transcode: initialize audio encoder after first packet arrives

Similar approach as video encoding does, helps for example with faad
module that in some cases only gets proper audio properties on decoding.

Ref #9022,#6298

modules/stream_out/transcode/audio.c

@@ -202,6 +202,34 @@ int transcode_audio_process( sout_stream_t *p_stream,
     while( (p_audio_buf = id->p_decoder->pf_decode_audio( id->p_decoder,
                                                           &in )) )
     {
+
+        if( unlikely( !id->p_encoder->p_module ) )
+        {
+            /* Complete destination format */
+            id->p_encoder->fmt_out.i_codec = p_sys->i_acodec;
+            id->p_encoder->fmt_out.audio.i_rate = p_sys->i_sample_rate > 0 ?
+                p_sys->i_sample_rate : id->p_decoder->fmt_out.audio.i_rate;
+            id->p_encoder->fmt_out.i_bitrate = p_sys->i_abitrate;
+            id->p_encoder->fmt_out.audio.i_bitspersample =
+                id->p_decoder->fmt_out.audio.i_bitspersample;
+            id->p_encoder->fmt_out.audio.i_channels = p_sys->i_channels > 0 ?
+                p_sys->i_channels : id->p_decoder->fmt_out.audio.i_channels;
+            /* Sanity check for audio channels */
+            id->p_encoder->fmt_out.audio.i_channels =
+                __MIN( id->p_encoder->fmt_out.audio.i_channels,
+                       id->p_decoder->fmt_out.audio.i_channels );
+            id->p_encoder->fmt_out.audio.i_original_channels =
+                id->p_decoder->fmt_in.audio.i_physical_channels;
+            id->p_encoder->fmt_out.audio.i_physical_channels =
+                pi_channels_maps[id->p_encoder->fmt_out.audio.i_channels];
+            if( transcode_audio_new( p_stream, id ) )
+            {
+                msg_Err( p_stream, "cannot create audio chain" );
+                return VLC_EGENERIC;
+            }
+            date_Init( &id->interpolated_pts, id->p_decoder->fmt_out.audio.i_rate, 1 );
+
+        }
         /* Check if audio format has changed, and filters need reinit */
         if( unlikely( ( id->p_decoder->fmt_out.audio.i_rate != p_sys->fmt_audio.i_rate ) ||
                       ( id->p_decoder->fmt_out.audio.i_physical_channels != p_sys->fmt_audio.i_physical_channels ) ) )
@@ -308,6 +336,15 @@ bool transcode_audio_add( sout_stream_t *p_stream, es_format_t *p_fmt,
         return false;
     }
 
+    module_unneed( id->p_encoder, id->p_encoder->p_module );
+    if( id->p_encoder->fmt_out.p_extra )
+    {
+        free( id->p_encoder->fmt_out.p_extra );
+        id->p_encoder->fmt_out.p_extra = NULL;
+        id->p_encoder->fmt_out.i_extra = 0;
+    }
+    id->p_encoder->p_module = NULL;
+
     date_Init( &id->interpolated_pts, p_fmt->audio.i_rate, 1 );
 
     return true;