This is an automated email from the ASF dual-hosted git repository.
acassis pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/nuttx.git
The following commit(s) were added to refs/heads/master by this push:
new 06c8c8ac256 include/nuttx/lib/math.h: Give NAN with the sign bit clear.
06c8c8ac256 is described below
commit 06c8c8ac25658baa98a532e9020339a54eaf27ae
Author: Marco Casaroli <[email protected]>
AuthorDate: Sun Nov 8 14:52:48 2026 +0100
include/nuttx/lib/math.h: Give NAN with the sign bit clear.
NAN, NAN_F and NAN_L were INFINITY * 0.0F. Where NAN is used in code,
GCC does not fold it (the multiplication raises a floating point
exception), so it runs at run time and gives the default NaN of the FPU:
on x86 that NaN has the sign bit set; on arm64 and RISC-V it is clear.
C does not specify the sign of NAN, but other C libraries give a
positive one, and programs and tests compare the text: a printf() that
prints the sign of a NaN, as C allows, prints "-nan" for NAN on x86 and
"nan" elsewhere.
Add CONFIG_HAVE_BUILTIN_NAN to the built-in list of compiler.h (GCC 4
and later, and Clang; not Green Hills), and with it use
__builtin_nan(""), __builtin_nanf("") and __builtin_nanl(""): a quiet
NaN with the sign bit clear on every architecture, also a constant
expression. The types do not change. Other compilers keep the old
expression.
Assisted-by: Claude Code:claude-opus-5-5
Signed-off-by: Marco Casaroli <[email protected]>
---
include/nuttx/compiler.h | 1 +
include/nuttx/lib/math.h | 17 ++++++++++++++---
2 files changed, 15 insertions(+), 3 deletions(-)
diff --git a/include/nuttx/compiler.h b/include/nuttx/compiler.h
index ee8e9e87206..4ae07992a4b 100644
--- a/include/nuttx/compiler.h
+++ b/include/nuttx/compiler.h
@@ -199,6 +199,7 @@
# define CONFIG_HAVE_BUILTIN_FFS 1
# define CONFIG_HAVE_BUILTIN_FFSL 1
# define CONFIG_HAVE_BUILTIN_FFSLL 1
+# define CONFIG_HAVE_BUILTIN_NAN 1
# endif
/* An __asm__ label carries __USER_LABEL_PREFIX__ ("_" on RH850 and RX) */
diff --git a/include/nuttx/lib/math.h b/include/nuttx/lib/math.h
index ad4967987be..e62c7fe949a 100644
--- a/include/nuttx/lib/math.h
+++ b/include/nuttx/lib/math.h
@@ -75,17 +75,28 @@
#endif
#define INFINITY ((double)(_HUGE_ENUF * _HUGE_ENUF))
-#define NAN ((double)(INFINITY * 0.0F))
#define HUGE_VAL INFINITY
#define INFINITY_F ((float)INFINITY)
-#define NAN_F ((float)(INFINITY * 0.0F))
#define HUGE_VALF ((float)INFINITY)
#define INFINITY_L ((long double)INFINITY)
-#define NAN_L ((long double)(INFINITY * 0.0F))
#define HUGE_VALL ((long double)INFINITY)
+/* A quiet NaN with the sign bit clear. INFINITY * 0 gives the default NaN
+ * of the FPU, and on x86 that NaN has the sign bit set.
+ */
+
+#ifdef CONFIG_HAVE_BUILTIN_NAN
+# define NAN __builtin_nan("")
+# define NAN_F __builtin_nanf("")
+# define NAN_L __builtin_nanl("")
+#else
+# define NAN ((double)(INFINITY * 0.0F))
+# define NAN_F ((float)(INFINITY * 0.0F))
+# define NAN_L ((long double)(INFINITY * 0.0F))
+#endif
+
#define isnan(x) ((x) != (x))
#define isnanf(x) ((x) != (x))
#define isnanl(x) ((x) != (x))
