The handling of 16bit samples in the peak calculations has a bug when a sample with 0x8000 is passed. As abs() treats 32bit int, it returns 0x8000. And yet the code stores back into 16bit value again.
To fix that overflow, use 32bit value (i.e. val instead of sval) for the further calculations.
Signed-off-by: Takashi Iwai tiwai@suse.de --- aplay/aplay.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/aplay/aplay.c b/aplay/aplay.c index 91af244edb60..c884346c9f25 100644 --- a/aplay/aplay.c +++ b/aplay/aplay.c @@ -1829,9 +1829,9 @@ static void compute_max_peak(u_char *data, size_t samples) else sval = be16toh(*valp); sval ^= mask; - sval = abs(sval); - if (max_peak[c] < sval) - max_peak[c] = sval; + val = abs(sval); + if (max_peak[c] < val) + max_peak[c] = val; valp++; if (vumeter == VUMETER_STEREO) c = !c;