Use copysign when implementing signbit, if the libc has it already.

1 files changed, 8 insertions, 0 deletions

diff --git a/lib/signbitf.c b/lib/signbitf.c
index c172b5fbef..d7e0663c7b 100644
--- a/lib/signbitf.c
+++ b/lib/signbitf.c
@@ -30,11 +30,19 @@ int
 gl_signbitf (float arg)
 {
 #if defined FLT_SIGNBIT_WORD && defined FLT_SIGNBIT_BIT
+  /* The use of a union to extract the bits of the representation of a
+     'long double' is safe in practice, despite of the "aliasing rules" of
+     C99, because the GCC docs say
+       "Even with '-fstrict-aliasing', type-punning is allowed, provided the
+        memory is accessed through the union type."
+     and similarly for other compilers.  */
 # define NWORDS \
     ((sizeof (float) + sizeof (unsigned int) - 1) / sizeof (unsigned int))
   union { float value; unsigned int word[NWORDS]; } m;
   m.value = arg;
   return (m.word[FLT_SIGNBIT_WORD] >> FLT_SIGNBIT_BIT) & 1;
+#elif HAVE_COPYSIGNF_IN_LIBC
+  return copysignf (1.0f, arg) < 0;
 #else
   /* This does not do the right thing for NaN, but this is irrelevant for
      most use cases.  */