| commit | 122971b8cc0ab46f0ab85927f109f89606800c3c | [log] [tgz] |
|---|---|---|
| author | Jean-Marc Valin <jmvalin@jmvalin.ca> | Tue Dec 10 13:55:35 2013 -0500 |
| committer | Jean-Marc Valin <jmvalin@jmvalin.ca> | Tue Dec 10 13:56:38 2013 -0500 |
| tree | 31da79aaef0cd38bf2751a80c4ad2bf75ce01f5b | |
| parent | fa85e3bf2823da90a031cbbb051022f346589da4 [diff] [blame] |
More NaN hardening in the analysis code
diff --git a/celt/arch.h b/celt/arch.h index 6693408..25cc0b7 100644 --- a/celt/arch.h +++ b/celt/arch.h
@@ -140,7 +140,7 @@ #ifdef FLOAT_APPROX /* This code should reliably detect NaN/inf even when -ffast-math is used. Assumes IEEE 754 format. */ -static inline int celt_isnan(float x) +static OPUS_INLINE int celt_isnan(float x) { union {float f; opus_uint32 i;} in; in.f = x;