use std::convert::TryInto;
use std::cmp;
use std::sync::{Arc, Mutex};
-use std::sync::atomic::{AtomicU64,AtomicUsize,Ordering};
+use std::sync::atomic::{AtomicU64,AtomicUsize,AtomicBool,Ordering};
use bech32::u5;
#[inline]
struct InputData {
data: Vec<u8>,
read_pos: AtomicUsize,
+ halt_fee_est_reads: AtomicBool,
}
impl InputData {
fn get_slice(&self, len: usize) -> Option<&[u8]> {
}
impl FeeEstimator for FuzzEstimator {
fn get_est_sat_per_1000_weight(&self, _: ConfirmationTarget) -> u32 {
+ if self.input.halt_fee_est_reads.load(Ordering::Acquire) {
+ return 253;
+ }
//TODO: We should actually be testing at least much more than 64k...
match self.input.get_slice(2) {
Some(slice) => cmp::max(slice_to_be16(slice) as u32, 253),
let input = Arc::new(InputData {
data: data.to_vec(),
read_pos: AtomicUsize::new(0),
+ halt_fee_est_reads: AtomicBool::new(false),
});
let fee_est = Arc::new(FuzzEstimator {
input: input.clone(),
11 => {
let mut txn = broadcast.txn_broadcasted.lock().unwrap().split_off(0);
if !txn.is_empty() {
+ input.halt_fee_est_reads.store(true, Ordering::Release);
loss_detector.connect_block(&txn[..]);
for _ in 2..100 {
loss_detector.connect_block(&[]);
}
+ input.halt_fee_est_reads.store(false, Ordering::Release);
}
for tx in txn.drain(..) {
loss_detector.funding_txn.push(tx);