F4: Headless test harness

Add an in-repo test runner that discovers test/**/test_*.gd and runs under --headless.
It uses Godot's Logger API, so engine and script errors during a test fail that test.
make test runs the suite, and make test-selfcheck runs deliberately failing fixtures
to prove each kind of failure is detected.

Closes #4

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-11-09 17:12:59 -05:00
co-authored by Claude Opus 5.5
parent a4b0c43cf0
commit dee417df3d
21 changed files with 388 additions and 2 deletions
+9 -2
View File
@@ -1,10 +1,10 @@
# Tidewall build entry points. Everything runs headless with the repo-local engine. # Tidewall build entry points. Everything runs headless with the repo-local engine.
GODOT := tools/godot.sh GODOT := tools/godot.sh
.PHONY: check import parse binaries verify-engine .PHONY: check import parse test test-selfcheck binaries verify-engine
## Everything that must pass before a push. ## Everything that must pass before a push.
check: verify-engine binaries parse check: verify-engine binaries parse test test-selfcheck
verify-engine: verify-engine:
@$(GODOT) --verify @$(GODOT) --verify
@@ -18,3 +18,10 @@ import:
parse: import parse: import
@$(GODOT) --script res://tools/parse_all.gd @$(GODOT) --script res://tools/parse_all.gd
test: import
@tools/run_tests.sh
## Proves the test runner detects each kind of failure.
test-selfcheck: import
@tools/test_selfcheck.sh
View File
+5
View File
@@ -0,0 +1,5 @@
extends TestCase
func test_failing_assertion() -> void:
assert_eq(1 + 1, 3, "deliberate fixture failure")
View File
+8
View File
@@ -0,0 +1,8 @@
extends TestCase
func test_engine_error() -> void:
var node: Node = Node.new()
# The engine reports a parameter error and carries on.
node.add_child(null)
node.free()
+5
View File
@@ -0,0 +1,5 @@
extends TestCase
func test_declared_error_never_happens() -> void:
expect_error("an error that is never raised")
+5
View File
@@ -0,0 +1,5 @@
extends RefCounted
func test_wrong_base_class() -> void:
pass
+5
View File
@@ -0,0 +1,5 @@
extends TestCase
func test_does_not_parse() -> void
this is not valid GDScript
+17
View File
@@ -0,0 +1,17 @@
extends TestCase
func test_passes() -> void:
assert_true(true)
assert_eq(1 + 1, 2)
assert_near(0.1 + 0.2, 0.3, 0.0001)
func test_expected_engine_error_does_not_fail() -> void:
expect_error("deliberate fixture error")
push_error("deliberate fixture error")
func test_may_await() -> void:
await Engine.get_main_loop().process_frame
assert_true(true)
+6
View File
@@ -0,0 +1,6 @@
extends TestCase
func test_null_dereference() -> void:
var missing: Node = null
missing.get_name()
+107
View File
@@ -0,0 +1,107 @@
## Base class for tests. Files named test_*.gd extend this and define test_* methods.
## Test methods may await. Any engine or script error raised during a test fails it,
## unless the test declared that error with expect_error().
class_name TestCase
extends RefCounted
var _failures: Array[String] = []
var _expected_errors: Array[String] = []
## Runs before each test. Override as needed; may await.
func before_each() -> void:
pass
## Runs after each test, even if the test failed. Override as needed; may await.
func after_each() -> void:
pass
## Records a failure. The test keeps running, so one run reports every failure.
func fail(message: String) -> void:
_failures.append(message)
## Declares that an engine error containing `fragment` is expected in this test.
## The test fails if the error does not happen.
func expect_error(fragment: String) -> void:
_expected_errors.append(fragment)
func assert_true(condition: bool, message: String = "") -> void:
if not condition:
fail(_describe("expected true", message))
func assert_false(condition: bool, message: String = "") -> void:
if condition:
fail(_describe("expected false", message))
func assert_eq(actual: Variant, expected: Variant, message: String = "") -> void:
if not _same(actual, expected):
fail(_describe("expected %s, got %s" % [_show(expected), _show(actual)], message))
func assert_ne(actual: Variant, unexpected: Variant, message: String = "") -> void:
if _same(actual, unexpected):
fail(_describe("expected a value other than %s" % _show(unexpected), message))
func assert_near(actual: float, expected: float, tolerance: float, message: String = "") -> void:
if absf(actual - expected) > tolerance:
var detail: String = "expected %f within %f, got %f" % [expected, tolerance, actual]
fail(_describe(detail, message))
func assert_null(value: Variant, message: String = "") -> void:
if value != null:
fail(_describe("expected null, got %s" % _show(value), message))
func assert_not_null(value: Variant, message: String = "") -> void:
if value == null:
fail(_describe("expected a value, got null", message))
## Clears per-test state. Called by the runner.
func reset_for_test() -> void:
_failures = []
_expected_errors = []
## Returns this test's failures, including unexpected and missing engine errors.
func finish_test(engine_errors: Array[String]) -> Array[String]:
var results: Array[String] = _failures.duplicate()
var expected: Array[String] = _expected_errors.duplicate()
for error: String in engine_errors:
var matched: int = -1
for index: int in expected.size():
if error.contains(expected[index]):
matched = index
break
if matched >= 0:
expected.remove_at(matched)
else:
results.append("engine error: %s" % error)
for missing: String in expected:
results.append("expected an engine error containing '%s', but none occurred" % missing)
return results
func _describe(detail: String, message: String) -> String:
return detail if message.is_empty() else "%s: %s" % [message, detail]
## Exact comparison that never treats values of different types as equal,
## except int and float, which compare by value.
func _same(a: Variant, b: Variant) -> bool:
var numeric: Array[int] = [TYPE_INT, TYPE_FLOAT]
if typeof(a) != typeof(b) and not (typeof(a) in numeric and typeof(b) in numeric):
return false
return a == b
func _show(value: Variant) -> String:
return "'%s'" % value if value is String else str(value)
+1
View File
@@ -0,0 +1 @@
uid://b8gr1lo80bhyb
+36
View File
@@ -0,0 +1,36 @@
## Records engine and script errors so the runner can fail the test that caused them.
## Godot may call a Logger from any thread, so all state is guarded by a mutex, and
## nothing here prints (printing from a logger callback would recurse).
class_name TestLogger
extends Logger
var _mutex: Mutex = Mutex.new()
var _errors: Array[String] = []
func _log_error(
function: String,
file: String,
line: int,
code: String,
rationale: String,
_editor_notify: bool,
error_type: int,
_script_backtrace: Array[ScriptBacktrace],
) -> void:
if error_type == Logger.ERROR_TYPE_WARNING:
return
var what: String = rationale if not rationale.is_empty() else code
var entry: String = "%s (at %s:%d in %s)" % [what, file, line, function]
_mutex.lock()
_errors.append(entry)
_mutex.unlock()
## Returns every error recorded since the last call, and clears the record.
func take_errors() -> Array[String]:
_mutex.lock()
var taken: Array[String] = _errors
_errors = []
_mutex.unlock()
return taken
+1
View File
@@ -0,0 +1 @@
uid://bpgk2h6e073tn
+126
View File
@@ -0,0 +1,126 @@
## Headless test runner. Discovers test_*.gd files, runs their test_* methods, and
## exits non-zero if anything fails.
##
## tools/run_tests.sh [--root=res://test/sim] [--filter=substring]
extends SceneTree
const DEFAULT_ROOT: String = "res://test"
const FIXTURES_ROOT: String = "res://test/fixtures"
var _logger: TestLogger = TestLogger.new()
var _root: String = DEFAULT_ROOT
var _filter: String = ""
var _passed: int = 0
var _failed: int = 0
var _reports: Array[String] = []
func _initialize() -> void:
OS.add_logger(_logger)
_parse_args()
_run.call_deferred()
func _parse_args() -> void:
for arg: String in OS.get_cmdline_user_args():
if arg.begins_with("--root="):
_root = arg.trim_prefix("--root=").trim_suffix("/")
elif arg.begins_with("--filter="):
_filter = arg.trim_prefix("--filter=")
else:
push_error("Unknown test runner argument: %s" % arg)
func _run() -> void:
var files: Array[String] = []
_discover(_root, files)
files.sort()
# A runner that finds nothing must not look like a passing suite.
var ran_nothing: bool = files.is_empty()
for path: String in files:
await _run_file(path)
if _passed + _failed == 0:
ran_nothing = true
_finish(ran_nothing)
func _discover(dir_path: String, out: Array[String]) -> void:
# Fixtures contain deliberate failures. They only run when targeted directly.
var skipping_fixtures: bool = dir_path == FIXTURES_ROOT and not _root.begins_with(FIXTURES_ROOT)
if skipping_fixtures:
return
for name: String in DirAccess.get_files_at(dir_path):
if name.begins_with("test_") and name.ends_with(".gd"):
out.append(dir_path.path_join(name))
for name: String in DirAccess.get_directories_at(dir_path):
if not name.begins_with("."):
_discover(dir_path.path_join(name), out)
func _run_file(path: String) -> void:
_logger.take_errors()
var script: GDScript = load(path) as GDScript
var load_errors: Array[String] = _logger.take_errors()
if script == null or not script.can_instantiate():
_record_load_failure(path, "could not compile the script", load_errors)
return
var instance: TestCase = script.new() as TestCase
if instance == null:
_record_load_failure(path, "script does not extend TestCase", _logger.take_errors())
return
var names: Array[String] = _test_names(script)
for test_name: String in names:
if _filter.is_empty() or (path + ":" + test_name).contains(_filter):
await _run_test(instance, path, test_name)
func _test_names(script: GDScript) -> Array[String]:
var names: Array[String] = []
for method: Dictionary in script.get_script_method_list():
var method_name: String = method["name"]
if method_name.begins_with("test_") and method_name not in names:
names.append(method_name)
names.sort()
return names
func _run_test(instance: TestCase, path: String, test_name: String) -> void:
_logger.take_errors()
instance.reset_for_test()
await instance.before_each()
await instance.call(test_name)
await instance.after_each()
# Let deferred calls raise their errors while this test is still current.
await process_frame
_record(path, test_name, instance.finish_test(_logger.take_errors()))
func _record_load_failure(path: String, reason: String, errors: Array[String]) -> void:
var failures: Array[String] = [reason]
failures.append_array(errors)
_record(path, "(load)", failures)
func _record(path: String, test_name: String, failures: Array[String]) -> void:
var label: String = "%s::%s" % [path.trim_prefix("res://"), test_name]
if failures.is_empty():
_passed += 1
print("PASS %s" % label)
return
_failed += 1
print("FAIL %s" % label)
for failure: String in failures:
print(" - %s" % failure)
_reports.append("%s: %s" % [label, failure])
func _finish(ran_nothing: bool) -> void:
print("")
for report: String in _reports:
print("FAILED: %s" % report)
if ran_nothing:
print("No tests ran under %s. Treating this as a failure." % _root)
print("Tests: %d passed, %d failed." % [_passed, _failed])
quit(1 if _failed > 0 or ran_nothing else 0)
+1
View File
@@ -0,0 +1 @@
uid://cr7xjx1o67nl0
+1
View File
@@ -0,0 +1 @@
uid://paoe2bxdshd7
+1
View File
@@ -0,0 +1 @@
uid://btff4b5o7gimu
+12
View File
@@ -0,0 +1,12 @@
extends TestCase
func test_engine_is_the_pinned_version() -> void:
var info: Dictionary = Engine.get_version_info()
assert_eq(info["major"], 4)
assert_eq(info["minor"], 7)
assert_eq(info["patch"], 2)
func test_simulation_runs_at_thirty_hertz() -> void:
assert_eq(ProjectSettings.get_setting("physics/common/physics_ticks_per_second"), 30)
+8
View File
@@ -0,0 +1,8 @@
#!/usr/bin/env bash
# Runs the headless test suite. Extra arguments are passed to the runner, for
# example: tools/run_tests.sh --root=res://test/sim
set -uo pipefail
cd "$(dirname "${BASH_SOURCE[0]}")/.."
# A hang in a test must fail the run, not block it forever.
exec timeout --signal=KILL "${TIDEWALL_TEST_TIMEOUT:-900}" \
tools/godot.sh --script res://test/framework/runner.gd -- "$@"
+34
View File
@@ -0,0 +1,34 @@
#!/usr/bin/env bash
# Proves the test runner detects each kind of failure, using deliberate fixtures.
set -uo pipefail
cd "$(dirname "${BASH_SOURCE[0]}")/.."
status=0
# check <fixture> <expected exit> <text that must appear in the output>
check() {
local fixture="$1" expected="$2" needle="$3" output code
output="$(tools/run_tests.sh "--root=res://test/fixtures/$fixture" 2>&1)"
code=$?
if [[ "$code" -ne "$expected" ]]; then
echo "selfcheck FAIL $fixture: exit $code, expected $expected"
echo "$output" | tail -n 20
status=1
elif ! grep -qF -- "$needle" <<<"$output"; then
echo "selfcheck FAIL $fixture: output lacks '$needle'"
echo "$output" | tail -n 20
status=1
else
echo "selfcheck ok $fixture (exit $code)"
fi
}
check passes 0 "Tests: 3 passed, 0 failed."
check assert 1 "deliberate fixture failure: expected 3, got 2"
check script_error 1 "engine error:"
check engine_error 1 "engine error:"
check parse_error 1 "could not compile the script"
check not_a_test 1 "does not extend TestCase"
check missing_error 1 "expected an engine error containing"
check empty 1 "No tests ran"
exit "$status"