[FFmpeg-cvslog] r22937 - in trunk: libavcodec/amrnbdec.c libavcodec/atrac1.c libavcodec/audioconvert.c libavcodec/qcelpdata.h libavcodec/qcelpdec.c libavcodec/ra288.c libavcodec/sipr.c libavcodec/sipr16k.c libavco...

[Andreas Öman]

rbultje wrote:
> Author: rbultje
> Date: Wed Apr 21 19:57:48 2010
> New Revision: 22937
> 
> Log:
> Move clipping of audio samples (for those codecs outputting float) from decoder
> to the audio conversion routines.
> 
> Modified: trunk/libavutil/common.h
> ==============================================================================
> --- trunk/libavutil/common.h	Wed Apr 21 19:51:37 2010	(r22936)
> +++ trunk/libavutil/common.h	Wed Apr 21 19:57:48 2010	(r22937)
> @@ -145,6 +145,17 @@ static inline av_const int16_t av_clip_i
>  }
>  
>  /**
> + * Clips a signed 64-bit integer value into the -2147483648,2147483647 range.
> + * @param a value to clip
> + * @return clipped value
> + */
> +static inline av_const int32_t av_clipl_int32(int64_t a)
> +{
> +    if ((a+2147483648) & ~2147483647) return (a>>63) ^ 2147483647;
> +    else                              return a;
> +}
> +
> +/**

This trips a warning if not compiled with c99 (for example an
application that happens to include libavutil/common.h). Not only
annoying but for applications compiling with -Werror this results in
build fail.

common.h:154: error: this decimal constant is unsigned only in ISO C90

I'm not sure but do we require that apps pulling include files from
FFmpeg to be compiled in c99 mode? I think not, but I don't really know.

I guess this could be easily fixed by adding the correct LL or ULL
postfixes to the constants.

Opinions?