replaygain: correctly parse peak values

According to the ReplayGain spec, the peak amplitude may overflow and may result
in peak amplitude values greater than 1.0 with psychoacoustically coded audio,
such as MP3. Fully compliant decoders must allow peak overflows.

Additionally, having peak values in the 0<->UINT32_MAX scale makes it more
difficult for applications to actually use the peak values (e.g. when
implementing clipping prevention) since values have to be rescaled down.

This patch corrects the peak parsing by removing the rescaling of the decoded
values between 0 and UINT32_MAX and the 1.0 upper limit.

Signed-off-by: Anton Khirnov <anton@khirnov.net>

diff --git a/doc/APIchanges b/doc/APIchanges
index 992a3b06e8df3e56dc02511fb9c0e699237912df..26271c86da995d52c0a2a4a89db9e93ecc7b300d 100644 (file)
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -13,6 +13,10 @@ libavutil:     2013-12-xx
 
 API changes, most recent first:
 
+2014-04-xx - xxxxxxx - lavu 53.10.0 - replaygain.h
+  Full scale for peak values is now 100000 (instead of UINT32_MAX) and values
+  may overflow.
+
 2014-04-xx - xxxxxxx - lavu 53.09.0 - log.h
   Add AV_LOG(c) macro to have 256 color debug messages.
 

diff --git a/libavformat/replaygain.c b/libavformat/replaygain.c
index 6983601491dc4c67e50f130212d8da8a98261529..c123a6398539803121a0493b248951820f928a8e 100644 (file)
--- a/libavformat/replaygain.c
+++ b/libavformat/replaygain.c
@@ -35,7 +35,7 @@
 #include "avformat.h"
 #include "replaygain.h"
 
-static int32_t parse_gain(const char *gain)
+static int32_t parse_value(const char *value, int32_t min)
 {
     char *fraction;
     int  scale = 10000;
@@ -43,15 +43,15 @@ static int32_t parse_gain(const char *gain)
     int sign   = 1;
     int db;
 
-    if (!gain)
-        return INT32_MIN;
+    if (!value)
+        return min;
 
-    gain += strspn(gain, " \t");
+    value += strspn(value, " \t");
 
-    if (*gain == '-')
+    if (*value == '-')
         sign = -1;
 
-    db = strtol(gain, &fraction, 0);
+    db = strtol(value, &fraction, 0);
     if (*fraction++ == '.') {
         while (av_isdigit(*fraction) && scale) {
             mb += scale * (*fraction - '0');
@@ -61,43 +61,11 @@ static int32_t parse_gain(const char *gain)
     }
 
     if (abs(db) > (INT32_MAX - mb) / 100000)
-        return INT32_MIN;
+        return min;
 
     return db * 100000 + sign * mb;
 }
 
-static uint32_t parse_peak(const uint8_t *peak)
-{
-    int64_t val = 0;
-    int64_t scale = 1;
-
-    if (!peak)
-        return 0;
-
-    peak += strspn(peak, " \t");
-
-    if (peak[0] == '1' && peak[1] == '.')
-        return UINT32_MAX;
-    else if (!(peak[0] == '0' && peak[1] == '.'))
-        return 0;
-
-    peak += 2;
-
-    while (av_isdigit(*peak)) {
-        int digit = *peak - '0';
-
-        if (scale > INT64_MAX / 10)
-            break;
-
-        val    = 10 * val + digit;
-        scale *= 10;
-
-        peak++;
-    }
-
-    return av_rescale(val, UINT32_MAX, scale);
-}
-
 static int replaygain_export(AVStream *st,
                              const uint8_t *track_gain, const uint8_t *track_peak,
                              const uint8_t *album_gain, const uint8_t *album_peak)
@@ -108,10 +76,10 @@ static int replaygain_export(AVStream *st,
     int32_t tg, ag;
     uint32_t tp, ap;
 
-    tg = parse_gain(track_gain);
-    ag = parse_gain(album_gain);
-    tp = parse_peak(track_peak);
-    ap = parse_peak(album_peak);
+    tg = parse_value(track_gain, INT32_MIN);
+    ag = parse_value(album_gain, INT32_MIN);
+    tp = parse_value(track_peak, 0);
+    ap = parse_value(album_peak, 0);
 
     if (tg == INT32_MIN && ag == INT32_MIN)
         return 0;

diff --git a/libavutil/replaygain.h b/libavutil/replaygain.h
index 8447703fbf8e56ca45d7ae776a09d76b538a0c91..7871d2c762e747f8958aeb888768445b96226b0b 100644 (file)
--- a/libavutil/replaygain.h
+++ b/libavutil/replaygain.h
@@ -34,8 +34,8 @@ typedef struct AVReplayGain {
      */
     int32_t track_gain;
     /**
-     * Peak track amplitude, with UINT32_MAX representing full scale. 0 when
-     * unknown.
+     * Peak track amplitude, with 100000 representing full scale (but values
+     * may overflow). 0 when unknown.
      */
     uint32_t track_peak;
     /**

diff --git a/libavutil/version.h b/libavutil/version.h
index f8bb4479e22d1c6add5132146ca89e54e5538923..e920653e8b963f190107ceed9f8c9debdc4c6bd9 100644 (file)
--- a/libavutil/version.h
+++ b/libavutil/version.h
@@ -54,7 +54,7 @@
  */
 
 #define LIBAVUTIL_VERSION_MAJOR 53
-#define LIBAVUTIL_VERSION_MINOR  9
+#define LIBAVUTIL_VERSION_MINOR 10
 #define LIBAVUTIL_VERSION_MICRO  0
 
 #define LIBAVUTIL_VERSION_INT   AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \