Remove misleading comment
This creates ambiguity about bitstream validity.
From a reference decoder point of view, there should be no ambiguity
whether one stream is valid or not.
And there should be no mention about the way the input
stream was created (e.g "fuzzed inputs").
Change-Id: I5afa8d6c615ed03275d122f2261ee1f492d628bc
diff --git a/av1/common/av1_inv_txfm1d.c b/av1/common/av1_inv_txfm1d.c
index 95a92fb..1e566ef 100644
--- a/av1/common/av1_inv_txfm1d.c
+++ b/av1/common/av1_inv_txfm1d.c
@@ -1087,8 +1087,6 @@
(void)cos_bit;
(void)stage_range;
for (int i = 0; i < 4; ++i) {
- // Normal input should fit into 32-bit. Cast to 64-bit here to avoid
- // overflow with corrupted/fuzzed input. The same for av1_iidentity/16/64_c.
output[i] = round_shift((int64_t)NewSqrt2 * input[i], NewSqrt2Bits);
}
assert(stage_range[0] + NewSqrt2Bits <= 32);
