shithub: sox

Info  •  Files  •  Log  •  Branches

fix to not read cooledit extra data as audio.

--- a/Changelog

+++ b/Changelog

@@ -17,6 +17,9 @@

     wasn't being flushed.

   o Guentcho Skordev pointed out ogg vorbis files were using

     the same value for serial numbers each time.

+  o Changed sox to only read the exact size of a WAV data chunk

+    if cooledit IFF LIST chunk is found at the end of the file.

+    Normally, this isn't done to allow reading > 2gig WAV files.

 sox-12.17.6

 -----------

--- a/src/wav.c

+++ b/src/wav.c

@@ -87,6 +87,7 @@

     unsigned short samplesPerBlock;

     unsigned short blockAlign;

     st_size_t dataStart;  /* need to for seeking */

+    int found_cooledit_list;

     /* following used by *ADPCM wav files */

     unsigned short nCoefs;          /* ADPCM: number of coef sets */

@@ -909,6 +910,7 @@

     /* Horrible way to find Cool Edit marker points. Taken from Quake source*/

     ft->loops[0].start = -1;

+    wav->found_cooledit_list = 0;

     if(ft->seekable){

         /*Got this from the quake source.  I think it 32bit aligns the chunks 

          * doubt any machine writing Cool Edit Chunks writes them at an odd 

@@ -916,6 +918,7 @@

         len = (len + 1) & ~1;

         st_seek(ft, len, SEEK_CUR);

         if( findChunk(ft, "LIST") != ST_EOF){

+            wav->found_cooledit_list = 1;

             ft->comment = (char*)malloc(256);

             /* Initialize comment to a NULL string */

             ft->comment[0] = 0;

@@ -1024,7 +1027,8 @@

              * wav handler.  Sometimes it accounts for stereo,

              * sometimes it does not.

              */

-            if (len > (wav->numSamples*ft->info.channels)) 

+            /* See reason for cooledit check in comments below */

+            if (wav->found_cooledit_list && len > (wav->numSamples*ft->info.channels)) 

                 len = (wav->numSamples*ft->info.channels);

             done = 0;

@@ -1073,7 +1077,8 @@

 #ifdef ENABLE_GSM

         case ST_ENCODING_GSM:

-            if (len > wav->numSamples) 

+            /* See reason for cooledit check in comments below */

+            if (wav->found_cooledit_list && len > wav->numSamples) 

                 len = wav->numSamples;

             done = wavgsmread(ft, buf, len);

@@ -1082,10 +1087,18 @@

         break;

 #endif

         default: /* assume PCM or float encoding */

-#if 0

-            if (len > wav->numSamples) 

+            /* Cooledit seems to put a non-standard IFF LIST at

+             * the end of the file.  When this is detected,

+             * go ahead and only read in the reported size

+             * of data chunk so the LIST data is not treated

+             * as audio.  

+             * In other cases, go ahead and read unit EOF

+             * This allows us to process WAV files that are

+             * greater then 2Gig and can't be represented

+             * by the 32-bit size field.

+             */

+            if (wav->found_cooledit_list && len > wav->numSamples) 

                 len = wav->numSamples;

-#endif

             done = st_rawread(ft, buf, len);

             /* If software thinks there are more samples but I/O */