Pest vs Rust libtest

Pest is a free PHP testing framework for developers and AI agents, installed in a PHP project as a Composer development dependency. Pest 5 requires PHP 8.4 or later. It supports parallel execution, coverage, mutation and snapshot testing, profiling, architectural testing and browser testing. Its Browser plugin uses Playwright for page interactions, while the Agent plugin lets coding agents run one-off application checks and can drive a real browser when the Browser plugin is installed. Pest Evals combines deterministic tests with AI-powered scorers for evaluating LLM output. The Tia Engine reruns tests affected by recent changes and requires PCOV or Xdebug as a coverage driver. Pest 5 includes PHPStan and Rector plugins; team-maintained plugins include Faker, Laravel and Livewire. JetBrains maintains the PhpStorm plugin, included by default since version 2023.3. Pest supports parameterized cases but does not include built-in mocks. Its maintainers say bug fixes for outdated major versions are available for 12 months after the latest version's release, without a guaranteed resolution time or fix for every issue.

Rust libtest is Rust’s built-in test harness for compiling and running crate tests. Building with `rustc --test` links the executable to libtest; tests are free functions marked `#[test]`. Cargo’s `cargo test` handles building and running unit, integration, and documentation tests, including code samples extracted from library documentation comments. A test succeeds when it returns without an error; a panic or non-zero termination result marks failure. You can mark tests to expect a panic or to be ignored, filter by name, choose exact matches, skip matches, select ignored tests, or list tests and benchmarks without running them. Tests run in parallel by default, with the thread count configurable. The harness can capture output, display successful-test output, and format results as pretty, terse, or JSON; JSON formatting is unstable. Rust libtest is free and supports Linux, macOS, and Windows. Catching panics requires the unwind strategy; abort-strategy support is experimental. Benchmarking and custom harnesses are experimental and nightly-only.