On Tue, Nov 8, 2011 at 12:51 AM, Kostya Shishkov <[email protected]> wrote: > On Mon, Nov 07, 2011 at 06:48:54PM -0800, Alex Converse wrote: >> Found with asan and the venerable 1-dog.avi sample. >> --- >> Input from the indeo experts is welcome. >> >> Furthermore if the indeo experts have extra cycles please look at 1-dog.avi >> http://samples.libav.org/samples/archive/extension/AVI/avi+indeo3+++1-dog.avi > > Looks like heavily zzuf'ed file - there are bits flipped everywhere. For > example, in header chunk LIST was turned in MIST (which is German word > describing that file), JUNK length was 0x10E92 instead of correct 0x0E92 and > 'movi' size was only 0x30714 instead of 0xB0714. Audio TWOCC is 0x2002 while > it's likely to be just 0x0002 (M$ ADPCM instead of RealAudio 14.4). 1025 > channels for it are unlikely too.
Yes the file is fuzzed. It was used to show that the previous indeo3 decoder was not only incomprehensible but a buffer overflow disaster. The file need not actually decode correctly. I mean I'd like the decoder to be valgrind/asan clean against it. _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
