Skip to content

testing.md

Latest commit

 

History

History
75 lines (51 loc) · 1.85 KB

testing.md

File metadata and controls

75 lines (51 loc) · 1.85 KB
Raw

Testing

macadmin has two test harnesses. The right one to use depends on what you're testing.

Zsh runner (tests/run.zsh)

The primary harness. Discovers tests/test_*.zsh, runs each in a subprocess, captures exit code and output.

Each test file:

#!/usr/bin/env zsh
# shellcheck shell=bash

emulate -L zsh
set -o errexit -o nounset -o pipefail

SCRIPT_DIR=${0:a:h}
source "$SCRIPT_DIR/assert.zsh"

# Mock system tools by injecting tests/mocks into PATH
export PATH="$SCRIPT_DIR/mocks:$PATH"

pass "boot"

# run_cmd "<prefix>" runs `<prefix> -- <cmd...>` and captures exit + stdout
out=$(run_cmd "system-info" -- zsh "$SCRIPT_DIR/../scripts/system_info.zsh")
assert_exit0 "$?"

# assert_contains <haystack> <needle>
assert_contains "$out" "Darwin"

tests/assert.zsh provides pass, fail, run_cmd, assert_exit0, assert_contains. Extend with new helpers rather than reinventing.

Mocks live in tests/mocks/. A mock is a shell script that imitates the real tool. make test does chmod +x tests/mocks/* before running, so mocks can be edited freely.

Bats library tests (tests/lib/*.bats)

Used for library-only tests where shell function assertions are more ergonomic in bats. Not every library needs bats coverage — duplicate only when bats is genuinely more comfortable than the zsh runner.

@test "macadmin_path_cache returns $HOME/Library/Caches" {
  source lib/paths.zsh
  result="$(macadmin_path_cache)"
  [[ "$result" == "$HOME/Library/Caches" ]]
}

Coverage

make coverage lists commands and reports which lack tests:

Coverage: 7 / 8 commands have tests
Untested: brew-tools

Target: 100% command coverage.

Linting

make lint runs shellcheck and shfmt in diff mode. make format auto-fixes.

CI

GitHub Actions runs make ci (= lint + test + coverage) on every push and PR. See .github/workflows/ci.yml.