Meson testing installation

Meson Testing & Installation

meson test runs project tests, and meson install puts built artifacts into their final location. For pkg-config exports, release tarballs, and packaging layout, use meson-package-export-distribution.

Runnable examples live under this skill's examples/ directory.

Quick Start

Run all tests:

meson test -C build

List available tests:

meson test -C build --list

Run specific test:

meson test -C build test_name

Install project:

meson install -C build

Dry-run (see what would be installed):

meson install -C build --dry-run

Why tests and installs are separate

Testing answers “does it work here?” Installation answers “does the project stage correctly?”

A healthy project should prove both.

meson test: Reference

Basic syntax

meson test [options] [test_names...]

Common options

Option	Purpose
-C BUILDDIR	Build directory
--list	List available tests (don't run)
-j JOBS	Parallel test jobs
-v, --verbose	Verbose output
--no-rebuild	Don't rebuild before testing
--gdb	Run failing tests under gdb
--benchmark	Run benchmark tests
--suite SUITE	Run only one suite
--print-errorlogs	Print failing test logs

Test suites

Use suites to keep test categories readable:

• unit
• integration
• slow
• compile
• install

Example:

test('core', core_test, suite: 'unit')

Install patterns

Install with custom prefix

meson setup build --prefix=/opt/myapp
meson install -C build

Install into a package root

meson install -C build --destdir "$pkgdir"

Use a staging tree in CI

meson install -C build --destdir "$PWD/stage"

Common install helpers

• install_headers()
• install_man()
• install_data()
• install_subdir()
• install_symlink()
• install_emptydir()

Install checklist

• public headers installed intentionally
• shared library versioning set if the ABI is stable
• pkg-config metadata exported if downstream users need it
• private files kept out of the public prefix
• staging installs tested in CI

Common mistakes

• confusing “built successfully” with “installed correctly”
• forgetting install: true on the target that should ship
• letting tests depend on the source directory layout
• writing tests that require a full install when a build-tree test would be simpler
• failing to check the install tree in CI