Introduce qt-testrunner.py

Script that wraps Qt test execution in order to iron-out flakiness.

Task-number: QTBUG-96353
Change-Id: Ie8ee780e3f4f8d086c080a7784f9f68fd1867be2
Reviewed-by: Daniel Smith <Daniel.Smith@qt.io>

8 files changed, 887 insertions, 43 deletions

diff --git a/cmake/QtBaseGlobalTargets.cmake b/cmake/QtBaseGlobalTargets.cmake
index bddc597bdb..c8499e3921 100644
--- a/cmake/QtBaseGlobalTargets.cmake
+++ b/cmake/QtBaseGlobalTargets.cmake
@@ -351,3 +351,5 @@ endif()
 qt_path_join(__qt_libexec_install_dir "${QT_INSTALL_DIR}" "${INSTALL_LIBEXECDIR}")
 qt_copy_or_install(FILES coin/instructions/qmake/ensure_pro_file.cmake
     DESTINATION "${__qt_libexec_install_dir}")
+qt_copy_or_install(PROGRAMS "util/testrunner/qt-testrunner.py"
+                   DESTINATION "${__qt_libexec_install_dir}")
diff --git a/cmake/QtWrapperScriptHelpers.cmake b/cmake/QtWrapperScriptHelpers.cmake
index cee9bfd1ed..e7a3b3c4bd 100644
--- a/cmake/QtWrapperScriptHelpers.cmake
+++ b/cmake/QtWrapperScriptHelpers.cmake
@@ -3,6 +3,7 @@
 # qt-cmake-private
 # qt-configure-module
 # qt-cmake-private-install
+# And other helper scripts.
 function(qt_internal_create_wrapper_scripts)
     # Provide a convenience cmake wrapper.
     if(CMAKE_HOST_UNIX)
diff --git a/coin/instructions/cmake_run_ctest_enforce_exit_code.yaml b/coin/instructions/cmake_run_ctest_enforce_exit_code.yaml
index fecf283905..c2560be37e 100644
--- a/coin/instructions/cmake_run_ctest_enforce_exit_code.yaml
+++ b/coin/instructions/cmake_run_ctest_enforce_exit_code.yaml
@@ -10,19 +10,14 @@ instructions:
       equals_value: Windows
   - type: EnvironmentVariable
     variableName: TESTRUNNER
-    variableValue: "python3 {{.SourceDir}}/coin_ctest_runner.py"
+    variableValue: "{{.InstallDir}}/libexec/qt-testrunner.py --"
     disable_if:
-      condition: or
-      conditions:
-        - condition: property
-          property: host.os
-          equals_value: Windows
-        - condition: property
-          property: features
-          contains_value: AndroidTestRun
+      condition: property
+      property: host.os
+      equals_value: Windows
   - type: EnvironmentVariable
     variableName: TESTRUNNER
-    variableValue: "{{.SourceDir}}\\coin_ctest_runner.py"
+    variableValue: "{{.InstallDir}}\\bin\\qt-testrunner.py --"
     enable_if:
       condition: property
       property: host.os
@@ -43,7 +38,7 @@ instructions:
       equals_value: Windows
   - type: EnvironmentVariable
     variableName: CTEST_ARGS
-    variableValue: "-V --rerun-failed --force-new-ctest-process --repeat until-pass:5"
+    variableValue: "-V"
   - type: AppendToEnvironmentVariable
     variableName: CTEST_ARGS
     variableValue: " --stop-on-failure"
diff --git a/coin/instructions/cmake_setup_running_tests_env_vars.yaml b/coin/instructions/cmake_setup_running_tests_env_vars.yaml
index c1a83eae07..b9387d74c6 100644
--- a/coin/instructions/cmake_setup_running_tests_env_vars.yaml
+++ b/coin/instructions/cmake_setup_running_tests_env_vars.yaml
@@ -5,38 +5,6 @@ instructions:
   - type: EnvironmentVariable
     variableName: QTEST_ENVIRONMENT
     variableValue: "ci"
-  - type: WriteFile
-    fileContents: |
-        #!/usr/bin/python3
-        import subprocess
-        import calendar
-        import datetime
-        import time
-        import sys
-        import os
-        import re
-        from os.path import expanduser
-        home = expanduser("~")
-
-        file=os.path.basename(sys.argv[1])
-        timestamp = str(round(time.time() * 1000))
-        results_file = home + "/work/testresults/" + file +"-" + timestamp + ".xml,xml"
-        testargs = ["-o", results_file, "-o", "-,txt"]
-        if re.search("testlib.selftests.tst_selftests", sys.argv[1]):
-            testargs = []
-        exit(subprocess.call([sys.argv[1]] + testargs))
-    filename: "{{.SourceDir}}/coin_ctest_runner.py"
-    fileMode: 755
-  - type: ExecuteCommand
-    command: "chmod 755 {{.SourceDir}}/coin_ctest_runner.py"
-    maxTimeInSeconds: 10
-    maxTimeBetweenOutput: 10
-    userMessageOnFailure: >
-      Failed to change file permission.
-    disable_if:
-      condition: property
-      property: host.os
-      equals_value: Windows
   - type: Group
     instructions:
       - type: EnvironmentVariable
diff --git a/util/testrunner/qt-testrunner.py b/util/testrunner/qt-testrunner.py
new file mode 100755
index 0000000000..3f29e9d1d9
--- /dev/null
+++ b/util/testrunner/qt-testrunner.py
@@ -0,0 +1,348 @@
+#!/usr/bin/env python3
+
+
+#############################################################################
+##
+## Copyright (C) 2021 The Qt Company Ltd.
+## Contact: https://www.qt.io/licensing/
+##
+## This file is part of the release tools of the Qt Toolkit.
+##
+## $QT_BEGIN_LICENSE:GPL-EXCEPT$
+## Commercial License Usage
+## Licensees holding valid commercial Qt licenses may use this file in
+## accordance with the commercial license agreement provided with the
+## Software or, alternatively, in accordance with the terms contained in
+## a written agreement between you and The Qt Company. For licensing terms
+## and conditions see https://www.qt.io/terms-conditions. For further
+## information use the contact form at https://www.qt.io/contact-us.
+##
+## GNU General Public License Usage
+## Alternatively, this file may be used under the terms of the GNU
+## General Public License version 3 as published by the Free Software
+## Foundation with exceptions as appearing in the file LICENSE.GPL3-EXCEPT
+## included in the packaging of this file. Please review the following
+## information to ensure the GNU General Public License requirements will
+## be met: https://www.gnu.org/licenses/gpl-3.0.html.
+##
+## $QT_END_LICENSE$
+##
+#############################################################################
+
+
+# !!!IMPORTANT!!!  If you change anything to this script, run the testsuite
+#    manually and make sure it still passes, as it doesn't run automatically.
+#    Just execute the command line as such:
+#
+#      ./util/testrunner/tests/tst_testrunner.py -v [--debug]
+#
+# This script wraps the execution of a Qt test executable, for example
+# tst_whatever, and tries to iron out unpredictable test failures.
+# In particular:
+#
+# + Appends output argument to it: "-o tst_whatever.xml,xml"
+# + Checks the exit code. If it is zero, the script exits with zero,
+#   otherwise proceeds.
+# + Reads the XML test log and Understands exactly which function
+#   of the test failed.
+#   + If no XML file is found or was invalid, the test executable
+#     probably CRASHed, so we *re-run the full test once again*.
+# + If some testcases failed it executes only those individually
+#   until they pass, or until max-repeats times is reached.
+#
+# The regular way to use is to set the environment variable TESTRUNNER to
+# point to this script before invoking ctest.
+#
+# NOTE: this script is crafted specifically for use with Qt tests and for
+#       using it in Qt's CI. For example it detects and acts specially if test
+#       executable is "tst_selftests" or "androidtestrunner".  It also detects
+#       env var "COIN_CTEST_RESULTSDIR" and uses it as log-dir.
+#
+# TODO implement --dry-run.
+
+# Exit codes of this script:
+#   0: PASS. Either no test failed, or failed initially but passed
+#      in the re-runs (FLAKY PASS).
+#   1: Some unexpected error of this script.
+#   2: FAIL! for at least one test, even after the re-runs.
+#   3: CRASH! for the test executable even after re-running it once.
+
+
+
+import sys
+if sys.version_info < (3, 6):
+    sys.stderr.write(
+        "Error: this test wrapper script requires Python version 3.6 at least\n")
+    sys.exit(1)
+
+import argparse
+import subprocess
+import os
+import traceback
+import timeit
+import xml.etree.ElementTree as ET
+import logging as L
+
+from pprint import pprint
+from typing import NamedTuple, Tuple, List, Optional
+
+# Define a custom type for returning a fail incident
+class WhatFailed(NamedTuple):
+    func: str
+    tag: Optional[str] = None
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(formatter_class=argparse.RawDescriptionHelpFormatter,
+        description="""
+Wrap Qt test execution. This is intended to be invoked via the TESTRUNNER
+environment variable before running ctest in the CI environment. The purpose
+of the script is to repeat failed tests in order to iron out transient errors
+caused by unpredictable factors. Individual test functions that failed are
+retried up to max-repeats times until the test passes.
+        """,
+        epilog="""
+Default flags: --max-repeats 5 --passes-needed 1
+        """
+    )
+    parser.add_argument("testargs", metavar="TESTARGS", nargs="+",
+                        help="Test executable and arguments")
+    parser.add_argument("--log-dir", metavar="DIR",
+                        help="Where to write the XML log files with the test results of the primary test run;"
+                        " by default write to CWD")
+    parser.add_argument("--max-repeats", type=int, default=5, metavar='N',
+                        help="In case the test FAILs, repeat the failed cases this many times")
+    parser.add_argument("--passes-needed", type=int, default=1, metavar='M',
+                        help="Number of repeats that need to succeed in order to return an overall PASS")
+    parser.add_argument("--parse-xml-testlog", metavar="file.xml",
+                        help="Do not run the full test the first time, but parse this XML test log;"
+                        " if the test log contains failures, then re-run the failed cases normally,"
+                        " as indicated by the other flags")
+    parser.add_argument("--dry-run", action="store_true",
+                        help="(TODO - not implemented yet) Do not run anything, just describe what would happen")
+    parser.add_argument("--timeout", metavar="T",
+                        help="Timeout for each test execution in seconds")
+    parser.add_argument("--no-extra-args", action="store_true",
+                        help="Do not append any extra arguments to the test command line, like"
+                        " -o log_file.xml -v2 -vs. This will disable some functionality like the"
+                        " failed test repetition and the verbose output on failure. This is"
+                        " activated by default when TESTARGS is tst_selftests.")
+    args = parser.parse_args()
+    args.self_name = os.path.basename(sys.argv[0])
+    args.specific_extra_args = []
+
+    logging_format = args.self_name + " %(levelname)8s: %(message)s"
+    L.basicConfig(format=logging_format, level=L.DEBUG)
+
+    if args.log_dir is None:
+        if "COIN_CTEST_RESULTSDIR" in os.environ:
+            args.log_dir = os.environ["COIN_CTEST_RESULTSDIR"]
+            L.info("Will write XML test logs to directory"
+                   " COIN_CTEST_RESULTSDIR=%s", args.log_dir)
+        else:
+            args.log_dir = "."
+
+    args.test_basename = os.path.basename(args.testargs[0])
+    if args.test_basename.endswith(".exe"):
+        args.test_basename = args.test_basename[:-4]
+
+    # On Android emulated platforms, "androidtestrunner" is invoked by CMake
+    # to wrap the tests.  We have to append the test arguments to it after
+    # "--". Besides that we have to detect the basename to avoid saving the
+    # XML log as "androidtestrunner.xml" for all tests.
+    if args.test_basename == "androidtestrunner":
+        args.specific_extra_args = [ "--" ]
+        apk_arg = False
+        for a in args.testargs[1:]:
+            if a == "--apk":
+                apk_arg = True
+            elif apk_arg:
+                apk_arg = False
+                if a.endswith(".apk"):
+                    args.test_basename = os.path.basename(a)[:-4]
+                    break
+        L.info("Detected androidtestrunner, test will be handled specially. Detected test basename: %s",
+               args.test_basename)
+
+    # The qtestlib selftests are implemented using an external test library
+    # (Catch), and they don't support the same command-line options.
+    if args.test_basename == "tst_selftests":
+        L.info("Detected special test not able to generate XML log! Will not repeat individual testcases.")
+        args.no_extra_args = True
+        args.max_repeats = 0
+
+    return args
+
+
+def parse_log(results_file) -> List[WhatFailed]:
+    """Parse the XML test log file. Return the failed testcases, if any.
+
+    Failures are considered the "fail" and "xpass" incidents.
+    A testcase is a function with an optional data tag."""
+    start_timer = timeit.default_timer()
+
+    try:
+        tree = ET.parse(results_file)
+    except FileNotFoundError:
+        L.error("XML log file not found: %s", results_file)
+        raise
+    except ET.ParseError:
+        L.error("Failed to parse the XML log file: %s", results_file)
+        with open(results_file, "rb") as f:
+            L.error("File Contents:\n%s\n\n", f.read().decode("utf-8", "ignore"))
+        raise
+
+    root = tree.getroot()
+    if root.tag != "TestCase":
+        raise AssertionError(
+            f"The XML test log must have <TestCase> as root tag, but has: <{root.tag}>")
+
+    failures = []
+    n_passes = 0
+    for e1 in root:
+        if e1.tag == "TestFunction":
+            for e2 in e1:                          # every <TestFunction> can have many <Incident>
+                if e2.tag == "Incident":
+                    if e2.attrib["type"] in ("fail", "xpass"):
+                        func = e1.attrib["name"]
+                        e3 = e2.find("DataTag")    # every <Incident> might have a <DataTag>
+                        if e3 is not None:
+                            failures.append(WhatFailed(func, tag=e3.text))
+                        else:
+                            failures.append(WhatFailed(func))
+                    else:
+                        n_passes += 1
+
+    end_timer = timeit.default_timer()
+    t = end_timer - start_timer
+    L.info(f"Parsed XML file {results_file} in {t:.3f} seconds")
+    L.info(f"Found {n_passes} passes and {len(failures)} failures")
+
+    return failures
+
+
+def run_test(arg_list: List[str], timeout=None):
+    L.debug("Running test command line: %s", arg_list)
+    proc = subprocess.run(arg_list, timeout=timeout)
+    L.info("Test process exited with code: %d", proc.returncode)
+
+    return proc
+
+# Returns tuple: (exit_code, xml_logfile)
+def run_full_test(test_basename, testargs: List[str], output_dir: str,
+                  no_extra_args=False, dryrun=False,
+                  timeout=None, specific_extra_args=[])  \
+        -> Tuple[int, Optional[str]]:
+
+    results_files = []
+    output_testargs = []
+
+    # Append arguments to write log to qtestlib XML file,
+    # to JUnit XML file, and text to stdout.
+    if not no_extra_args:
+        results_files.append(os.path.join(output_dir, test_basename + ".xml"))
+        results_files.append(os.path.join(output_dir, test_basename + "-junitxml.xml"))
+        output_testargs.extend(["-o", results_files[0] + ",xml"])
+        output_testargs.extend(["-o", results_files[1] + ",junitxml"])
+        output_testargs.extend(["-o", "-,txt"])
+
+    proc = run_test(testargs + specific_extra_args + output_testargs, timeout)
+
+    return (proc.returncode, results_files[0] if results_files else None)
+
+
+# TODO alter environment for logging:
+# QT_LOGGING_RULES="*=true"
+# QT_MESSAGE_PATTERN="[%{time process} %{if-debug}D%{endif}%{if-warning}W%{endif}%{if-critical}C%{endif}%{if-fatal}F%{endif}] %{category} %{file}:%{line} %{function}()  -  %{message}"
+# add arg: -maxwarnings 0 (maybe -v2 -vs?)
+def rerun_failed_testcase(testargs: List[str], what_failed: WhatFailed,
+                          max_repeats, passes_needed,
+                          dryrun=False, timeout=None) -> bool:
+    """Run a specific function:tag of a test, until it passes enough times, or
+    until max_repeats is reached.
+
+    Return True if it passes eventually, False if it fails.
+    """
+    assert passes_needed <= max_repeats
+    failed_arg = what_failed.func
+    if what_failed.tag:
+        failed_arg += ":" + what_failed.tag
+
+    n_passes = 0
+    for i in range(max_repeats):
+        L.info("Re-running testcase: %s", failed_arg)
+        proc = run_test(testargs + [failed_arg], timeout)
+        if proc.returncode == 0:
+            n_passes += 1
+        if n_passes == passes_needed:
+            L.info("Test has PASSed as FLAKY after re-runs:%d, passes:%d, failures:%d",
+                   i+1, n_passes, i+1-n_passes)
+            return True
+
+    assert n_passes <  passes_needed
+    assert n_passes <= max_repeats
+    n_failures = max_repeats - n_passes
+    L.info("Test has FAILed despite all repetitions! re-runs:%d failures:%d",
+           max_repeats, n_failures)
+    return False
+
+
+def main():
+    args = parse_args()
+    n_full_runs = 1 if args.parse_xml_testlog else 2
+
+    for i in range(n_full_runs):
+        try:
+            if i != 0:
+                L.info("Re-running the full test!")
+            if args.parse_xml_testlog:
+                retcode = 1    # pretend the test returned error
+                results_file = args.parse_xml_testlog
+            else:
+                (retcode, results_file) = \
+                    run_full_test(args.test_basename, args.testargs, args.log_dir,
+                                  args.no_extra_args, args.dry_run, args.timeout,
+                                  args.specific_extra_args)
+                if retcode == 0:
+                    sys.exit(0)    # PASS
+
+            failed_functions = parse_log(results_file)
+
+            if not args.parse_xml_testlog:
+                assert len(failed_functions) > 0, \
+                    "The XML test log should contain at least one failure!"
+
+            break    # go to re-running individual failed testcases
+
+        except Exception as e:
+            L.exception("Uncontrolled test CRASH! Details:", exc_info=e)
+            if i < n_full_runs - 1:
+                L.info("Will re-run the full test executable again!")
+            else:    # Failed on the final run
+                L.error("Full test run failed repeatedly, aborting!")
+                sys.exit(3)
+
+    if args.max_repeats == 0:
+        sys.exit(2)          # Some tests failed but no re-runs were asked
+
+    L.info("Some tests failed, will re-run at most %d times.\n",
+           args.max_repeats)
+
+    for what_failed in failed_functions:
+        try:
+            ret = rerun_failed_testcase(args.testargs, what_failed,
+                                        args.max_repeats, args.passes_needed,
+                                        dryrun=args.dry_run, timeout=args.timeout)
+        except Exception as e:
+            L.exception("Uncontrolled test CRASH! Details:", exc_info=e)
+            L.error("Test re-run exited unxpectedly, aborting!")
+            sys.exit(3)                                    # Test re-run CRASH
+
+        if not ret:
+            sys.exit(2)                                    # Test re-run FAIL
+
+    sys.exit(0)                                  # All testcase re-runs PASSed
+
+
+if __name__ == "__main__":
+    main()
diff --git a/util/testrunner/tests/qt_mock_test-log.xml b/util/testrunner/tests/qt_mock_test-log.xml
new file mode 100644
index 0000000000..0c316d71c3
--- /dev/null
+++ b/util/testrunner/tests/qt_mock_test-log.xml
@@ -0,0 +1,32 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<TestCase name="qt_mock_test">
+  <Environment>
+    <QtVersion>MOCK</QtVersion>
+    <QtBuild>MOCK</QtBuild>
+    <QTestVersion>6.3.0</QTestVersion>
+  </Environment>
+  <TestFunction name="always_pass">
+    <Incident type="{{always_pass_result}}" file="" line="0" />
+    <Duration msecs="0.71704"/>
+  </TestFunction>
+  <TestFunction name="always_fail">
+    <Incident type="{{always_fail_result}}" file="" line="0" />
+    <Duration msecs="0.828272"/>
+  </TestFunction>
+  <TestFunction name="always_crash">
+    <Incident type="{{always_crash_result}}" file="" line="0" />
+    <Duration msecs="0.828272"/>
+  </TestFunction>
+  <TestFunction name="fail_then_pass">
+    <Incident type="{{fail_then_pass:2_result}}" file="" line="0">
+      <DataTag><![CDATA[2]]></DataTag>
+    </Incident>
+    <Incident type="{{fail_then_pass:5_result}}" file="" line="0">
+      <DataTag><![CDATA[5]]></DataTag>
+    </Incident>
+    <Incident type="{{fail_then_pass:6_result}}" file="" line="0">
+      <DataTag><![CDATA[6]]></DataTag>
+    </Incident>
+  </TestFunction>
+  <Duration msecs="1904.9"/>
+</TestCase>
diff --git a/util/testrunner/tests/qt_mock_test.py b/util/testrunner/tests/qt_mock_test.py
new file mode 100755
index 0000000000..e170f3a6c0
--- /dev/null
+++ b/util/testrunner/tests/qt_mock_test.py
@@ -0,0 +1,203 @@
+#!/usr/bin/env python3
+
+
+#############################################################################
+##
+## Copyright (C) 2021 The Qt Company Ltd.
+## Contact: https://www.qt.io/licensing/
+##
+## This file is part of the release tools of the Qt Toolkit.
+##
+## $QT_BEGIN_LICENSE:GPL-EXCEPT$
+## Commercial License Usage
+## Licensees holding valid commercial Qt licenses may use this file in
+## accordance with the commercial license agreement provided with the
+## Software or, alternatively, in accordance with the terms contained in
+## a written agreement between you and The Qt Company. For licensing terms
+## and conditions see https://www.qt.io/terms-conditions. For further
+## information use the contact form at https://www.qt.io/contact-us.
+##
+## GNU General Public License Usage
+## Alternatively, this file may be used under the terms of the GNU
+## General Public License version 3 as published by the Free Software
+## Foundation with exceptions as appearing in the file LICENSE.GPL3-EXCEPT
+## included in the packaging of this file. Please review the following
+## information to ensure the GNU General Public License requirements will
+## be met: https://www.gnu.org/licenses/gpl-3.0.html.
+##
+## $QT_END_LICENSE$
+##
+#############################################################################
+
+
+# This is an artificial test, mimicking the Qt tests, for example tst_whatever.
+# Its purpose is to assist in testing qt-testrunner.py.
+#
+# Mode A:
+#
+# If invoked with a test function argument, it runs that test function.
+#
+# Usage:
+#
+# $0 always_pass
+# $0 always_fail
+# $0 always_crash
+# $0 fail_then_pass:N   # where N is the number of failing runs before passing
+#
+# Needs environment variable:
+#   + QT_MOCK_TEST_STATE_FILE :: points to a unique filename, to be written
+#     for keeping the state of the fail_then_pass:N tests.
+#
+# Mode B:
+#
+# If invoked without any argument, it runs the tests listed in the
+# variable QT_MOCK_TEST_FAIL_LIST. If variable is empty it just runs
+# the always_pass test. It also understands qtestlib's `-o outfile.xml,xml`
+# option for writing a mock testlog in a file. Requires environment variables:
+#   + QT_MOCK_TEST_STATE_FILE :: See above
+#   + QT_MOCK_TEST_XML_TEMPLATE_FILE :: may point to the template XML file
+#     located in the same source directory. Without this variable, the
+#     option `-o outfile.xml,xml` will be ignored.
+#   + QT_MOCK_TEST_FAIL_LIST :: may contain a comma-separated list of test
+#     that should run.
+
+
+import sys
+import os
+import traceback
+from tst_testrunner import write_xml_log
+
+
+MY_NAME         = os.path.basename(sys.argv[0])
+STATE_FILE      = None
+XML_TEMPLATE    = None
+XML_OUTPUT_FILE = None
+
+
+def put_failure(test_name):
+    with open(STATE_FILE, "a") as f:
+        f.write(test_name + "\n")
+def get_failures(test_name):
+    n = 0
+    try:
+        with open(STATE_FILE) as f:
+            for line in f:
+                if line.strip() == test_name:
+                    n += 1
+    except FileNotFoundError:
+        return 0
+    return n
+
+# Only care about the XML log output file.
+def parse_output_argument(a):
+    global XML_OUTPUT_FILE
+    if a.endswith(",xml"):
+        XML_OUTPUT_FILE = a[:-4]
+
+# Strip qtestlib specific arguments.
+# Only care about the "-o ...,xml" argument.
+def clean_cmdline():
+    args = []
+    prev_arg = None
+    skip_next_arg = True    # Skip argv[0]
+    for a in sys.argv:
+        if skip_next_arg:
+            if prev_arg == "-o":
+                parse_output_argument(a)
+            prev_arg = None
+            skip_next_arg = False
+            continue
+        if a in ("-o", "-maxwarnings"):
+            skip_next_arg = True
+            prev_arg = a
+            continue
+        if a in ("-v1", "-v2", "-vs"):
+            continue
+        args.append(a)
+    return args
+
+
+def log_test(testcase, result,
+             testsuite=MY_NAME.rpartition(".")[0]):
+    print("%-7s: %s::%s()" % (result, testsuite, testcase))
+
+# Return the exit code
+def run_test(testname):
+    if   testname == "always_pass":
+        exit_code = 0
+    elif testname == "always_fail":
+        exit_code = 1
+    elif testname == "always_crash":
+        exit_code = 130
+    elif testname.startswith("fail_then_pass"):
+        wanted_fails   = int(testname.partition(":")[2])
+        previous_fails = get_failures(testname)
+        if previous_fails < wanted_fails:
+            put_failure(testname)
+            exit_code = 1
+        else:
+            exit_code = 0
+    else:
+        assert False, "Unknown argument: %s" % testname
+
+    if exit_code == 0:
+        log_test(testname, "PASS")
+    elif exit_code == 1:
+        log_test(testname, "FAIL!")
+    else:
+        log_test(testname, "CRASH!")
+
+    return exit_code
+
+def no_args_run():
+    try:
+        run_list = os.environ["QT_MOCK_TEST_RUN_LIST"].split(",")
+    except KeyError:
+        run_list = ["always_pass"]
+
+    total_result = True
+    fail_list = []
+    for test in run_list:
+        test_exit_code = run_test(test)
+        if test_exit_code not in (0, 1):
+            sys.exit(130)                                       # CRASH!
+        if test_exit_code != 0:
+            fail_list.append(test)
+        total_result = total_result and (test_exit_code == 0)
+
+    if XML_TEMPLATE and XML_OUTPUT_FILE:
+        write_xml_log(XML_OUTPUT_FILE, failure=fail_list)
+
+    if total_result:
+        sys.exit(0)
+    else:
+        sys.exit(1)
+
+
+def main():
+    global STATE_FILE
+    # Will fail if env var is not set.
+    STATE_FILE = os.environ["QT_MOCK_TEST_STATE_FILE"]
+
+    global XML_TEMPLATE
+    if "QT_MOCK_TEST_XML_TEMPLATE_FILE" in os.environ:
+        with open(os.environ["QT_MOCK_TEST_XML_TEMPLATE_FILE"]) as f:
+            XML_TEMPLATE = f.read()
+
+    args = clean_cmdline()
+
+    if len(args) == 0:
+        no_args_run()
+        assert False, "Unreachable!"
+    else:
+        sys.exit(run_test(args[0]))
+
+
+# TODO write XPASS test that does exit(1)
+
+if __name__ == "__main__":
+    try:
+        main()
+    except Exception as e:
+        traceback.print_exc()
+        exit(128)                      # Something went wrong with this script
diff --git a/util/testrunner/tests/tst_testrunner.py b/util/testrunner/tests/tst_testrunner.py
new file mode 100755
index 0000000000..02a839409f
--- /dev/null
+++ b/util/testrunner/tests/tst_testrunner.py
@@ -0,0 +1,295 @@
+#!/usr/bin/env python3
+
+
+#############################################################################
+##
+## Copyright (C) 2021 The Qt Company Ltd.
+## Contact: https://www.qt.io/licensing/
+##
+## This file is part of the release tools of the Qt Toolkit.
+##
+## $QT_BEGIN_LICENSE:GPL-EXCEPT$
+## Commercial License Usage
+## Licensees holding valid commercial Qt licenses may use this file in
+## accordance with the commercial license agreement provided with the
+## Software or, alternatively, in accordance with the terms contained in
+## a written agreement between you and The Qt Company. For licensing terms
+## and conditions see https://www.qt.io/terms-conditions. For further
+## information use the contact form at https://www.qt.io/contact-us.
+##
+## GNU General Public License Usage
+## Alternatively, this file may be used under the terms of the GNU
+## General Public License version 3 as published by the Free Software
+## Foundation with exceptions as appearing in the file LICENSE.GPL3-EXCEPT
+## included in the packaging of this file. Please review the following
+## information to ensure the GNU General Public License requirements will
+## be met: https://www.gnu.org/licenses/gpl-3.0.html.
+##
+## $QT_END_LICENSE$
+##
+#############################################################################
+
+
+import sys
+import os
+import re
+import subprocess
+
+from subprocess import STDOUT, PIPE
+from tempfile   import TemporaryDirectory, mkstemp
+
+MY_NAME          = os.path.basename(__file__)
+my_dir           = os.path.dirname(__file__)
+testrunner       = os.path.join(my_dir, "..", "qt-testrunner.py")
+mock_test        = os.path.join(my_dir, "qt_mock_test.py")
+xml_log_template = os.path.join(my_dir, "qt_mock_test-log.xml")
+
+with open(xml_log_template) as f:
+    XML_TEMPLATE = f.read()
+
+
+import unittest
+
+def setUpModule():
+    global TEMPDIR
+    TEMPDIR = TemporaryDirectory(prefix="tst_testrunner-")
+
+    filename = os.path.join(TEMPDIR.name, "file_1")
+    print("setUpModule(): setting up temporary directory and env var"
+          " QT_MOCK_TEST_STATE_FILE=" + filename + " and"
+          " QT_MOCK_TEST_XML_TEMPLATE_FILE=" + xml_log_template)
+
+    os.environ["QT_MOCK_TEST_STATE_FILE"]        = filename
+    os.environ["QT_MOCK_TEST_XML_TEMPLATE_FILE"] = xml_log_template
+
+def tearDownModule():
+    print("\ntearDownModule(): Cleaning up temporary directory:",
+          TEMPDIR.name)
+    del os.environ["QT_MOCK_TEST_STATE_FILE"]
+    TEMPDIR.cleanup()
+
+
+# Helper to run a command and always capture output
+def run(*args, **kwargs):
+    if DEBUG:
+        print("Running: ", args, flush=True)
+    proc = subprocess.run(*args, stdout=PIPE, stderr=STDOUT, **kwargs)
+    if DEBUG and proc.stdout:
+        print(proc.stdout.decode(), flush=True)
+    return proc
+
+# Helper to run qt-testrunner.py with proper testing arguments.
+def run_testrunner(xml_filename=None, extra_args=None, env=None):
+
+    args = [ testrunner, mock_test ]
+    if xml_filename:
+        args += [ "--parse-xml-testlog", xml_filename ]
+    if extra_args:
+        args += extra_args
+
+    return run(args, env=env)
+
+# Write the XML_TEMPLATE to filename, replacing the templated results.
+def write_xml_log(filename, failure=None):
+    data = XML_TEMPLATE
+    # Replace what was asked to fail with "fail"
+    if type(failure) in (list, tuple):
+        for template in failure:
+            data = data.replace("{{"+template+"_result}}", "fail")
+    elif type(failure) is str:
+        data = data.replace("{{"+failure+"_result}}", "fail")
+    # Replace the rest with "pass"
+    data = re.sub(r"{{[^}]+}}", "pass", data)
+    with open(filename, "w") as f:
+        f.write(data)
+
+
+# Test that qt_mock_test.py behaves well. This is necessary to properly
+# test qt-testrunner.
+class Test_qt_mock_test(unittest.TestCase):
+    def setUp(self):
+        state_file = os.environ["QT_MOCK_TEST_STATE_FILE"]
+        if os.path.exists(state_file):
+            os.remove(state_file)
+    def test_always_pass(self):
+        proc = run([mock_test, "always_pass"])
+        self.assertEqual(proc.returncode, 0)
+    def test_always_fail(self):
+        proc = run([mock_test, "always_fail"])
+        self.assertEqual(proc.returncode, 1)
+    def test_fail_then_pass_2(self):
+        proc = run([mock_test, "fail_then_pass:2"])
+        self.assertEqual(proc.returncode, 1)
+        proc = run([mock_test, "fail_then_pass:2"])
+        self.assertEqual(proc.returncode, 1)
+        proc = run([mock_test, "fail_then_pass:2"])
+        self.assertEqual(proc.returncode, 0)
+    def test_fail_then_pass_1(self):
+        proc = run([mock_test, "fail_then_pass:1"])
+        self.assertEqual(proc.returncode, 1)
+        proc = run([mock_test, "fail_then_pass:1"])
+        self.assertEqual(proc.returncode, 0)
+    def test_fail_then_pass_many_tests(self):
+        proc = run([mock_test, "fail_then_pass:1"])
+        self.assertEqual(proc.returncode, 1)
+        proc = run([mock_test, "fail_then_pass:2"])
+        self.assertEqual(proc.returncode, 1)
+        proc = run([mock_test, "fail_then_pass:1"])
+        self.assertEqual(proc.returncode, 0)
+        proc = run([mock_test, "fail_then_pass:2"])
+        self.assertEqual(proc.returncode, 1)
+        proc = run([mock_test, "fail_then_pass:2"])
+        self.assertEqual(proc.returncode, 0)
+    def test_xml_file_is_written(self):
+        filename = os.path.join(TEMPDIR.name, "testlog.xml")
+        proc = run([mock_test, "-o", filename+",xml"])
+        self.assertEqual(proc.returncode, 0)
+        self.assertTrue(os.path.exists(filename))
+        self.assertGreater(os.path.getsize(filename), 0)
+        os.remove(filename)
+
+# Test regular invocations of qt-testrunner.
+class Test_testrunner(unittest.TestCase):
+    def setUp(self):
+        state_file = os.environ["QT_MOCK_TEST_STATE_FILE"]
+        if os.path.exists(state_file):
+            os.remove(state_file)
+        old_logfile = os.path.join(TEMPDIR.name, os.path.basename(mock_test) + ".xml")
+        if os.path.exists(old_logfile):
+            os.remove(old_logfile)
+        self.env = dict()
+        self.env["QT_MOCK_TEST_XML_TEMPLATE_FILE"] = os.environ["QT_MOCK_TEST_XML_TEMPLATE_FILE"]
+        self.env["QT_MOCK_TEST_STATE_FILE"]        = state_file
+        self.extra_args = [ "--log-dir", TEMPDIR.name ]
+    def prepare_env(self, run_list=None):
+        if run_list is not None:
+            self.env['QT_MOCK_TEST_RUN_LIST'] = ",".join(run_list)
+    def run2(self):
+        return run_testrunner(extra_args=self.extra_args, env=self.env)
+    def test_simple_invocation(self):
+        # All tests pass.
+        proc = self.run2()
+        self.assertEqual(proc.returncode, 0)
+    def test_always_pass(self):
+        self.prepare_env(run_list=["always_pass"])
+        proc = self.run2()
+        self.assertEqual(proc.returncode, 0)
+    def test_always_fail(self):
+        self.prepare_env(run_list=["always_fail"])
+        proc = self.run2()
+        # TODO verify that re-runs==max_repeats
+        self.assertEqual(proc.returncode, 2)
+    def test_flaky_pass_1(self):
+        self.prepare_env(run_list=["always_pass,fail_then_pass:1"])
+        proc = self.run2()
+        self.assertEqual(proc.returncode, 0)
+    def test_flaky_pass_5(self):
+        self.prepare_env(run_list=["always_pass,fail_then_pass:1,fail_then_pass:5"])
+        proc = self.run2()
+        self.assertEqual(proc.returncode, 0)
+    def test_flaky_fail(self):
+        self.prepare_env(run_list=["always_pass,fail_then_pass:6"])
+        proc = self.run2()
+        self.assertEqual(proc.returncode, 2)
+    def test_flaky_pass_fail(self):
+        self.prepare_env(run_list=["always_pass,fail_then_pass:1,fail_then_pass:6"])
+        proc = self.run2()
+        # TODO verify that one func was re-run and passed but the other failed.
+        self.assertEqual(proc.returncode, 2)
+
+    # If no XML file is found by qt-testrunner, it is usually considered a
+    # CRASH and the whole test is re-run. But when the return code is zero, it
+    # doesn't care about XML file and passes anyway.
+    def test_no_xml_log_written_pass(self):
+        del self.env["QT_MOCK_TEST_XML_TEMPLATE_FILE"]
+        self.prepare_env(run_list=["always_pass"])
+        proc = self.run2()
+        self.assertEqual(proc.returncode, 0)
+    # On the 2nd iteration of the full test, both of the tests pass.
+    def test_no_xml_log_written_fail_then_pass(self):
+        del self.env["QT_MOCK_TEST_XML_TEMPLATE_FILE"]
+        self.prepare_env(run_list=["always_pass,fail_then_pass:1"])
+        proc = self.run2()
+        # TODO verify that the whole test has run twice.
+        self.assertEqual(proc.returncode, 0)
+    # Even after 2 iterations of the full test we still get failures but no XML file,
+    # and this is considered a CRASH.
+    def test_no_xml_log_written_crash(self):
+        del self.env["QT_MOCK_TEST_XML_TEMPLATE_FILE"]
+        self.prepare_env(run_list=["fail_then_pass:2"])
+        proc = self.run2()
+        self.assertEqual(proc.returncode, 3)
+
+# Test qt-testrunner script with an existing XML log file:
+#   qt-testrunner.py qt_mock_test.py --parse-xml-testlog file.xml
+# qt-testrunner should repeat the testcases that are logged as
+# failures and fail or pass depending on how the testcases behave.
+# Different XML files are generated for the following test cases.
+#   + No failure logged.                  qt-testrunner should exit(0)
+#   + The "always_pass" test has failed.  qt-testrunner should exit(0).
+#   + The "always_fail" test has failed.  qt-testrunner should exit(2).
+#   + The "always_crash" test has failed. qt-testrunner should exit(2).
+#   + The "fail_then_pass:2" test failed. qt-testrunner should exit(0).
+#   + The "fail_then_pass:5" test failed. qt-testrunner should exit(2).
+class Test_testrunner_with_xml_logfile(unittest.TestCase):
+    # Runs before every single test function, creating a unique temp file.
+    def setUp(self):
+        (_handle, self.xml_file) = mkstemp(
+            suffix=".xml", prefix="qt_mock_test-log-",
+            dir=TEMPDIR.name)
+        if os.path.exists(os.environ["QT_MOCK_TEST_STATE_FILE"]):
+            os.remove(os.environ["QT_MOCK_TEST_STATE_FILE"])
+    def tearDown(self):
+        os.remove(self.xml_file)
+        del self.xml_file
+
+    def test_no_failure(self):
+        write_xml_log(self.xml_file, failure=None)
+        proc = run_testrunner(self.xml_file)
+        self.assertEqual(proc.returncode, 0)
+    def test_always_pass_failed(self):
+        write_xml_log(self.xml_file, failure="always_pass")
+        proc = run_testrunner(self.xml_file)
+        self.assertEqual(proc.returncode, 0)
+    def test_always_pass_failed_max_repeats_0(self):
+        write_xml_log(self.xml_file, failure="always_pass")
+        proc = run_testrunner(self.xml_file,
+                              extra_args=["--max-repeats", "0"])
+        self.assertEqual(proc.returncode, 2)
+    def test_always_fail_failed(self):
+        write_xml_log(self.xml_file, failure="always_fail")
+        proc = run_testrunner(self.xml_file)
+        self.assertEqual(proc.returncode, 2)
+    def test_always_crash_failed(self):
+        write_xml_log(self.xml_file, failure="always_crash")
+        proc = run_testrunner(self.xml_file)
+        self.assertEqual(proc.returncode, 2)
+    def test_fail_then_pass_2_failed(self):
+        write_xml_log(self.xml_file, failure="fail_then_pass:2")
+        proc = run_testrunner(self.xml_file)
+        self.assertEqual(proc.returncode, 0)
+    def test_fail_then_pass_5_failed(self):
+        write_xml_log(self.xml_file, failure="fail_then_pass:5")
+        proc = run_testrunner(self.xml_file)
+        self.assertEqual(proc.returncode, 2)
+    def test_with_two_failures(self):
+        write_xml_log(self.xml_file,
+                      failure=["always_pass", "fail_then_pass:2"])
+        proc = run_testrunner(self.xml_file)
+        self.assertEqual(proc.returncode, 0)
+        # Check that test output is properly interleaved with qt-testrunner's logging.
+        matches = re.findall(r"(PASS|FAIL!).*\n.*Test process exited with code",
+                             proc.stdout.decode())
+        self.assertEqual(len(matches), 4)
+
+
+if __name__ == "__main__":
+
+    DEBUG = False
+    if "--debug" in sys.argv:
+        sys.argv.remove("--debug")
+        DEBUG = True
+
+    # We set failfast=True as we do not want the test suite to continue if the
+    # tests of qt_mock_test failed. The next ones depend on it.
+    unittest.main(failfast=True)