cargo-nextest is a faster, more reliable test runner for Rust that executes each test in its own process for better isolation and parallel performance.

Installation

# Install cargo-nextest
cargo install cargo-nextest --locked

# Verify installation
cargo nextest --version

Basic Usage

# Run all tests with nextest
cargo nextest run

# Run specific test
cargo nextest run test_name

# Run tests in specific package
cargo nextest run -p package_name

# Run with verbose output
cargo nextest run --verbose

# Run ignored tests
cargo nextest run -- --ignored

# Run all tests including ignored
cargo nextest run -- --include-ignored

Configuration File

Create .config/nextest.toml in your project root:

[profile.default]
# Number of retries for flaky tests
retries = 0

# Test threads (default: number of logical CPUs)
test-threads = 8

# Fail fast - stop on first failure
fail-fast = false

# Show output for passing tests
success-output = "never"  # never, immediate, final, immediate-final

# Show output for failing tests
failure-output = "immediate"  # never, immediate, final, immediate-final

[profile.ci]
# CI-specific configuration
retries = 2
fail-fast = true
success-output = "never"
failure-output = "immediate-final"

# Slow test timeout
slow-timeout = { period = "60s", terminate-after = 2 }

[profile.ci.junit]
# JUnit XML output for CI
path = "target/nextest/ci/junit.xml"

Test Filtering with Expression Language

# Run tests matching pattern
cargo nextest run -E 'test(auth)'

# Run tests in specific binary
cargo nextest run -E 'binary(my_app)'

# Run tests in specific package
cargo nextest run -E 'package(my_crate)'

# Complex expressions with operators
cargo nextest run -E 'test(auth) and not test(slow)'

# Run all integration tests
cargo nextest run -E 'kind(test)'

# Run only unit tests in library
cargo nextest run -E 'kind(lib) and test(/)'

Expression Operators

test(pattern) - Match test name (regex)
binary(pattern) - Match binary name
package(pattern) - Match package name
kind(type) - Match test kind (lib, bin, test, bench, example)
platform(os) - Match target platform
not expr - Logical NOT
expr and expr - Logical AND
expr or expr - Logical OR

Parallel Execution

Nextest runs each test in its own process by default, providing:

Better isolation - Tests cannot interfere with each other
True parallelism - No global state conflicts
Fault isolation - One test crash doesn't affect others
Better resource management - Each test has clean environment

# Control parallelism
cargo nextest run --test-threads 4

# Single-threaded execution
cargo nextest run --test-threads 1

# Use all available cores
cargo nextest run --test-threads 0

Output Formats

# Human-readable output (default)
cargo nextest run

# JSON output for programmatic parsing
cargo nextest run --message-format json

# JSON with test output
cargo nextest run --message-format json-pretty

# Generate JUnit XML for CI
cargo nextest run --profile ci

# Combine with cargo-llvm-cov for coverage
cargo llvm-cov nextest --html

Flaky Test Management

# In .config/nextest.toml
[profile.default]
# Retry flaky tests automatically
retries = 3

# Detect tests that pass on retry
[profile.default.overrides]
filter = 'test(flaky_)'
retries = 5

# Run with retries
cargo nextest run --retries 3

# Show retry statistics
cargo nextest run --retries 3 --failure-output immediate-final

GitHub Actions Integration

name: Test

on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - uses: dtolnay/rust-toolchain@stable

      - name: Install nextest
        uses: taiki-e/install-action@v2
        with:
          tool: nextest

      - name: Run tests
        run: cargo nextest run --profile ci --all-features

      - name: Upload test results
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: test-results
          path: target/nextest/ci/junit.xml

      - name: Publish test results
        uses: EnricoMi/publish-unit-test-result-action@v2
        if: always()
        with:
          files: target/nextest/ci/junit.xml

Advanced Configuration

# .config/nextest.toml
[profile.default]
# Timeout for slow tests
slow-timeout = { period = "30s", terminate-after = 3 }

# Test groups for resource management
[test-groups.database]
max-threads = 1  # Serialize database tests

[profile.default.overrides]
filter = 'test(db_)'
test-group = 'database'

# Platform-specific overrides
[[profile.default.overrides]]
platform = 'x86_64-pc-windows-msvc'
slow-timeout = { period = "60s" }

Doctests Limitation

Important: nextest does not support doctests. Use cargo test for doctests:

# Run doctests separately
cargo test --doc

# Run both nextest and doctests in CI
cargo nextest run && cargo test --doc

Workspace Configuration

# In workspace root .config/nextest.toml
[profile.default]
default-filter = 'all()'

# Per-package overrides
[[profile.default.overrides]]
filter = 'package(slow_tests)'
test-threads = 1
slow-timeout = { period = "120s" }

Comparison with cargo test

| Execution model | In-process | Per-test process

| Parallelism | Thread-based | Process-based

| Test isolation | Shared state | Complete isolation

| Output format | Limited | Rich (JSON, JUnit)

| Flaky test handling | Manual | Built-in retries

| Doctests | Supported | Not supported

| Performance | Good | Excellent

Best Practices

Use profiles for different environments:

default for local development

ci for continuous integration
coverage for coverage analysis
Configure flaky test detection:

[profile.default]
retries = 2

Group resource-intensive tests:

[test-groups.expensive]
max-threads = 2

Set appropriate timeouts:

[profile.default]
slow-timeout = { period = "60s", terminate-after = 2 }

Use expression filters effectively:

# Run fast tests during development
cargo nextest run -E 'not test(slow_)'

Always run doctests separately in CI:

- run: cargo nextest run --all-features
- run: cargo test --doc --all-features

Troubleshooting

Tests timeout too quickly:

[profile.default]
slow-timeout = { period = "120s", terminate-after = 3 }

Too much parallel execution:

cargo nextest run --test-threads 4

Need test output for debugging:

cargo nextest run --success-output immediate --failure-output immediate

Flaky tests in CI:

[profile.ci]
retries = 3
failure-output = "immediate-final"

cargo-nextest

安装