avfilter/asubprocess: use wav header instead of headerless stream data

Signed-off-by: Stefan Westerfeld <stefan@space.twc.de>

libavfilter/af_asubprocess.c

 #include "libavutil/opt.h"
+#include "libavutil/mem.h"
+#include "libavformat/avio.h"
+#include "libavformat/avio_internal.h"
 
 #include "audio.h"
 #include "avfilter.h"
@@ -186,12 +189,22 @@ sp_write (SP *sp, char *buffer, int count)
 }
 // =============================================
 
+enum {
+    STATE_EXPECT_RIFF = 0,
+    STATE_EXPECT_CHUNK = 1,
+    STATE_SKIP_CHUNK = 2,
+    STATE_IN_DATA_CHUNK = 3,
+    STATE_ERROR = 4
+};
+
 typedef struct AndioWMarkContext {
     const AVClass *class;
 
     double strength;
 
     int    fd;
+    int    state;
+    unsigned int    state_skip;
     SP    *sp;
     int    nb_samples;
     int    eof;
@@ -223,7 +236,32 @@ static int config_input(AVFilterLink *inlink)
     AVFilterContext *ctx = inlink->dst;
     AudioWMarkContext *s = ctx->priv;
     // TODO: use inlink->sample_rate, inlink->ch_layout.nb_channels
-    s->sp = sp_new ("audiowmark -q add --format raw --raw-rate 44100 - - f0");
+    s->sp = sp_new ("audiowmark -q add - - f0");
+    int bit_depth = 16;
+    int sample_rate = inlink->sample_rate;
+    int nb_channels = inlink->ch_layout.nb_channels;
+    AVIOContext *wav_header = NULL;
+    uint8_t *buffer = NULL;
+    avio_open_dyn_buf (&wav_header);
+    ffio_wfourcc(wav_header, "RIFF");
+    avio_wl32(wav_header, -1);
+    ffio_wfourcc(wav_header, "WAVE");
+    // subchunk 1
+    ffio_wfourcc(wav_header, "fmt ");
+    avio_wl32 (wav_header, 16); // subchunk size
+    avio_wl16 (wav_header, 1);  // uncompressed audio
+    avio_wl16 (wav_header, nb_channels);
+    avio_wl32 (wav_header, sample_rate);
+    avio_wl32 (wav_header, sample_rate * nb_channels * bit_depth / 8); // byte rate
+    avio_wl16 (wav_header, nb_channels * bit_depth / 8); // block align
+    avio_wl16 (wav_header, bit_depth); // bits per sample
+
+    // subchunk 2
+    ffio_wfourcc (wav_header, "data");
+    avio_wl32 (wav_header, -1);
+    int sz = avio_close_dyn_buf (wav_header, &buffer);
+    sp_write (s->sp, buffer, sz);
+    av_free (buffer);
 #if 0
     int opts = s->transients|s->detector|s->phase|s->window|
                s->smoothing|s->formant|s->opitch|s->channels|
@@ -239,6 +277,7 @@ static int config_input(AVFilterLink *inlink)
     s->first_pts = AV_NOPTS_VALUE;
 #endif
     s->nb_samples = 1024;
+    s->state = STATE_EXPECT_RIFF;
 
     return 0;
 }
@@ -251,49 +290,82 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
     AVFrame *out;
     int ret = 0, nb_samples;
 
+    short buffer[100000];
+    int k = 0;
     for (int i = 0; i < in->nb_samples; i++)
       {
         for (int j = 0; j < inlink->ch_layout.nb_channels; j++)
           {
             float **xd = (float **)in->extended_data;
-            short ss = xd[j][i]*32000;
-            sp_write (s->sp, &ss, 2);
+            buffer[k++] = xd[j][i]*32000;
           }
       }
+    sp_write (s->sp, buffer, 2 * k);
     if (s->eof)
     {
-      while (!sp_write (s->sp, 0, 0))
-        ;
-      printf ("eof\n");
+        while (!sp_write (s->sp, 0, 0))
+            ;
+        printf ("eof\n");
     }
-    bool gen_output = true;
-    while (gen_output)
-    {
-    int avail = sp_can_read (s->sp) / 2 / inlink->ch_layout.nb_channels;
-    if (avail > 2048)
-      avail = 2048;
-    if (avail > 0)
+    if (s->state == STATE_EXPECT_RIFF)
+        if (sp_can_read (s->sp) >= 12)
+        {
+            sp_read (s->sp, buffer, 12);
+            s->state = STATE_EXPECT_CHUNK;
+        }
+    if (s->state == STATE_EXPECT_CHUNK)
+        if (sp_can_read (s->sp) >= 8)
+        {
+            unsigned char x[8];
+            sp_read (s->sp, x, 8);
+            if (x[0] == 'd' && x[1] == 'a' && x[2] == 't' && x[3] == 'a')
+                s->state = STATE_IN_DATA_CHUNK;
+            else
+            {
+                s->state = STATE_SKIP_CHUNK;
+                s->state_skip = x[4] + (x[5] << 8) + (x[6] << 16) + (x[7] << 24);
+            }
+        }
+    if (s->state == STATE_SKIP_CHUNK && sp_can_read (s->sp) >= s->state_skip)
     {
-        out = ff_get_audio_buffer(outlink, avail);
-        if (!out) {
-            av_frame_free(&in);
-            return AVERROR(ENOMEM);
+        while (s->state_skip)
+        {
+            char buffer[1024];
+            unsigned int n = FFMIN (sizeof (buffer), s->state_skip);
+            sp_read (s->sp, buffer, n);
+            s->state_skip -= n;
         }
-        short buffer[100000]; // FIXME avail * inlink->ch_layout.nb_channels];
-        sp_read (s->sp, buffer, avail * 2 * inlink->ch_layout.nb_channels);
-        int k = 0;
-        for (int i = 0; i < avail; i++)
-          {
-            for (int j = 0; j < inlink->ch_layout.nb_channels; j++)
-              {
-                float **xd = (float **)out->extended_data;
-                xd[j][i] = buffer[k++] * (1/32000.);
-              }
-          }
-        ret = ff_filter_frame(outlink, out);
+        s->state = STATE_EXPECT_CHUNK;
     }
-    else
-      gen_output = false;
+    bool gen_output = s->state == STATE_IN_DATA_CHUNK;
+    while (gen_output)
+    {
+        int avail = sp_can_read (s->sp) / 2 / inlink->ch_layout.nb_channels;
+        if (avail > 2048)
+            avail = 2048;
+        if (avail > 0)
+        {
+            out = ff_get_audio_buffer(outlink, avail);
+            if (!out) {
+                av_frame_free(&in);
+                return AVERROR(ENOMEM);
+            }
+            short buffer[100000]; // FIXME avail * inlink->ch_layout.nb_channels];
+            sp_read (s->sp, buffer, avail * 2 * inlink->ch_layout.nb_channels);
+            int k = 0;
+            for (int i = 0; i < avail; i++)
+            {
+                for (int j = 0; j < inlink->ch_layout.nb_channels; j++)
+                {
+                    float **xd = (float **)out->extended_data;
+                    xd[j][i] = buffer[k++] * (1/32000.);
+                }
+            }
+            ret = ff_filter_frame(outlink, out);
+        }
+        else {
+            gen_output = false;
+        }
     }
 #if 0
     if (s->first_pts == AV_NOPTS_VALUE)