shithub: sox

Info  •  Files  •  Log  •  Branches

make SOX_FLOAT_32BIT_TO_SAMPLE alias of SOX_FLOAT_64BIT_TO_SAMPLE

Conversions from floating-point formats should use the same formula.
The C language conversion from float to integer truncates the fractional
part toward zero.  To get the nearest integer, an adjustment by 0.5 away
from zero is thus needed.  The "64-bit" macro does this while the
"32-bit" one does not.  Fix this discrepancy by making the latter an
alias of the former.

--- a/src/sox.h

+++ b/src/sox.h

@@ -917,18 +917,7 @@

 @param clips Variable to increment if the input sample is too large or too small.

 @returns SoX native sample value.

 */

-#define SOX_FLOAT_32BIT_TO_SAMPLE(d,clips)                      \

-  (sox_sample_t)(                                               \

-    LSX_USE_VAR(sox_macro_temp_sample),                         \

-    sox_macro_temp_double = (d) * (SOX_SAMPLE_MAX + 1.0),       \

-    sox_macro_temp_double < SOX_SAMPLE_MIN ?                    \

-      ++(clips), SOX_SAMPLE_MIN :                               \

-      sox_macro_temp_double >= SOX_SAMPLE_MAX + 1.0 ?           \

-        sox_macro_temp_double > SOX_SAMPLE_MAX + 1.0 ?          \

-          ++(clips), SOX_SAMPLE_MAX :                           \

-          SOX_SAMPLE_MAX :                                      \

-        sox_macro_temp_double                                   \

-  )

+#define SOX_FLOAT_32BIT_TO_SAMPLE(d,clips) SOX_FLOAT_64BIT_TO_SAMPLE(d, clips)

 /**

 Client API: