[Gnash-commit] gnash ChangeLog server/parser/video_stream_def.cpp

[Sandro Santilli]

CVSROOT:        /sources/gnash
Module name:    gnash
Changes by:     Sandro Santilli <strk>  07/10/19 10:29:17

Modified files:
        .              : ChangeLog 
        server/parser  : video_stream_def.cpp 

Log message:
                * server/parser/video_stream_def.cpp (read): rework the input
                  buffer padding to be cleaner, safer and a bit faster (only
                  memset the pad); (destructor): get rid of the embedded video
                  frames. Add TODO items.

CVSWeb URLs:
http://cvs.savannah.gnu.org/viewcvs/gnash/ChangeLog?cvsroot=gnash&r1=1.4644&r2=1.4645
http://cvs.savannah.gnu.org/viewcvs/gnash/server/parser/video_stream_def.cpp?cvsroot=gnash&r1=1.20&r2=1.21

Patches:
Index: ChangeLog
===================================================================
RCS file: /sources/gnash/gnash/ChangeLog,v
retrieving revision 1.4644
retrieving revision 1.4645
diff -u -b -r1.4644 -r1.4645
--- ChangeLog   19 Oct 2007 09:30:23 -0000      1.4644
+++ ChangeLog   19 Oct 2007 10:29:17 -0000      1.4645
@@ -1,5 +1,12 @@
 2007-10-19 Sandro Santilli <address@hidden>
 
+       * server/parser/video_stream_def.cpp (read): rework the input
+         buffer padding to be cleaner, safer and a bit faster (only
+         memset the pad); (destructor): get rid of the embedded video
+         frames. Add TODO items.
+
+2007-10-19 Sandro Santilli <address@hidden>
+
        * server/parser/video_stream_def.{cpp,h}: keep _decoder by
          boost::scoped_ptr. Fixes bug #21284.
        * server/swf/tag_loaders.cpp (video_loader): don't abort on malformed

Index: server/parser/video_stream_def.cpp
===================================================================
RCS file: /sources/gnash/gnash/server/parser/video_stream_def.cpp,v
retrieving revision 1.20
retrieving revision 1.21
diff -u -b -r1.20 -r1.21
--- server/parser/video_stream_def.cpp  19 Oct 2007 09:30:24 -0000      1.20
+++ server/parser/video_stream_def.cpp  19 Oct 2007 10:29:17 -0000      1.21
@@ -16,7 +16,7 @@
 // Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 
 // 
-// $Id: video_stream_def.cpp,v 1.20 2007/10/19 09:30:24 strk Exp $
+// $Id: video_stream_def.cpp,v 1.21 2007/10/19 10:29:17 strk Exp $
 
 #include "video_stream_def.h"
 #include "video_stream_instance.h"
@@ -42,7 +42,11 @@
 
 video_stream_definition::~video_stream_definition()
 {
-       m_video_frames.clear(); // TODO: isn't this useless ?
+       for (EmbedFrameMap::iterator i=m_video_frames.begin(), 
e=m_video_frames.end();
+                       i!=e; ++i)
+       {
+               delete i->second;
+       }
 }
 
 
@@ -88,32 +92,39 @@
                // it belongs.
                in->skip_bytes(2); //int frameNum = in->read_u16();
 
-               // We need to make the buffer a bit bigger than the data
-               // to avoid libavcodec (ffmpeg) making illegal reads.
-               // The reason is a bit sketchy, but it seems that the h263
-               // decoder (perhaps other decoders as well) assumes that the
-               // buffer with the data is bigger than the data it contains.
-               // We make the buffer 4 bytes bigger than the data, and set
-               // them to 0.
-
-               // The data size is 4 bytes smaller than this, but because of 
-               // what is mentioned above we don't subtract the 4 bytes.
-               int size = in->get_tag_length();
-               uint8_t* data = new uint8_t[size];
-               memset(data, 0, size);
-               for (int i = 0; i < size-4; i++) // The size-variable 4 bytes 
bigger than the data
+               // We need to make the buffer a FF_INPUT_BUFFER_PADDING_SIZE
+               // bigger than the data to avoid libavcodec (ffmpeg) making
+               // illegal reads. Also, we must ensure first 23 bits to be 
+               // zeroed out. We'll zero out all padding.
+               unsigned int padding =  FF_INPUT_BUFFER_PADDING_SIZE;
+               unsigned int dataSize = in->get_tag_end_position() - 
in->get_position();
+               unsigned int totSize = dataSize+padding;
+
+               boost::scoped_array<uint8_t> data ( new uint8_t[totSize] );
+               for (unsigned int i = 0; i < dataSize; ++i) 
                {
                        data[i] = in->read_u8();
                }
+               memset(&data[dataSize], 0, padding);  // padd with zeroes
 
-               image::image_base* img = _decoder->decodeToImage(data, size);
+               // TODO: should we pass dataSize instead of totSize to 
decodeToImage ?
+               std::auto_ptr<image::image_base> img ( 
_decoder->decodeToImage(data.get(), totSize) );
 
-               if (img) {
-                       m_video_frames[m->get_loading_frame()] = img;
-               } else {
+               if ( img.get() )
+               {
+                       // TODO: why don't we use  the frame number specified
+                       //       in the tag instead of skipping it ?
+                       size_t frameNum = m->get_loading_frame();
+
+                       // TODO: use a private function for this as we
+                       //       should eventually delete any previously
+                       //       existing element for key frameNum
+                       m_video_frames[frameNum] = img.release();
+               }
+               else
+               {
                        log_error(_("An error occured while decoding video 
frame in frame %d"), m->get_loading_frame());
                }
-               delete [] data;
        }
 
 }