Use AV_R/WN32 instead of direct integer copying.
authorJeff Downs <heydowns@borg.com>
Wed, 27 May 2009 18:59:24 +0000 (18:59 +0000)
committerJeff Downs <heydowns@borg.com>
Wed, 27 May 2009 18:59:24 +0000 (18:59 +0000)
Fixes decoding on at least Sparc/Solaris where unaligned accesses cause
crashes.
ML thread: http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/91414

Originally committed as revision 18967 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/lcldec.c

index e4114f8..8b31b5f 100644 (file)
@@ -129,7 +129,7 @@ static unsigned int mszh_decomp(unsigned char * srcptr, int srclen, unsigned cha
         if ((mask & (1 << (--maskbit))) == 0) {
             if (destptr + 4 > destptr_end)
                 break;
         if ((mask & (1 << (--maskbit))) == 0) {
             if (destptr + 4 > destptr_end)
                 break;
-            *(int*)destptr = *(int*)srcptr;
+            AV_WN32(destptr, AV_RN32(srcptr));
             srclen -= 4;
             destptr += 4;
             srcptr += 4;
             srclen -= 4;
             destptr += 4;
             srcptr += 4;