fixing slice decoding, dunno why the regression tests didnt catch that ...

Originally committed as revision 1744 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/error_resilience.c

@@ -614,13 +614,13 @@ void ff_er_add_slice(MpegEncContext *s, int startx, int starty, int endx, int en
             s->error_status_table[i] &= mask;
         }
     }
-
+   
-    s->error_status_table[start_xy] |= VP_START;
-    
     if(end_xy < s->mb_num){
         s->error_status_table[end_xy] &= mask;
         s->error_status_table[end_xy] |= status;
     }
+ 
+    s->error_status_table[start_xy] |= VP_START;
 }
 
 void ff_er_frame_end(MpegEncContext *s){

libavcodec/h263dec.c

@@ -484,7 +484,7 @@ retry:
     avctx->has_b_frames= !s->low_delay;
 
     if(s->workaround_bugs&FF_BUG_AUTODETECT){
-        if(s->padding_bug_score > -2 && !s->data_partitioning)
+        if(s->padding_bug_score > -2 && !s->data_partitioning && !s->resync_marker)
             s->workaround_bugs |=  FF_BUG_NO_PADDING;
         else
             s->workaround_bugs &= ~FF_BUG_NO_PADDING;