Quick start
From nothing to a shrunk counterexample in five steps.
1. Install fe and Node
You need fe 26.6 and Node 22.18 or newer:
curl -fsSL https://raw.githubusercontent.com/argotorg/fe/master/feup/feup.sh | bash -s -- --version v26.6.0
node --versiondiff and --backend native also need fe's native backend, which runs on Linux and macOS. On Windows, install fe inside WSL as well; Fuzzloom then runs native tests there by itself.
2. Check your setup
Run this from inside any Fe ingot (a directory with fe.toml and src/):
npx fuzzloom doctorok Node v22.18.0
ok fe fe 26.6.0 (5b3f96a) at /home/you/.fe/bin/fe
ok Fe library .../node_modules/fuzzloom/ingot
ok Library compiles fe check passed
ok Native backend fe 26.6.0 (5b3f96a) in this machine
warn Project ingot my_ingot: 0 prop_* and 0 diff_* functions, 0 skipped
write `pub fn prop_<name>(args) -> bool` (or diff_<name>) in the ingot's src/The project warning goes away once you write your first property.
3. Write a property
Add a public function whose name starts with prop_ and which returns bool, anywhere under src/:
pub fn fee(_ amount: u64) -> u64 {
amount.wrapping_mul(30) / 10000
}
/// The fee matches exact arithmetic for every amount.
pub fn prop_fee_matches_exact_math(_ amount: u64) -> bool {
let exact: u128 = (amount as u128) * 30 / 10000
(fee(amount) as u128) == exact
}4. Fuzz it
npx fuzzloom fuzz --seed 7FAIL prop_fee_matches_exact_math(u64) counterexample at input 2
amount = 614891469123651721
shrunk from 4762291152399575021 in 1414 evaluations
reproduce: fuzzloom reproduce prop_fee_matches_exact_math-0e05b819614891469123651721 is u64::MAX / 30 + 1, the smallest amount for which amount * 30 wraps. The shrinker found the exact threshold. The seed makes the run repeatable; leave it out and Fuzzloom picks one at random and prints it.
5. Fix it and replay
Widen before multiplying, then check the saved failure again:
pub fn fee(_ amount: u64) -> u64 {
((amount as u128) * 30 / 10000).downcast_truncate()
}npx fuzzloom replayFIXED prop_fee_matches_exact_math-0e05b819 now passesCommit .fuzzloom/failures/: it is your regression suite. Every later fuzz also mutates those saved inputs, to catch a fix that only handles the one input that was saved.
Next
- Writing properties: types, labels and what gets skipped.
- Effects and fixtures: properties that use state or context.
- Differential testing: compare sonatina and native.
- Running in CI: exit codes and JUnit reports.