Test Tools

Beyond the built-in cargo test, these tools give deeper visibility into test runs.

cargo-nextest

Drop-in replacement for cargo test with better output.

What it gives you

• Per-test timing on stable (no nightly needed)
• Live progress bar showing which tests are running
• Retry support for flaky tests
• JUnit XML output for CI or browser viewing
• Better failure diagnostics

Install

cargo install cargo-nextest

Usage

# Run all core tests (replaces cargo test -p nine65 --lib --release)
cargo nextest run -p nine65 --lib --release

# With filter
cargo nextest run -p nine65 --lib --release -- bootstrap

# List tests
cargo nextest list -p nine65 --lib --release

# Generate JUnit XML report
cargo nextest run -p nine65 --lib --release --message-format libtest-json > results.xml

cargo-pretty-test

Renders test results as a module tree with colors and timing.

Install

cargo install cargo-pretty-test

Usage

cargo pretty-test -p nine65 --lib --release

cargo-llvm-cov

Code coverage — shows which lines of code your tests actually execute.

What it gives you

• HTML report you open in a browser
• Every file with green (covered) / red (not covered) line highlighting
• Summary showing percentage coverage per module
• Identifies dead code and untested paths

Install

cargo install cargo-llvm-cov

You also need the llvm-tools component:

rustup component add llvm-tools-preview

Usage

# Generate HTML coverage report
cargo llvm-cov --release --workspace --exclude nine65-python --exclude nine65-wasm --html

# Open the report
# Output goes to target/llvm-cov/html/index.html
xdg-open target/llvm-cov/html/index.html

# Just core library coverage
cargo llvm-cov -p nine65 --lib --release --html

# Text summary (no HTML)
cargo llvm-cov -p nine65 --lib --release

# JSON output for tooling
cargo llvm-cov -p nine65 --lib --release --json

Reading the Report

The HTML report shows:

• Green lines: Executed by at least one test
• Red lines: Never executed — potential gaps
• Per-file percentage: How much of each file is covered
• Per-function coverage: Which functions were tested

Focus on red lines in critical files:

• ops/bootstrap.rs — every bootstrap path should be green
• arithmetic/k_elimination.rs — CRT reconstruction must be covered
• noise/budget.rs — all budget operations should be tested
• security/ — constant-time operations need full coverage

Miri (Nightly Only)

Memory safety checker. Detects undefined behavior that normal tests can’t catch.

What it catches

• Use-after-free
• Out-of-bounds access
• Uninitialized memory reads
• Data races
• Invalid pointer arithmetic

Install

rustup toolchain install nightly
rustup +nightly component add miri

Usage

# Run a specific test under miri (very slow)
cargo +nightly miri test -p nine65 --lib -- montgomery::tests::test_montgomery_mul

# Run with more diagnostics
MIRIFLAGS="-Zmiri-backtrace=full" cargo +nightly miri test -p nine65 --lib -- specific_test

Miri is 100-1000x slower than normal test execution. Use it for targeted verification of unsafe code or security-sensitive arithmetic, not for full suite runs.

Limitations

• Cannot handle FFI calls
• Cannot handle inline assembly
• Some system calls are unsupported
• Very slow — minutes per test

Comparison

Tool	Speed	What it shows	When to use
cargo test	Fast	Pass/fail	Daily workflow
cargo test -- --nocapture	Fast	Pass/fail + println output	Debugging specific behavior
cargo nextest run	Fast	Pass/fail + timing + progress	When you want timing data
cargo pretty-test	Fast	Module tree view	Visual overview
cargo llvm-cov	Medium	Line-level coverage	Finding untested code
cargo +nightly miri test	Very slow	Memory safety violations	Verifying unsafe code