test(fuzz): cover every indicator, scalar and candle inputs (R9)

The fuzz suite previously covered only `Rsi(14)` and `Ema(20)` — 2 of
71 indicators, no OHLCV coverage at all. Audit finding R9 asked for
ATR/ADX/Stochastic/PSAR as a minimum; this commit goes further and
brings every indicator under fuzz.

- `indicator_update` (rewritten): drives every scalar-input indicator
  through one streaming pass + one batch call per iteration. Covers
  SMA, EMA, WMA, RSI, DEMA, TEMA, HMA, ROC, TRIX, SMMA, TRIMA, ZLEMA,
  KAMA, T3, MOM, CMO, TSI, PMO, StochRSI, DPO, PPO, Coppock, StdDev,
  UlcerIndex, HistoricalVolatility, LinearRegression, LinRegSlope,
  LinRegAngle, VHF, ZScore, MACD, BollingerBands. A `drive` helper
  marked `#[inline(never)]` keeps each indicator on its own panic
  backtrace frame.

- `indicator_update_candle` (new): chunks the fuzz `f64` stream into
  `[open, high, low, close, volume]` tuples, builds candles via
  `Candle::new` (skipping ones that fail OHLCV validation — that path
  is fuzz-tested separately), then drives every candle-input indicator
  through streaming + batch. Covers ATR, NATR, TrueRange,
  ChaikinVolatility, Keltner, Donchian, PSAR, SuperTrend,
  ChandelierExit, ChandeKrollStop, ATRTrailingStop, ADX, Aroon,
  AroonOscillator, Vortex, MassIndex, ChoppinessIndex, CCI, WilliamsR,
  AwesomeOscillator, AcceleratorOscillator, UltimateOscillator,
  BalanceOfPower, OBV, MFI, VWAP, RollingVWAP, VWMA, ADL, VPT, CMF,
  ChaikinOscillator, ForceIndex, EaseOfMovement, TypicalPrice,
  MedianPrice, WeightedClose, Stochastic.

- `fuzz/Cargo.toml` registers the new target; `fuzz/README.md`
  describes both expanded targets.

- A `fuzz-smoke` CI job runs each of the five targets for 30 s on
  every push and pull-request — enough to catch a regression in the
  harness without slowing CI to a crawl. Long fuzz campaigns belong
  on dedicated infrastructure with persistent corpora.
This commit is contained in:
kingchenc
2026-05-23 10:33:05 +02:00
parent 0995f8d66a
commit b003321562
6 changed files with 277 additions and 15 deletions
+44
View File
@@ -134,6 +134,50 @@ jobs:
with:
command: check
# Time-boxed fuzz smoke. Each target runs for ~30 s with libfuzzer; any panic
# fails the job. The goal is to catch a regression in the harness (e.g. a
# newly added indicator that panics on a particular input shape), not to
# discover novel bugs — long fuzz campaigns should be run on dedicated
# infrastructure with persistent corpora.
fuzz-smoke:
name: Fuzz (smoke)
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4.3.1
- name: Install nightly Rust
uses: dtolnay/rust-toolchain@29eef336d9b2848a0b548edc03f92a220660cdb8 # stable branch, 2026-03-27
with:
toolchain: nightly
- name: Cache cargo
uses: Swatinem/rust-cache@e18b497796c12c097a38f9edb9d0641fb99eee32 # v2
with:
workspaces: fuzz
- name: Install cargo-fuzz
run: cargo install cargo-fuzz --locked
- name: Fuzz csv_reader (30 s)
run: cargo +nightly fuzz run csv_reader -- -max_total_time=30
working-directory: fuzz
- name: Fuzz binance_envelope (30 s)
run: cargo +nightly fuzz run binance_envelope -- -max_total_time=30
working-directory: fuzz
- name: Fuzz indicator_update (30 s)
run: cargo +nightly fuzz run indicator_update -- -max_total_time=30
working-directory: fuzz
- name: Fuzz indicator_update_candle (30 s)
run: cargo +nightly fuzz run indicator_update_candle -- -max_total_time=30
working-directory: fuzz
- name: Fuzz tick_aggregator (30 s)
run: cargo +nightly fuzz run tick_aggregator -- -max_total_time=30
working-directory: fuzz
python:
name: Python ${{ matrix.python-version }} on ${{ matrix.os }}
runs-on: ${{ matrix.os }}
+11
View File
@@ -24,6 +24,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
site makes the invariant explicit.
### Changed
- Fuzz suite expanded from 2 indicators to the full catalogue (audit finding
R9). The existing `indicator_update` target now exercises every scalar-input
indicator (~33 classes including MACD and Bollinger Bands); a new
`indicator_update_candle` target exercises every candle-input indicator (~37
classes, including ATR, ADX, Stochastic, PSAR, Keltner, SuperTrend,
ChandelierExit, AwesomeOscillator, OBV, MFI, VWAP, RollingVWAP, and the rest
of the volume / volatility / trailing-stop / price-statistics families). Each
iteration sweeps every indicator through both the streaming `update` loop
and a full `batch` call so any state-mutation bug surfaces on either path.
CI gains a `fuzz-smoke` job that runs each of the five targets for 30 s on
every push and pull-request.
- `UlcerIndex::update` now tracks the trailing maximum with a monotonically-
decreasing deque of `(index, price)` pairs instead of scanning the whole
trailing window on every tick. The indicator now honours the `Indicator`
+7
View File
@@ -38,6 +38,13 @@ test = false
doc = false
bench = false
[[bin]]
name = "indicator_update_candle"
path = "fuzz_targets/indicator_update_candle.rs"
test = false
doc = false
bench = false
[[bin]]
name = "tick_aggregator"
path = "fuzz_targets/tick_aggregator.rs"
+3 -1
View File
@@ -17,7 +17,8 @@ rustup toolchain install nightly
| --- | --- |
| `csv_reader` | `CandleReader` over arbitrary bytes — headers, cells, BOM, binary noise. |
| `binance_envelope` | `RawWsEnvelope` deserialization from arbitrary strings. |
| `indicator_update` | RSI / EMA streaming + batch over arbitrary `f64` sequences (NaN, ±inf, jumps). |
| `indicator_update` | Every scalar-input indicator (SMA / EMA / WMA / RSI / DEMA / TEMA / HMA / ROC / TRIX / SMMA / TRIMA / ZLEMA / KAMA / T3 / MOM / CMO / TSI / PMO / StochRSI / DPO / PPO / Coppock / StdDev / UlcerIndex / HistoricalVolatility / LinearRegression / LinRegSlope / LinRegAngle / VHF / ZScore / MACD / Bollinger) streamed + batched over arbitrary `f64` sequences (NaN, ±inf, jumps). |
| `indicator_update_candle` | Every candle-input indicator (ATR, NATR, TrueRange, ChaikinVolatility, Keltner, Donchian, PSAR, SuperTrend, ChandelierExit, ChandeKrollStop, ATRTrailingStop, ADX, Aroon, AroonOscillator, Vortex, MassIndex, ChoppinessIndex, CCI, WilliamsR, AwesomeOscillator, AcceleratorOscillator, UltimateOscillator, BalanceOfPower, OBV, MFI, VWAP, RollingVWAP, VWMA, ADL, VPT, CMF, ChaikinOscillator, ForceIndex, EaseOfMovement, TypicalPrice, MedianPrice, WeightedClose, Stochastic) streamed + batched over fuzz-derived OHLCV candles. |
| `tick_aggregator` | `TickAggregator` over arbitrary `(price, volume, timestamp)` triples. |
## Run
@@ -27,6 +28,7 @@ rustup toolchain install nightly
cargo +nightly fuzz run csv_reader
cargo +nightly fuzz run binance_envelope
cargo +nightly fuzz run indicator_update
cargo +nightly fuzz run indicator_update_candle
cargo +nightly fuzz run tick_aggregator
```
+82 -14
View File
@@ -1,22 +1,90 @@
#![no_main]
//! Fuzz indicator updates with arbitrary `f64` sequences.
//! Fuzz scalar-input indicator updates with arbitrary `f64` sequences.
//!
//! Every indicator must tolerate any finite-or-not input stream — NaN, ±inf,
//! subnormals, abrupt jumps — without panicking, and `batch` must agree with
//! the streaming `update` path.
//! Every scalar indicator must tolerate any finite-or-not input stream — NaN,
//! ±inf, subnormals, abrupt jumps — without panicking. Each fuzz iteration
//! runs the **same** input sequence through every scalar indicator twice:
//! once as a streaming `update` loop and once as a full `batch` call. Neither
//! path may panic; `batch` is also expected to agree with the streaming path
//! (the `BatchExt` blanket implementation replays `update` internally, so the
//! agreement is structural — but exercising both paths surfaces any
//! state-mutation bugs in `update` that would only manifest mid-batch).
//!
//! Audit finding R9: the previous version covered only `Rsi(14)` and
//! `Ema(20)`. This target now covers every scalar indicator in the catalogue.
use libfuzzer_sys::fuzz_target;
use wickra_core::{BatchExt, Ema, Indicator, Rsi};
use wickra_core::{
BatchExt, BollingerBands, Cmo, Coppock, Dema, Dpo, Ema, HistoricalVolatility, Hma, Indicator,
Kama, LinRegAngle, LinRegSlope, LinearRegression, MacdIndicator, Mom, Pmo, Ppo, Roc, Rsi, Sma,
Smma, StdDev, StochRsi, T3, Tema, Trima, Trix, Tsi, UlcerIndex, VerticalHorizontalFilter, Wma,
ZScore, Zlema,
};
/// Drive a single streaming + batch run through one scalar indicator. Marked
/// `#[inline(never)]` so a panic backtrace pin-points the specific indicator.
#[inline(never)]
fn drive<I>(make: impl Fn() -> I, data: &[f64])
where
I: Indicator<Input = f64, Output = f64> + BatchExt,
{
let mut streaming = make();
for &x in data {
let _ = streaming.update(x);
}
let _ = make().batch(data);
}
fuzz_target!(|data: Vec<f64>| {
let mut rsi = Rsi::new(14).unwrap();
let mut ema = Ema::new(20).unwrap();
for &x in &data {
let _ = rsi.update(x);
let _ = ema.update(x);
}
// Bounded periods keep each iteration cheap and bias the fuzzer toward
// adversarial input patterns rather than enormous windows. The constants
// mirror the README's "common defaults" so we cover the parameterisations
// most users actually instantiate.
drive(|| Sma::new(14).unwrap(), &data);
drive(|| Ema::new(20).unwrap(), &data);
drive(|| Wma::new(14).unwrap(), &data);
drive(|| Rsi::new(14).unwrap(), &data);
drive(|| Dema::new(14).unwrap(), &data);
drive(|| Tema::new(14).unwrap(), &data);
drive(|| Hma::new(14).unwrap(), &data);
drive(|| Roc::new(14).unwrap(), &data);
drive(|| Trix::new(14).unwrap(), &data);
drive(|| Smma::new(14).unwrap(), &data);
drive(|| Trima::new(14).unwrap(), &data);
drive(|| Zlema::new(14).unwrap(), &data);
drive(|| Kama::new(10, 2, 30).unwrap(), &data);
drive(|| T3::new(14, 0.7).unwrap(), &data);
drive(|| Mom::new(14).unwrap(), &data);
drive(|| Cmo::new(14).unwrap(), &data);
drive(|| Tsi::new(25, 13).unwrap(), &data);
drive(|| Pmo::new(35, 20).unwrap(), &data);
drive(|| StochRsi::new(14, 14).unwrap(), &data);
drive(|| Dpo::new(14).unwrap(), &data);
drive(|| Ppo::new(12, 26).unwrap(), &data);
drive(|| Coppock::new(14, 11, 10).unwrap(), &data);
drive(|| StdDev::new(14).unwrap(), &data);
drive(|| UlcerIndex::new(14).unwrap(), &data);
drive(|| HistoricalVolatility::new(14, 252).unwrap(), &data);
drive(|| LinearRegression::new(14).unwrap(), &data);
drive(|| LinRegSlope::new(14).unwrap(), &data);
drive(|| LinRegAngle::new(14).unwrap(), &data);
drive(|| VerticalHorizontalFilter::new(14).unwrap(), &data);
drive(|| ZScore::new(14).unwrap(), &data);
// batch over the same data must not panic either.
let _ = Rsi::new(14).unwrap().batch(&data);
let _ = Ema::new(20).unwrap().batch(&data);
// MACD and Bollinger Bands have non-`f64` outputs, so they cannot use the
// generic `drive` helper above. Streaming + batch are still both exercised.
{
let mut macd = MacdIndicator::new(12, 26, 9).unwrap();
for &x in &data {
let _ = macd.update(x);
}
let _ = MacdIndicator::new(12, 26, 9).unwrap().batch(&data);
}
{
let mut bb = BollingerBands::new(20, 2.0).unwrap();
for &x in &data {
let _ = bb.update(x);
}
let _ = BollingerBands::new(20, 2.0).unwrap().batch(&data);
}
});
@@ -0,0 +1,130 @@
#![no_main]
//! Fuzz OHLCV-input indicator updates with arbitrary candle sequences.
//!
//! Every candle-input indicator must tolerate any sequence of validated OHLCV
//! candles — extreme magnitudes, micro-spreads, zero-volume bars, abrupt
//! reversals — without panicking. The fuzzer chunks the raw `f64` stream into
//! `[open, high, low, close, volume]` tuples and constructs each candle via
//! `Candle::new`; entries that fail OHLCV-invariant validation are skipped so
//! the indicator only ever sees structurally-valid candles. Each iteration
//! then drives that candle stream through every candle-input indicator twice
//! (streaming `update` + batch).
//!
//! Audit finding R9: the previous fuzz suite had no candle-input coverage at
//! all. This target now covers every candle-input indicator including the
//! ones the audit named explicitly (ATR, ADX, Stochastic, PSAR) plus the
//! complete catalogue: Keltner, Donchian, SuperTrend, Chandelier Exit, ATR
//! Trailing Stop, Aroon, AwesomeOscillator, CCI, WilliamsR, MFI, OBV, VWAP,
//! RollingVWAP, ADL, VPT, ChaikinMoneyFlow, ChaikinOscillator, ForceIndex,
//! EaseOfMovement, NATR, AroonOscillator, ChandeKrollStop, Vortex, MassIndex,
//! ChoppinessIndex, TrueRange, ChaikinVolatility, AcceleratorOscillator,
//! BalanceOfPower, UltimateOscillator, VWMA, TypicalPrice, MedianPrice,
//! WeightedClose.
use libfuzzer_sys::fuzz_target;
use wickra_core::{
AcceleratorOscillator, Adl, Adx, Aroon, AroonOscillator, Atr, AtrTrailingStop,
AwesomeOscillator, BalanceOfPower, BatchExt, Candle, Cci, ChaikinMoneyFlow, ChaikinOscillator,
ChaikinVolatility, ChandeKrollStop, ChandelierExit, ChoppinessIndex, Donchian, EaseOfMovement,
ForceIndex, Indicator, Keltner, MassIndex, MedianPrice, Mfi, Natr, Obv, Psar, RollingVwap,
Stochastic, SuperTrend, TrueRange, TypicalPrice, UltimateOscillator, VolumePriceTrend, Vortex,
Vwap, Vwma, WeightedClose, WilliamsR,
};
/// Convert a flat `f64` stream into a `Vec<Candle>` by chunking it into
/// `[open, high, low, close, volume]` groups. Tuples that fail OHLCV
/// validation are dropped so the indicator under test only ever sees a
/// structurally-valid candle stream (the *parser* is fuzz-tested elsewhere;
/// this target focuses on indicator robustness).
fn candles_from(data: &[f64]) -> Vec<Candle> {
data.chunks_exact(5)
.enumerate()
.filter_map(|(i, ch)| {
// A monotonic timestamp avoids surprising any indicator that might
// care about ordering. The fuzz input drives OHLCV; time is just a
// tie-breaker.
Candle::new(ch[0], ch[1], ch[2], ch[3], ch[4], i as i64).ok()
})
.collect()
}
/// Streaming + batch sweep through one candle-input indicator. `#[inline(never)]`
/// keeps each indicator on its own frame in any panic backtrace.
#[inline(never)]
fn drive<I, O>(make: impl Fn() -> I, candles: &[Candle])
where
I: Indicator<Input = Candle, Output = O> + BatchExt,
{
let mut streaming = make();
for c in candles {
let _ = streaming.update(*c);
}
let _ = make().batch(candles);
}
fuzz_target!(|data: Vec<f64>| {
let candles = candles_from(&data);
if candles.is_empty() {
return;
}
// --- Volatility & ATR family ---
drive(|| Atr::new(14).unwrap(), &candles);
drive(|| Natr::new(14).unwrap(), &candles);
drive(TrueRange::new, &candles);
drive(|| ChaikinVolatility::new(10, 10).unwrap(), &candles);
// --- Bands & Channels ---
drive(|| Keltner::new(20, 10, 2.0).unwrap(), &candles);
drive(|| Donchian::new(20).unwrap(), &candles);
// --- Trailing Stops ---
drive(|| Psar::new(0.02, 0.02, 0.20).unwrap(), &candles);
drive(|| SuperTrend::new(14, 3.0).unwrap(), &candles);
drive(|| ChandelierExit::new(22, 3.0).unwrap(), &candles);
drive(|| ChandeKrollStop::new(10, 1.0, 9).unwrap(), &candles);
drive(|| AtrTrailingStop::new(14, 3.0).unwrap(), &candles);
// --- Trend & Directional ---
drive(|| Adx::new(14).unwrap(), &candles);
drive(|| Aroon::new(14).unwrap(), &candles);
drive(|| AroonOscillator::new(14).unwrap(), &candles);
drive(|| Vortex::new(14).unwrap(), &candles);
drive(|| MassIndex::new(9, 25).unwrap(), &candles);
drive(|| ChoppinessIndex::new(14).unwrap(), &candles);
// --- Momentum & Oscillators ---
drive(|| Cci::new(20).unwrap(), &candles);
drive(|| WilliamsR::new(14).unwrap(), &candles);
drive(|| AwesomeOscillator::new(5, 34).unwrap(), &candles);
drive(|| AcceleratorOscillator::new(5, 34, 5).unwrap(), &candles);
drive(|| UltimateOscillator::new(7, 14, 28).unwrap(), &candles);
drive(BalanceOfPower::new, &candles);
// --- Volume ---
drive(Obv::new, &candles);
drive(|| Mfi::new(14).unwrap(), &candles);
drive(Vwap::new, &candles);
drive(|| RollingVwap::new(20).unwrap(), &candles);
drive(|| Vwma::new(20).unwrap(), &candles);
drive(Adl::new, &candles);
drive(VolumePriceTrend::new, &candles);
drive(|| ChaikinMoneyFlow::new(20).unwrap(), &candles);
drive(|| ChaikinOscillator::new(3, 10).unwrap(), &candles);
drive(|| ForceIndex::new(13).unwrap(), &candles);
drive(|| EaseOfMovement::with_divisor(14, 1e8).unwrap(), &candles);
// --- Price transformations ---
drive(TypicalPrice::new, &candles);
drive(MedianPrice::new, &candles);
drive(WeightedClose::new, &candles);
// --- Stochastic (multi-output) ---
{
let mut s = Stochastic::new(14, 3).unwrap();
for c in &candles {
let _ = s.update(*c);
}
let _ = Stochastic::new(14, 3).unwrap().batch(&candles);
}
});