Am Mittwoch, den 18.02.2015, 12:11 +0100 schrieb Fabian Greffrath: > This time, it was a simple logical error in the lame sources: The "fake" > sample rate of the fuzzed input file is 16319999 kHz which lame tries to > sample down to 48 kHz in the process of encoding. The ratio between > input and output samplerates is thus 16319999/48000=339.999979 which is > very close, but only close, to the integer value 340.
Actually, the bug is easy to reproduce even without a fuzzed sample. Take a random valid WAV file and convert it to another sample rate, e.g. sox /path/to/dummy.wav --rate 95999 crash.wav This time, I set the input sample frequency to 95999 Hz, which is 2*48kHz-1Hz. The "intratio" variable will be set to 1 again, although fabs(resample_ratio - floor(.5 + resample_ratio)) == 0.00002 < 0.0001 != 0. The crash.wav sample will thus crash LAME. - Fabian -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
