Merge pull request #4958 from bonzini/tap

Add initial TAP test support

12 files changed, 554 insertions, 29 deletions

diff --git a/docs/markdown/Reference-manual.md b/docs/markdown/Reference-manual.md
index 6c3b3ad7a..f2b0416d3 100644
--- a/docs/markdown/Reference-manual.md
+++ b/docs/markdown/Reference-manual.md
@@ -1406,10 +1406,7 @@ executable to run.  The executable can be an [executable build target
 object](#build-target-object) returned by
 [`executable()`](#executable) or an [external program
 object](#external-program-object) returned by
-[`find_program()`](#find_program). The executable's exit code is used
-by the test harness to record the outcome of the test, for example
-exit code zero indicates success. For more on the Meson test harness
-protocol read [Unit Tests](Unit-tests.md).
+[`find_program()`](#find_program).
 
 Keyword arguments are the following:
 
@@ -1446,6 +1443,12 @@ Keyword arguments are the following:
   before test is executed even if they have `build_by_default : false`.
   Since 0.46.0
 
+- `protocol` specifies how the test results are parsed and can be one
+  of `exitcode` (the executable's exit code is used by the test harness
+  to record the outcome of the test) or `tap` ([Test Anything
+  Protocol](https://www.testanything.org/)). For more on the Meson test
+  harness protocol read [Unit Tests](Unit-tests.md). Since 0.50.0
+
 Defined tests can be run in a backend-agnostic way by calling
 `meson test` inside the build dir, or by using backend-specific
 commands, such as `ninja test` or `msbuild RUN_TESTS.vcxproj`.
diff --git a/docs/markdown/Unit-tests.md b/docs/markdown/Unit-tests.md
index a8e72737c..9e6173910 100644
--- a/docs/markdown/Unit-tests.md
+++ b/docs/markdown/Unit-tests.md
@@ -51,9 +51,15 @@ By default Meson uses as many concurrent processes as there are cores on the tes
 $ MESON_TESTTHREADS=5 ninja test
 ```
 
-## Skipped tests
+## Skipped tests and hard errors
 
-Sometimes a test can only determine at runtime that it can not be run. The GNU standard approach in this case is to exit the program with error code 77. Meson will detect this and report these tests as skipped rather than failed. This behavior was added in version 0.37.0.
+Sometimes a test can only determine at runtime that it can not be run.
+
+For the default `exitcode` testing protocol, the GNU standard approach in this case is to exit the program with error code 77. Meson will detect this and report these tests as skipped rather than failed. This behavior was added in version 0.37.0.
+
+For TAP-based tests, skipped tests should print a single line starting with `1..0 # SKIP`.
+
+In addition, sometimes a test fails set up so that it should fail even if it is marked as an expected failure. The GNU standard approach in this case is to exit the program with error code 99. Again, Meson will detect this and report these tests as `ERROR`, ignoring the setting of `should_fail`. This behavior was added in version 0.50.0.
 
 ## Testing tool
 
diff --git a/mesonbuild/backend/backends.py b/mesonbuild/backend/backends.py
index 506276708..4d35d223d 100644
--- a/mesonbuild/backend/backends.py
+++ b/mesonbuild/backend/backends.py
@@ -84,7 +84,7 @@ class ExecutableSerialisation:
 
 class TestSerialisation:
     def __init__(self, name, project, suite, fname, is_cross_built, exe_wrapper, is_parallel,
-                 cmd_args, env, should_fail, timeout, workdir, extra_paths):
+                 cmd_args, env, should_fail, timeout, workdir, extra_paths, protocol):
         self.name = name
         self.project_name = project
         self.suite = suite
@@ -100,6 +100,7 @@ class TestSerialisation:
         self.timeout = timeout
         self.workdir = workdir
         self.extra_paths = extra_paths
+        self.protocol = protocol
 
 class OptionProxy:
     def __init__(self, name, value):
@@ -756,7 +757,7 @@ class Backend:
                     raise MesonException('Bad object in test command.')
             ts = TestSerialisation(t.get_name(), t.project_name, t.suite, cmd, is_cross,
                                    exe_wrapper, t.is_parallel, cmd_args, t.env,
-                                   t.should_fail, t.timeout, t.workdir, extra_paths)
+                                   t.should_fail, t.timeout, t.workdir, extra_paths, t.protocol)
             arr.append(ts)
         return arr
 
diff --git a/mesonbuild/interpreter.py b/mesonbuild/interpreter.py
index 5f19bc5db..8bde727a4 100644
--- a/mesonbuild/interpreter.py
+++ b/mesonbuild/interpreter.py
@@ -850,7 +850,7 @@ class RunTargetHolder(InterpreterObject, ObjectHolder):
 
 class Test(InterpreterObject):
     def __init__(self, name, project, suite, exe, depends, is_parallel,
-                 cmd_args, env, should_fail, timeout, workdir):
+                 cmd_args, env, should_fail, timeout, workdir, protocol):
         InterpreterObject.__init__(self)
         self.name = name
         self.suite = suite
@@ -863,6 +863,7 @@ class Test(InterpreterObject):
         self.should_fail = should_fail
         self.timeout = timeout
         self.workdir = workdir
+        self.protocol = protocol
 
     def get_exe(self):
         return self.exe
@@ -1973,7 +1974,8 @@ permitted_kwargs = {'add_global_arguments': {'language', 'native'},
                     'library': known_library_kwargs,
                     'subdir': {'if_found'},
                     'subproject': {'version', 'default_options', 'required'},
-                    'test': {'args', 'depends', 'env', 'is_parallel', 'should_fail', 'timeout', 'workdir', 'suite'},
+                    'test': {'args', 'depends', 'env', 'is_parallel', 'should_fail', 'timeout', 'workdir',
+                             'suite', 'protocol'},
                     'vcs_tag': {'input', 'output', 'fallback', 'command', 'replace_string'},
                     }
 
@@ -3269,6 +3271,9 @@ This will become a hard error in the future.''' % kwargs['input'], location=self
             workdir = None
         if not isinstance(timeout, int):
             raise InterpreterException('Timeout must be an integer.')
+        protocol = kwargs.get('protocol', 'exitcode')
+        if protocol not in ('exitcode', 'tap'):
+            raise InterpreterException('Protocol must be "exitcode" or "tap".')
         suite = []
         prj = self.subproject if self.is_subproject() else self.build.project_name
         for s in mesonlib.stringlistify(kwargs.get('suite', '')):
@@ -3280,7 +3285,7 @@ This will become a hard error in the future.''' % kwargs['input'], location=self
             if not isinstance(dep, (build.CustomTarget, build.BuildTarget)):
                 raise InterpreterException('Depends items must be build targets.')
         t = Test(args[0], prj, suite, exe.held_object, depends, par, cmd_args,
-                 env, should_fail, timeout, workdir)
+                 env, should_fail, timeout, workdir, protocol)
         if is_base_test:
             self.build.tests.append(t)
             mlog.debug('Adding test', mlog.bold(args[0], True))
diff --git a/mesonbuild/mtest.py b/mesonbuild/mtest.py
index 6536558de..02b728e45 100644
--- a/mesonbuild/mtest.py
+++ b/mesonbuild/mtest.py
@@ -23,6 +23,9 @@ from mesonbuild.dependencies import ExternalProgram
 from mesonbuild.mesonlib import substring_is_in_list, MesonException
 from mesonbuild import mlog
 
+from collections import namedtuple
+import io
+import re
 import tempfile
 import time, datetime, multiprocessing, json
 import concurrent.futures as conc
@@ -36,6 +39,10 @@ import enum
 # mean that the test should be skipped.
 GNU_SKIP_RETURNCODE = 77
 
+# GNU autotools interprets a return code of 99 from tests it executes to
+# mean that the test failed even before testing what it is supposed to test.
+GNU_ERROR_RETURNCODE = 99
+
 def is_windows():
     platname = platform.system().lower()
     return platname == 'windows' or 'mingw' in platname
@@ -146,11 +153,202 @@ class TestResult(enum.Enum):
     FAIL = 'FAIL'
     EXPECTEDFAIL = 'EXPECTEDFAIL'
     UNEXPECTEDPASS = 'UNEXPECTEDPASS'
+    ERROR = 'ERROR'
+
+
+class TAPParser(object):
+    Plan = namedtuple('Plan', ['count', 'late', 'skipped', 'explanation'])
+    Bailout = namedtuple('Bailout', ['message'])
+    Test = namedtuple('Test', ['number', 'name', 'result', 'explanation'])
+    Error = namedtuple('Error', ['message'])
+    Version = namedtuple('Version', ['version'])
+
+    _MAIN = 1
+    _AFTER_TEST = 2
+    _YAML = 3
+
+    _RE_BAILOUT = r'Bail out!\s*(.*)'
+    _RE_DIRECTIVE = r'(?:\s*\#\s*([Ss][Kk][Ii][Pp]\S*|[Tt][Oo][Dd][Oo])\b\s*(.*))?'
+    _RE_PLAN = r'1\.\.([0-9]+)' + _RE_DIRECTIVE
+    _RE_TEST = r'((?:not )?ok)\s*(?:([0-9]+)\s*)?([^#]*)' + _RE_DIRECTIVE
+    _RE_VERSION = r'TAP version ([0-9]+)'
+    _RE_YAML_START = r'(\s+)---.*'
+    _RE_YAML_END = r'\s+\.\.\.\s*'
+
+    def __init__(self, io):
+        self.io = io
+
+    def parse_test(self, ok, num, name, directive, explanation):
+        name = name.strip()
+        explanation = explanation.strip() if explanation else None
+        if directive is not None:
+            directive = directive.upper()
+            if directive == 'SKIP':
+                if ok:
+                    yield self.Test(num, name, TestResult.SKIP, explanation)
+                    return
+            elif directive == 'TODO':
+                yield self.Test(num, name, TestResult.UNEXPECTEDPASS if ok else TestResult.EXPECTEDFAIL, explanation)
+                return
+            else:
+                yield self.Error('invalid directive "%s"' % (directive,))
+
+        yield self.Test(num, name, TestResult.OK if ok else TestResult.FAIL, explanation)
+
+    def parse(self):
+        found_late_test = False
+        bailed_out = False
+        plan = None
+        lineno = 0
+        num_tests = 0
+        yaml_lineno = None
+        yaml_indent = None
+        state = self._MAIN
+        version = 12
+        while True:
+            lineno += 1
+            try:
+                line = next(self.io).rstrip()
+            except StopIteration:
+                break
+
+            # YAML blocks are only accepted after a test
+            if state == self._AFTER_TEST:
+                if version >= 13:
+                    m = re.match(self._RE_YAML_START, line)
+                    if m:
+                        state = self._YAML
+                        yaml_lineno = lineno
+                        yaml_indent = m.group(1)
+                        continue
+                state = self._MAIN
+
+            elif state == self._YAML:
+                if re.match(self._RE_YAML_END, line):
+                    state = self._MAIN
+                    continue
+                if line.startswith(yaml_indent):
+                    continue
+                yield self.Error('YAML block not terminated (started on line %d)' % (yaml_lineno,))
+                state = self._MAIN
+
+            assert state == self._MAIN
+            if line.startswith('#'):
+                continue
+
+            m = re.match(self._RE_TEST, line)
+            if m:
+                if plan and plan.late and not found_late_test:
+                    yield self.Error('unexpected test after late plan')
+                    found_late_test = True
+                num_tests += 1
+                num = num_tests if m.group(2) is None else int(m.group(2))
+                if num != num_tests:
+                    yield self.Error('out of order test numbers')
+                yield from self.parse_test(m.group(1) == 'ok', num,
+                                           m.group(3), m.group(4), m.group(5))
+                state = self._AFTER_TEST
+                continue
+
+            m = re.match(self._RE_PLAN, line)
+            if m:
+                if plan:
+                    yield self.Error('more than one plan found')
+                else:
+                    count = int(m.group(1))
+                    skipped = (count == 0)
+                    if m.group(2):
+                        if m.group(2).upper().startswith('SKIP'):
+                            if count > 0:
+                                yield self.Error('invalid SKIP directive for plan')
+                            skipped = True
+                        else:
+                            yield self.Error('invalid directive for plan')
+                    plan = self.Plan(count=count, late=(num_tests > 0),
+                                     skipped=skipped, explanation=m.group(3))
+                    yield plan
+                continue
+
+            m = re.match(self._RE_BAILOUT, line)
+            if m:
+                yield self.Bailout(m.group(1))
+                bailed_out = True
+                continue
+
+            m = re.match(self._RE_VERSION, line)
+            if m:
+                # The TAP version is only accepted as the first line
+                if lineno != 1:
+                    yield self.Error('version number must be on the first line')
+                    continue
+                version = int(m.group(1))
+                if version < 13:
+                    yield self.Error('version number should be at least 13')
+                else:
+                    yield self.Version(version=version)
+                continue
+
+            yield self.Error('unexpected input at line %d' % (lineno,))
+
+        if state == self._YAML:
+            yield self.Error('YAML block not terminated (started on line %d)' % (yaml_lineno,))
+
+        if not bailed_out and plan and num_tests != plan.count:
+            if num_tests < plan.count:
+                yield self.Error('Too few tests run (expected %d, got %d)' % (plan.count, num_tests))
+            else:
+                yield self.Error('Too many tests run (expected %d, got %d)' % (plan.count, num_tests))
 
 
 class TestRun:
-    def __init__(self, res, returncode, should_fail, duration, stdo, stde, cmd,
-                 env):
+    @staticmethod
+    def make_exitcode(test, returncode, duration, stdo, stde, cmd):
+        if returncode == GNU_SKIP_RETURNCODE:
+            res = TestResult.SKIP
+        elif returncode == GNU_ERROR_RETURNCODE:
+            res = TestResult.ERROR
+        elif test.should_fail:
+            res = TestResult.EXPECTEDFAIL if bool(returncode) else TestResult.UNEXPECTEDPASS
+        else:
+            res = TestResult.FAIL if bool(returncode) else TestResult.OK
+        return TestRun(test, res, returncode, duration, stdo, stde, cmd)
+
+    def make_tap(test, returncode, duration, stdo, stde, cmd):
+        res = None
+        num_tests = 0
+        failed = False
+        num_skipped = 0
+
+        for i in TAPParser(io.StringIO(stdo)).parse():
+            if isinstance(i, TAPParser.Bailout):
+                res = TestResult.ERROR
+            elif isinstance(i, TAPParser.Test):
+                if i.result == TestResult.SKIP:
+                    num_skipped += 1
+                elif i.result in (TestResult.FAIL, TestResult.UNEXPECTEDPASS):
+                    failed = True
+                num_tests += 1
+            elif isinstance(i, TAPParser.Error):
+                res = TestResult.ERROR
+                stde += '\nTAP parsing error: ' + i.message
+
+        if returncode != 0:
+            res = TestResult.ERROR
+            stde += '\n(test program exited with status code %d)' % (returncode,)
+
+        if res is None:
+            # Now determine the overall result of the test based on the outcome of the subcases
+            if num_skipped == num_tests:
+                # This includes the case where num_tests is zero
+                res = TestResult.SKIP
+            elif test.should_fail:
+                res = TestResult.EXPECTEDFAIL if failed else TestResult.UNEXPECTEDPASS
+            else:
+                res = TestResult.FAIL if failed else TestResult.OK
+
+        return TestRun(test, res, returncode, duration, stdo, stde, cmd)
+
+    def __init__(self, test, res, returncode, duration, stdo, stde, cmd):
         assert isinstance(res, TestResult)
         self.res = res
         self.returncode = returncode
@@ -158,8 +356,8 @@ class TestRun:
         self.stdo = stdo
         self.stde = stde
         self.cmd = cmd
-        self.env = env
-        self.should_fail = should_fail
+        self.env = test.env
+        self.should_fail = test.should_fail
 
     def get_log(self):
         res = '--- command ---\n'
@@ -257,9 +455,8 @@ class SingleTestRunner:
         cmd = self._get_cmd()
         if cmd is None:
             skip_stdout = 'Not run because can not execute cross compiled binaries.'
-            return TestRun(res=TestResult.SKIP, returncode=GNU_SKIP_RETURNCODE,
-                           should_fail=self.test.should_fail, duration=0.0,
-                           stdo=skip_stdout, stde=None, cmd=None, env=self.test.env)
+            return TestRun(test=self.test, res=TestResult.SKIP, returncode=GNU_SKIP_RETURNCODE,
+                           duration=0.0, stdo=skip_stdout, stde=None, cmd=None)
         else:
             wrap = TestHarness.get_wrapper(self.options)
             if self.options.gdb:
@@ -388,14 +585,12 @@ class SingleTestRunner:
             stdo = ""
             stde = additional_error
         if timed_out:
-            res = TestResult.TIMEOUT
-        elif p.returncode == GNU_SKIP_RETURNCODE:
-            res = TestResult.SKIP
-        elif self.test.should_fail:
-            res = TestResult.EXPECTEDFAIL if bool(p.returncode) else TestResult.UNEXPECTEDPASS
+            return TestRun(self.test, TestResult.TIMEOUT, p.returncode, duration, stdo, stde, cmd)
         else:
-            res = TestResult.FAIL if bool(p.returncode) else TestResult.OK
-        return TestRun(res, p.returncode, self.test.should_fail, duration, stdo, stde, cmd, self.test.env)
+            if self.test.protocol == 'exitcode':
+                return TestRun.make_exitcode(self.test, p.returncode, duration, stdo, stde, cmd)
+            else:
+                return TestRun.make_tap(self.test, p.returncode, duration, stdo, stde, cmd)
 
 
 class TestHarness:
@@ -471,7 +666,7 @@ class TestHarness:
             self.skip_count += 1
         elif result.res is TestResult.OK:
             self.success_count += 1
-        elif result.res is TestResult.FAIL:
+        elif result.res is TestResult.FAIL or result.res is TestResult.ERROR:
             self.fail_count += 1
         elif result.res is TestResult.EXPECTEDFAIL:
             self.expectedfail_count += 1
@@ -493,9 +688,11 @@ class TestHarness:
             (num, name, padding1, result.res.value, padding2, result.duration,
              status)
         ok_statuses = (TestResult.OK, TestResult.EXPECTEDFAIL)
+        bad_statuses = (TestResult.FAIL, TestResult.TIMEOUT, TestResult.UNEXPECTEDPASS,
+                        TestResult.ERROR)
         if not self.options.quiet or result.res not in ok_statuses:
             if result.res not in ok_statuses and mlog.colorize_console:
-                if result.res in (TestResult.FAIL, TestResult.TIMEOUT, TestResult.UNEXPECTEDPASS):
+                if result.res in bad_statuses:
                     decorator = mlog.red
                 elif result.res is TestResult.SKIP:
                     decorator = mlog.yellow
@@ -505,8 +702,7 @@ class TestHarness:
             else:
                 print(result_str)
         result_str += "\n\n" + result.get_log()
-        if (result.returncode != GNU_SKIP_RETURNCODE) \
-                and (result.returncode != 0) != result.should_fail:
+        if result.res in bad_statuses:
             if self.options.print_errorlogs:
                 self.collected_logs.append(result_str)
         if self.logfile:
diff --git a/run_unittests.py b/run_unittests.py
index 32c7875fb..e13903a65 100755
--- a/run_unittests.py
+++ b/run_unittests.py
@@ -27,6 +27,7 @@ import unittest
 import platform
 import pickle
 import functools
+import io
 from itertools import chain
 from unittest import mock
 from configparser import ConfigParser
@@ -54,6 +55,8 @@ from mesonbuild.dependencies import PkgConfigDependency, ExternalProgram
 from mesonbuild.build import Target
 import mesonbuild.modules.pkgconfig
 
+from mesonbuild.mtest import TAPParser, TestResult
+
 from run_tests import (
     Backend, FakeBuild, FakeCompilerOptions,
     ensure_backend_detects_changes, exe_suffix, get_backend_commands,
@@ -5728,6 +5731,272 @@ class CrossFileTests(BasePlatformTests):
                               '-Ddef_sysconfdir=sysconfbar'])
 
 
+class TAPParserTests(unittest.TestCase):
+    def assert_test(self, events, **kwargs):
+        if 'explanation' not in kwargs:
+            kwargs['explanation'] = None
+        self.assertEqual(next(events), TAPParser.Test(**kwargs))
+
+    def assert_plan(self, events, **kwargs):
+        if 'skipped' not in kwargs:
+            kwargs['skipped'] = False
+        if 'explanation' not in kwargs:
+            kwargs['explanation'] = None
+        self.assertEqual(next(events), TAPParser.Plan(**kwargs))
+
+    def assert_version(self, events, **kwargs):
+        self.assertEqual(next(events), TAPParser.Version(**kwargs))
+
+    def assert_error(self, events):
+        self.assertEqual(type(next(events)), TAPParser.Error)
+
+    def assert_bailout(self, events, **kwargs):
+        self.assertEqual(next(events), TAPParser.Bailout(**kwargs))
+
+    def assert_last(self, events):
+        with self.assertRaises(StopIteration):
+            next(events)
+
+    def parse_tap(self, s):
+        parser = TAPParser(io.StringIO(s))
+        return iter(parser.parse())
+
+    def parse_tap_v13(self, s):
+        events = self.parse_tap('TAP version 13\n' + s)
+        self.assert_version(events, version=13)
+        return events
+
+    def test_empty(self):
+        events = self.parse_tap('')
+        self.assert_last(events)
+
+    def test_empty_plan(self):
+        events = self.parse_tap('1..0')
+        self.assert_plan(events, count=0, late=False, skipped=True)
+        self.assert_last(events)
+
+    def test_plan_directive(self):
+        events = self.parse_tap('1..0 # skipped for some reason')
+        self.assert_plan(events, count=0, late=False, skipped=True,
+                         explanation='for some reason')
+        self.assert_last(events)
+
+        events = self.parse_tap('1..1 # skipped for some reason\nok 1')
+        self.assert_error(events)
+        self.assert_plan(events, count=1, late=False, skipped=True,
+                         explanation='for some reason')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+        events = self.parse_tap('1..1 # todo not supported here\nok 1')
+        self.assert_error(events)
+        self.assert_plan(events, count=1, late=False, skipped=False,
+                         explanation='not supported here')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_one_test_ok(self):
+        events = self.parse_tap('ok')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_one_test_with_number(self):
+        events = self.parse_tap('ok 1')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_one_test_with_name(self):
+        events = self.parse_tap('ok 1 abc')
+        self.assert_test(events, number=1, name='abc', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_one_test_not_ok(self):
+        events = self.parse_tap('not ok')
+        self.assert_test(events, number=1, name='', result=TestResult.FAIL)
+        self.assert_last(events)
+
+    def test_one_test_todo(self):
+        events = self.parse_tap('not ok 1 abc # TODO')
+        self.assert_test(events, number=1, name='abc', result=TestResult.EXPECTEDFAIL)
+        self.assert_last(events)
+
+        events = self.parse_tap('ok 1 abc # TODO')
+        self.assert_test(events, number=1, name='abc', result=TestResult.UNEXPECTEDPASS)
+        self.assert_last(events)
+
+    def test_one_test_skip(self):
+        events = self.parse_tap('ok 1 abc # SKIP')
+        self.assert_test(events, number=1, name='abc', result=TestResult.SKIP)
+        self.assert_last(events)
+
+    def test_one_test_skip_failure(self):
+        events = self.parse_tap('not ok 1 abc # SKIP')
+        self.assert_test(events, number=1, name='abc', result=TestResult.FAIL)
+        self.assert_last(events)
+
+    def test_many_early_plan(self):
+        events = self.parse_tap('1..4\nok 1\nnot ok 2\nok 3\nnot ok 4')
+        self.assert_plan(events, count=4, late=False)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_test(events, number=3, name='', result=TestResult.OK)
+        self.assert_test(events, number=4, name='', result=TestResult.FAIL)
+        self.assert_last(events)
+
+    def test_many_late_plan(self):
+        events = self.parse_tap('ok 1\nnot ok 2\nok 3\nnot ok 4\n1..4')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_test(events, number=3, name='', result=TestResult.OK)
+        self.assert_test(events, number=4, name='', result=TestResult.FAIL)
+        self.assert_plan(events, count=4, late=True)
+        self.assert_last(events)
+
+    def test_directive_case(self):
+        events = self.parse_tap('ok 1 abc # skip')
+        self.assert_test(events, number=1, name='abc', result=TestResult.SKIP)
+        self.assert_last(events)
+
+        events = self.parse_tap('ok 1 abc # ToDo')
+        self.assert_test(events, number=1, name='abc', result=TestResult.UNEXPECTEDPASS)
+        self.assert_last(events)
+
+    def test_directive_explanation(self):
+        events = self.parse_tap('ok 1 abc # skip why')
+        self.assert_test(events, number=1, name='abc', result=TestResult.SKIP,
+                         explanation='why')
+        self.assert_last(events)
+
+        events = self.parse_tap('ok 1 abc # ToDo Because')
+        self.assert_test(events, number=1, name='abc', result=TestResult.UNEXPECTEDPASS,
+                         explanation='Because')
+        self.assert_last(events)
+
+    def test_one_test_early_plan(self):
+        events = self.parse_tap('1..1\nok')
+        self.assert_plan(events, count=1, late=False)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_one_test_late_plan(self):
+        events = self.parse_tap('ok\n1..1')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_plan(events, count=1, late=True)
+        self.assert_last(events)
+
+    def test_out_of_order(self):
+        events = self.parse_tap('ok 2')
+        self.assert_error(events)
+        self.assert_test(events, number=2, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_middle_plan(self):
+        events = self.parse_tap('ok 1\n1..2\nok 2')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_plan(events, count=2, late=True)
+        self.assert_error(events)
+        self.assert_test(events, number=2, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_too_many_plans(self):
+        events = self.parse_tap('1..1\n1..2\nok 1')
+        self.assert_plan(events, count=1, late=False)
+        self.assert_error(events)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_too_many(self):
+        events = self.parse_tap('ok 1\nnot ok 2\n1..1')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_plan(events, count=1, late=True)
+        self.assert_error(events)
+        self.assert_last(events)
+
+        events = self.parse_tap('1..1\nok 1\nnot ok 2')
+        self.assert_plan(events, count=1, late=False)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_error(events)
+        self.assert_last(events)
+
+    def test_too_few(self):
+        events = self.parse_tap('ok 1\nnot ok 2\n1..3')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_plan(events, count=3, late=True)
+        self.assert_error(events)
+        self.assert_last(events)
+
+        events = self.parse_tap('1..3\nok 1\nnot ok 2')
+        self.assert_plan(events, count=3, late=False)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_error(events)
+        self.assert_last(events)
+
+    def test_too_few_bailout(self):
+        events = self.parse_tap('1..3\nok 1\nnot ok 2\nBail out! no third test')
+        self.assert_plan(events, count=3, late=False)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_bailout(events, message='no third test')
+        self.assert_last(events)
+
+    def test_diagnostics(self):
+        events = self.parse_tap('1..1\n# ignored\nok 1')
+        self.assert_plan(events, count=1, late=False)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+        events = self.parse_tap('# ignored\n1..1\nok 1\n# ignored too')
+        self.assert_plan(events, count=1, late=False)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+        events = self.parse_tap('# ignored\nok 1\n1..1\n# ignored too')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_plan(events, count=1, late=True)
+        self.assert_last(events)
+
+    def test_unexpected(self):
+        events = self.parse_tap('1..1\ninvalid\nok 1')
+        self.assert_plan(events, count=1, late=False)
+        self.assert_error(events)
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+    def test_version(self):
+        events = self.parse_tap('TAP version 13\n')
+        self.assert_version(events, version=13)
+        self.assert_last(events)
+
+        events = self.parse_tap('TAP version 12\n')
+        self.assert_error(events)
+        self.assert_last(events)
+
+        events = self.parse_tap('1..0\nTAP version 13\n')
+        self.assert_plan(events, count=0, late=False, skipped=True)
+        self.assert_error(events)
+        self.assert_last(events)
+
+    def test_yaml(self):
+        events = self.parse_tap_v13('ok\n ---\n foo: abc\n  bar: def\n ...\nok 2')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_test(events, number=2, name='', result=TestResult.OK)
+        self.assert_last(events)
+
+        events = self.parse_tap_v13('ok\n ---\n foo: abc\n  bar: def')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_error(events)
+        self.assert_last(events)
+
+        events = self.parse_tap_v13('ok 1\n ---\n foo: abc\n  bar: def\nnot ok 2')
+        self.assert_test(events, number=1, name='', result=TestResult.OK)
+        self.assert_error(events)
+        self.assert_test(events, number=2, name='', result=TestResult.FAIL)
+        self.assert_last(events)
+
 def unset_envs():
     # For unit tests we must fully control all command lines
     # so that there are no unexpected changes coming from the
@@ -5741,6 +6010,8 @@ def main():
     unset_envs()
     cases = ['InternalTests', 'DataTests', 'AllPlatformTests', 'FailureTests',
              'PythonTests', 'NativeFileTests', 'RewriterTests', 'CrossFileTests',
+             'TAPParserTests',
+
              'LinuxlikeTests', 'LinuxCrossArmTests', 'LinuxCrossMingwTests',
              'WindowsTests', 'DarwinTests']
 
diff --git a/test cases/common/212 tap tests/meson.build b/test cases/common/212 tap tests/meson.build
new file mode 100644
index 000000000..58529a72e
--- /dev/null
+++ b/test cases/common/212 tap tests/meson.build
@@ -0,0 +1,10 @@
+project('test features', 'c')
+
+tester = executable('tester', 'tester.c')
+test('pass', tester, args : ['ok'], protocol: 'tap')
+test('fail', tester, args : ['not ok'], should_fail: true, protocol: 'tap')
+test('xfail', tester, args : ['not ok # todo'], protocol: 'tap')
+test('xpass', tester, args : ['ok # todo'], should_fail: true, protocol: 'tap')
+test('skip', tester, args : ['ok # skip'], protocol: 'tap')
+test('skip failure', tester, args : ['not ok # skip'], should_fail: true, protocol: 'tap')
+test('no tests', tester, args : ['1..0 # skip'], protocol: 'tap')
diff --git a/test cases/common/212 tap tests/tester.c b/test cases/common/212 tap tests/tester.c
new file mode 100644
index 000000000..ac582e7c0
--- /dev/null
+++ b/test cases/common/212 tap tests/tester.c
@@ -0,0 +1,10 @@
+#include <stdio.h>
+
+int main(int argc, char **argv) {
+    if (argc != 2) {
+        fprintf(stderr, "Incorrect number of arguments, got %i\n", argc);
+        return 1;
+    }
+    puts(argv[1]);
+    return 0;
+}
diff --git a/test cases/failing test/4 hard error/main.c b/test cases/failing test/4 hard error/main.c
new file mode 100644
index 000000000..a1e705ade
--- /dev/null
+++ b/test cases/failing test/4 hard error/main.c
@@ -0,0 +1,3 @@
+int main(void) {
+    return 99;
+}
diff --git a/test cases/failing test/4 hard error/meson.build b/test cases/failing test/4 hard error/meson.build
new file mode 100644
index 000000000..6979b0416
--- /dev/null
+++ b/test cases/failing test/4 hard error/meson.build
@@ -0,0 +1,4 @@
+project('trivial', 'c')
+
+# Exit code 99 even overrides should_fail
+test('My Test', executable('main', 'main.c'), should_fail: true)
diff --git a/test cases/failing test/5 tap tests/meson.build b/test cases/failing test/5 tap tests/meson.build
new file mode 100644
index 000000000..844c1f990
--- /dev/null
+++ b/test cases/failing test/5 tap tests/meson.build
@@ -0,0 +1,6 @@
+project('test features', 'c')
+
+tester = executable('tester', 'tester.c')
+test('nonzero return code', tester, args : [], protocol: 'tap')
+test('missing test', tester, args : ['1..1'], protocol: 'tap')
+test('incorrect skip', tester, args : ['1..1 # skip\nok 1'], protocol: 'tap')
diff --git a/test cases/failing test/5 tap tests/tester.c b/test cases/failing test/5 tap tests/tester.c
new file mode 100644
index 000000000..ac582e7c0
--- /dev/null
+++ b/test cases/failing test/5 tap tests/tester.c
@@ -0,0 +1,10 @@
+#include <stdio.h>
+
+int main(int argc, char **argv) {
+    if (argc != 2) {
+        fprintf(stderr, "Incorrect number of arguments, got %i\n", argc);
+        return 1;
+    }
+    puts(argv[1]);
+    return 0;
+}