Reimar Döffinger <b...@reimardoeffinger.de> added the comment: On Sat, Feb 05, 2011 at 07:49:34PM +0000, Ronald S. Bultje wrote: > ==56585== Invalid read of size 4 > ==56585== Address 0x101059597 is 1,687 bytes inside a block of size > 1,690 alloc > 'd > ==56585== at 0x100CDFD06: memalign (vg_replace_malloc.c:581) > ==56585== by 0x100CDFD5F: posix_memalign (vg_replace_malloc.c:709) > ==56585== by 0x1004B50DC: av_mallocz (in ./ffmpeg_g) > ==56585== by 0x7FFF5FBFE91F: ??? > [..] > > ==56585== Invalid read of size 4 > ==56585== Address 0x101059598 is 1,688 bytes inside a block of size > 1,690 alloc > 'd > ==56585== at 0x100CDFD06: memalign (vg_replace_malloc.c:581) > ==56585== by 0x100CDFD5F: posix_memalign (vg_replace_malloc.c:709) > ==56585== by 0x1004B50DC: av_mallocz (in ./ffmpeg_g) > ==56585== by 0x7FFF5FBFE91F: ???
Those are invalid reads and not even on stack memory, so I see no relation with this issue. And they are probably fixed by the patch I sent quite some time ago. Probably it's this one: Index: libavcodec/vc1dec.c =================================================================== --- libavcodec/vc1dec.c (revision 26402) +++ libavcodec/vc1dec.c (working copy) @@ -1375,7 +1375,7 @@ if (index != vc1_ac_sizes[codingset] - 1) { run = vc1_index_decode_table[codingset][index][0]; level = vc1_index_decode_table[codingset][index][1]; - lst = index >= vc1_last_decode_table[codingset]; + lst = index >= vc1_last_decode_table[codingset] || get_bits_left(gb) < 0; if(get_bits1(gb)) level = -level; } else { ________________________________________________ FFmpeg issue tracker <iss...@roundup.ffmpeg.org> <https://roundup.ffmpeg.org/issue2584> ________________________________________________