: 0, mantissa0: 0x80000, exponent: 0x7ff, negative: 0 } };
+
+  if (tagp[0] != '\0')
+    {
+      char buf[6 + strlen (tagp)];
+      sprintf (buf, "NAN(%s)", tagp);
+      return strtod (buf, NULL);
+    }
+
   return nan_value.d;
-#endif
 }
 weak_alias (__nan, nan)