stream_cdda, demux_raw: always use s16le

stream_cdda's output format is linked to demux_raw's default audio
format, and at least we don't care enough to provide a separate
mechanism to let stream_cdda explicitly set the format, so they must
match.

Judging from the existing code, it looks like CDDA always outputs little
endian. stream_cdda.c changed this back to native endian (what demux_raw
expects). Just make them both little endian. This requires less code,
and also having a raw demuxer's behavior depend on the endianness of the
machine isn't very sane anyway.

demux/demux_raw.c

@@ -78,7 +78,7 @@ const struct m_sub_options demux_rawaudio_conf = {
         // Note that currently, stream_cdda expects exactly these parameters!
         .channels = MP_CHMAP_INIT_STEREO,
         .samplerate = 44100,
-        .aformat = PCM(1, 0, 16, NE), // s16
+        .aformat = PCM(1, 0, 16, 0), // s16le
     },
 };
 

stream/stream_cdda.c

@@ -38,15 +38,12 @@
 #include <stdlib.h>
 #include <stdbool.h>
 
-#include "osdep/endian.h"
-
 #include "talloc.h"
 
 #include "stream.h"
 #include "options/m_option.h"
 #include "options/m_config.h"
 #include "options/options.h"
-#include "osdep/mpbswap.h"
 
 #include "common/msg.h"
 
@@ -176,11 +173,6 @@ static int fill_buffer(stream_t *s, char *buffer, int max_len)
     if (!buf)
         return 0;
 
-#if BYTE_ORDER == BIG_ENDIAN
-    for (i = 0; i < CDIO_CD_FRAMESIZE_RAW / 2; i++)
-        buf[i] = le2me_16(buf[i]);
-#endif
-
     p->sector++;
     memcpy(buffer, buf, CDIO_CD_FRAMESIZE_RAW);