Reimar Döffinger <[email protected]> added the comment: On Sat, Feb 05, 2011 at 08:29:06PM +0000, Ronald S. Bultje wrote: > It fixes some, but not all. wc -l of valgrind ffmpeg goes from ~2000 to > ~400, but still more warnings remain: > > ==61513== Invalid read of size 4 > ==61513== at 0x10036E8BA: vc1_decode_i_blocks_adv (in ./ffmpeg_g) > ==61513== by 0x3000000017: ??? > ==61513== by 0x1010DC41F: ??? > ==61513== by 0x7FFF0000000C: ??? > ==61513== by 0x300000019: ??? > ==61513== by 0x16FFFFFFFF: ??? > ==61513== by 0x7FFF5FBFE46B: ??? > ==61513== by 0x7FFF5FBFE467: ??? > ==61513== by 0x7FFF5FBFE463: ??? > ==61513== by 0x10105FFF2: ??? > ==61513== by 0x58FF: ??? > ==61513== by 0x11: ??? > ==61513== Address 0x1010beedc is 3,436 bytes inside a block of size > 3,439 alloc'd
*shrug* that's a one-byte overread, I do not care that much about that. I don't think this is really the right place to discuss this, the bug report was about a "stack overflow", assuming that means a "stack buffer overflow" (as in, writes+on stack) that is really, really serious and needs to be fixed ASAP. These overreads do not. ________________________________________________ FFmpeg issue tracker <[email protected]> <https://roundup.ffmpeg.org/issue2584> ________________________________________________
