Don't manually -DFLT_EVAL_METHOD.
FLT_EVAL_METHOD comes from <float.h> and is how the compiler communicates with us, not the other way around.
Also, the FreeBSD source #includes <float.h> in all the files that actually need the STRICT_ASSIGN() macro that uses FLT_EVAL_METHOD, so we don't need the forced include now (if we ever did).
Change-Id: I12d2867c6ca382f2849c8fa0adb5078f95792fbf
diff --git a/libm/Android.bp b/libm/Android.bp
index fa43a16..6a51df7 100644
--- a/libm/Android.bp
+++ b/libm/Android.bp
@@ -389,7 +389,6 @@
cflags: [
"-D_BSD_SOURCE",
- "-DFLT_EVAL_METHOD=0",
"-include freebsd-compat.h",
"-include fenv-access.h",
"-fno-builtin",
diff --git a/libm/freebsd-compat.h b/libm/freebsd-compat.h
index a3c7cd4..f1985d2 100644
--- a/libm/freebsd-compat.h
+++ b/libm/freebsd-compat.h
@@ -18,7 +18,6 @@
// Some FreeBSD source includes <complex.h> and assumes <math.h> from that.
#include <math.h>
-#include <float.h>
#define __weak_reference(sym,alias) \
__asm__(".weak " #alias); \