grpc/tools/run_tests/run_tests_matrix.py

#!/usr/bin/env python
# Copyright 2015 gRPC authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Run test matrix."""

from __future__ import print_function

import argparse
import multiprocessing
import os
import sys

import python_utils.jobset as jobset
import python_utils.report_utils as report_utils
from python_utils.filter_pull_request_tests import filter_tests

_ROOT = os.path.abspath(os.path.join(os.path.dirname(sys.argv[0]), '../..'))
os.chdir(_ROOT)

_DEFAULT_RUNTESTS_TIMEOUT = 1 * 60 * 60

# Set the timeout high to allow enough time for sanitizers and pre-building
# clang docker.
_CPP_RUNTESTS_TIMEOUT = 4 * 60 * 60

# C++ TSAN takes longer than other sanitizers
_CPP_TSAN_RUNTESTS_TIMEOUT = 8 * 60 * 60

# Set timeout high for ObjC for Cocoapods to install pods
_OBJC_RUNTESTS_TIMEOUT = 90 * 60

# Number of jobs assigned to each run_tests.py instance
_DEFAULT_INNER_JOBS = 2

# Name of the top-level umbrella report that includes all the run_tests.py invocations
# Note that the starting letter 't' matters so that the targets are listed AFTER
# the per-test breakdown items that start with 'run_tests/' (it is more readable that way)
_MATRIX_REPORT_NAME = 'toplevel_run_tests_invocations'

def _safe_report_name(name):
    """Reports with '+' in target name won't show correctly in ResultStore"""
    return name.replace('+', 'p')


def _report_filename(name):
    """Generates report file name with directory structure that leads to better presentation by internal CI"""
    # 'sponge_log.xml' suffix must be there for results to get recognized by kokoro.
    return '%s/%s' % (_safe_report_name(name), 'sponge_log.xml')


def _matrix_job_logfilename(shortname_for_multi_target):
    """Generate location for log file that will match the sponge_log.xml from the top-level matrix report."""
    # 'sponge_log.log' suffix must be there for log to get recognized as "target log"
    # for the corresponding 'sponge_log.xml' report.
    # the shortname_for_multi_target component must be set to match the sponge_log.xml location
    # because the top-level render_junit_xml_report is called with multi_target=True
    s = '%s/%s/%s' % (_MATRIX_REPORT_NAME, shortname_for_multi_target, 'sponge_log.log')
    print(s)
    return s


def _docker_jobspec(name,
                    runtests_args=[],
                    runtests_envs={},
                    inner_jobs=_DEFAULT_INNER_JOBS,
                    timeout_seconds=None):
    """Run a single instance of run_tests.py in a docker container"""
    if not timeout_seconds:
        timeout_seconds = _DEFAULT_RUNTESTS_TIMEOUT
    shortname = 'run_tests_%s' % name
    test_job = jobset.JobSpec(
        cmdline=[
            'python', 'tools/run_tests/run_tests.py', '--use_docker', '-t',
            '-j',
            str(inner_jobs), '-x',
            'run_tests/%s' % _report_filename(name), '--report_suite_name',
            '%s' % _safe_report_name(name)
        ] + runtests_args,
        environ=runtests_envs,
        shortname=shortname,
        timeout_seconds=timeout_seconds,
        logfilename=_matrix_job_logfilename(shortname))
    return test_job


def _workspace_jobspec(name,
                       runtests_args=[],
                       workspace_name=None,
                       runtests_envs={},
                       inner_jobs=_DEFAULT_INNER_JOBS,
                       timeout_seconds=None):
    """Run a single instance of run_tests.py in a separate workspace"""
    if not workspace_name:
        workspace_name = 'workspace_%s' % name
    if not timeout_seconds:
        timeout_seconds = _DEFAULT_RUNTESTS_TIMEOUT
    shortname = 'run_tests_%s' % name
    env = {'WORKSPACE_NAME': workspace_name}
    env.update(runtests_envs)
    test_job = jobset.JobSpec(
        cmdline=[
            'bash', 'tools/run_tests/helper_scripts/run_tests_in_workspace.sh',
            '-t', '-j',
            str(inner_jobs), '-x',
            '../run_tests/%s' % _report_filename(name), '--report_suite_name',
            '%s' % _safe_report_name(name)
        ] + runtests_args,
        environ=env,
        shortname=shortname,
        timeout_seconds=timeout_seconds,
        logfilename=_matrix_job_logfilename(shortname))
    return test_job


def _generate_jobs(languages,
                   configs,
                   platforms,
                   iomgr_platforms=['native'],
                   arch=None,
                   compiler=None,
                   labels=[],
                   extra_args=[],
                   extra_envs={},
                   inner_jobs=_DEFAULT_INNER_JOBS,
                   timeout_seconds=None):
    result = []
    for language in languages:
        for platform in platforms:
            for iomgr_platform in iomgr_platforms:
                for config in configs:
                    name = '%s_%s_%s_%s' % (language, platform, config,
                                            iomgr_platform)
                    runtests_args = [
                        '-l', language, '-c', config, '--iomgr_platform',
                        iomgr_platform
                    ]
                    if arch or compiler:
                        name += '_%s_%s' % (arch, compiler)
                        runtests_args += [
                            '--arch', arch, '--compiler', compiler
                        ]
                    if '--build_only' in extra_args:
                        name += '_buildonly'
                    for extra_env in extra_envs:
                        name += '_%s_%s' % (extra_env, extra_envs[extra_env])

                    runtests_args += extra_args
                    if platform == 'linux':
                        job = _docker_jobspec(
                            name=name,
                            runtests_args=runtests_args,
                            runtests_envs=extra_envs,
                            inner_jobs=inner_jobs,
                            timeout_seconds=timeout_seconds)
                    else:
                        job = _workspace_jobspec(
                            name=name,
                            runtests_args=runtests_args,
                            runtests_envs=extra_envs,
                            inner_jobs=inner_jobs,
                            timeout_seconds=timeout_seconds)

                    job.labels = [platform, config, language, iomgr_platform
                                 ] + labels
                    result.append(job)
    return result


def _create_test_jobs(extra_args=[], inner_jobs=_DEFAULT_INNER_JOBS):
    test_jobs = []
    # sanity tests
    test_jobs += _generate_jobs(
        languages=['sanity'],
        configs=['dbg', 'opt'],
        platforms=['linux'],
        labels=['basictests'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs)

    # supported on linux only
    test_jobs += _generate_jobs(
        languages=['php7'],
        configs=['dbg', 'opt'],
        platforms=['linux'],
        labels=['basictests', 'multilang'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs)

    # supported on all platforms.
    test_jobs += _generate_jobs(
        languages=['c'],
        configs=['dbg', 'opt'],
        platforms=['linux', 'macos', 'windows'],
        labels=['basictests', 'corelang'],
        extra_args=extra_args,  # don't use multi_target report because C has too many test cases
        inner_jobs=inner_jobs,
        timeout_seconds=_CPP_RUNTESTS_TIMEOUT)

    # C# tests on .NET desktop/mono
    test_jobs += _generate_jobs(
        languages=['csharp'],
        configs=['dbg', 'opt'],
        platforms=['linux', 'macos', 'windows'],
        labels=['basictests', 'multilang'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs)
    # C# tests on .NET core
    test_jobs += _generate_jobs(
        languages=['csharp'],
        configs=['dbg', 'opt'],
        platforms=['linux', 'macos', 'windows'],
        arch='default',
        compiler='coreclr',
        labels=['basictests', 'multilang'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs)

    test_jobs += _generate_jobs(
        languages=['python'],
        configs=['opt'],
        platforms=['linux', 'macos', 'windows'],
        iomgr_platforms=['native', 'gevent'],
        labels=['basictests', 'multilang'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs)

    # supported on linux and mac.
    test_jobs += _generate_jobs(
        languages=['c++'],
        configs=['dbg', 'opt'],
        platforms=['linux', 'macos'],
        labels=['basictests', 'corelang'],
        extra_args=extra_args,  # don't use multi_target report because C++ has too many test cases
        inner_jobs=inner_jobs,
        timeout_seconds=_CPP_RUNTESTS_TIMEOUT)

    test_jobs += _generate_jobs(
        languages=['grpc-node', 'ruby', 'php'],
        configs=['dbg', 'opt'],
        platforms=['linux', 'macos'],
        labels=['basictests', 'multilang'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs)

    # supported on mac only.
    test_jobs += _generate_jobs(
        languages=['objc'],
        configs=['opt'],
        platforms=['macos'],
        labels=['basictests', 'multilang'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs,
        timeout_seconds=_OBJC_RUNTESTS_TIMEOUT)

    # sanitizers
    test_jobs += _generate_jobs(
        languages=['c'],
        configs=['msan', 'asan', 'tsan', 'ubsan'],
        platforms=['linux'],
        arch='x64',
        compiler='clang7.0',
        labels=['sanitizers', 'corelang'],
        extra_args=extra_args,
        inner_jobs=inner_jobs,
        timeout_seconds=_CPP_RUNTESTS_TIMEOUT)
    test_jobs += _generate_jobs(
        languages=['c++'],
        configs=['asan'],
        platforms=['linux'],
        arch='x64',
        compiler='clang7.0',
        labels=['sanitizers', 'corelang'],
        extra_args=extra_args,
        inner_jobs=inner_jobs,
        timeout_seconds=_CPP_RUNTESTS_TIMEOUT)
    test_jobs += _generate_jobs(
        languages=['c++'],
        configs=['tsan'],
        platforms=['linux'],
        arch='x64',
        compiler='clang7.0',
        labels=['sanitizers', 'corelang'],
        extra_args=extra_args,
        inner_jobs=inner_jobs,
        timeout_seconds=_CPP_TSAN_RUNTESTS_TIMEOUT)

    return test_jobs


def _create_portability_test_jobs(extra_args=[],
                                  inner_jobs=_DEFAULT_INNER_JOBS):
    test_jobs = []
    # portability C x86
    test_jobs += _generate_jobs(
        languages=['c'],
        configs=['dbg'],
        platforms=['linux'],
        arch='x86',
        compiler='default',
        labels=['portability', 'corelang'],
        extra_args=extra_args,
        inner_jobs=inner_jobs)

    # portability C and C++ on x64
    for compiler in [
            'gcc4.8', 'gcc5.3', 'gcc7.2', 'gcc_musl', 'clang3.5', 'clang3.6',
            'clang3.7', 'clang7.0'
    ]:
        test_jobs += _generate_jobs(
            languages=['c', 'c++'],
            configs=['dbg'],
            platforms=['linux'],
            arch='x64',
            compiler=compiler,
            labels=['portability', 'corelang'],
            extra_args=extra_args,
            inner_jobs=inner_jobs,
            timeout_seconds=_CPP_RUNTESTS_TIMEOUT)

    # portability C on Windows 64-bit (x86 is the default)
    test_jobs += _generate_jobs(
        languages=['c'],
        configs=['dbg'],
        platforms=['windows'],
        arch='x64',
        compiler='default',
        labels=['portability', 'corelang'],
        extra_args=extra_args,
        inner_jobs=inner_jobs)

    # portability C++ on Windows
    # TODO(jtattermusch): some of the tests are failing, so we force --build_only
    test_jobs += _generate_jobs(
        languages=['c++'],
        configs=['dbg'],
        platforms=['windows'],
        arch='default',
        compiler='default',
        labels=['portability', 'corelang'],
        extra_args=extra_args + ['--build_only'],
        inner_jobs=inner_jobs)

    # portability C and C++ on Windows using VS2017 (build only)
    # TODO(jtattermusch): some of the tests are failing, so we force --build_only
    test_jobs += _generate_jobs(
        languages=['c', 'c++'],
        configs=['dbg'],
        platforms=['windows'],
        arch='x64',
        compiler='cmake_vs2017',
        labels=['portability', 'corelang'],
        extra_args=extra_args + ['--build_only'],
        inner_jobs=inner_jobs)

    # C and C++ with the c-ares DNS resolver on Linux
    test_jobs += _generate_jobs(
        languages=['c', 'c++'],
        configs=['dbg'],
        platforms=['linux'],
        labels=['portability', 'corelang'],
        extra_args=extra_args,
        extra_envs={'GRPC_DNS_RESOLVER': 'ares'},
        timeout_seconds=_CPP_RUNTESTS_TIMEOUT)

    # C and C++ with no-exceptions on Linux
    test_jobs += _generate_jobs(
        languages=['c', 'c++'],
        configs=['noexcept'],
        platforms=['linux'],
        labels=['portability', 'corelang'],
        extra_args=extra_args,
        timeout_seconds=_CPP_RUNTESTS_TIMEOUT)

    # TODO(zyc): Turn on this test after adding c-ares support on windows.
    # C with the c-ares DNS resolver on Windows
    # test_jobs += _generate_jobs(languages=['c'],
    #                             configs=['dbg'], platforms=['windows'],
    #                             labels=['portability', 'corelang'],
    #                             extra_args=extra_args,
    #                             extra_envs={'GRPC_DNS_RESOLVER': 'ares'})

    # C and C++ build with cmake on Linux
    # TODO(jtattermusch): some of the tests are failing, so we force --build_only
    # to make sure it's buildable at least.
    test_jobs += _generate_jobs(
        languages=['c', 'c++'],
        configs=['dbg'],
        platforms=['linux'],
        arch='default',
        compiler='cmake',
        labels=['portability', 'corelang'],
        extra_args=extra_args + ['--build_only'],
        inner_jobs=inner_jobs)

    test_jobs += _generate_jobs(
        languages=['python'],
        configs=['dbg'],
        platforms=['linux'],
        arch='default',
        compiler='python_alpine',
        labels=['portability', 'multilang'],
        extra_args=extra_args + ['--report_multi_target'],
        inner_jobs=inner_jobs)

    # TODO(jtattermusch): a large portion of the libuv tests is failing,
    # which can end up killing the kokoro job due to gigabytes of error logs
    # generated. Remove the --build_only flag
    # once https://github.com/grpc/grpc/issues/17556 is fixed.
    test_jobs += _generate_jobs(
        languages=['c'],
        configs=['dbg'],
        platforms=['linux'],
        iomgr_platforms=['uv'],
        labels=['portability', 'corelang'],
        extra_args=extra_args + ['--build_only'],
        inner_jobs=inner_jobs,
        timeout_seconds=_CPP_RUNTESTS_TIMEOUT)

    return test_jobs


def _allowed_labels():
    """Returns a list of existing job labels."""
    all_labels = set()
    for job in _create_test_jobs() + _create_portability_test_jobs():
        for label in job.labels:
            all_labels.add(label)
    return sorted(all_labels)


def _runs_per_test_type(arg_str):
    """Auxiliary function to parse the "runs_per_test" flag."""
    try:
        n = int(arg_str)
        if n <= 0: raise ValueError
        return n
    except:
        msg = '\'{}\' is not a positive integer'.format(arg_str)
        raise argparse.ArgumentTypeError(msg)


if __name__ == "__main__":
    argp = argparse.ArgumentParser(
        description='Run a matrix of run_tests.py tests.')
    argp.add_argument(
        '-j',
        '--jobs',
        default=multiprocessing.cpu_count() / _DEFAULT_INNER_JOBS,
        type=int,
        help='Number of concurrent run_tests.py instances.')
    argp.add_argument(
        '-f',
        '--filter',
        choices=_allowed_labels(),
        nargs='+',
        default=[],
        help='Filter targets to run by label with AND semantics.')
    argp.add_argument(
        '--exclude',
        choices=_allowed_labels(),
        nargs='+',
        default=[],
        help='Exclude targets with any of given labels.')
    argp.add_argument(
        '--build_only',
        default=False,
        action='store_const',
        const=True,
        help='Pass --build_only flag to run_tests.py instances.')
    argp.add_argument(
        '--force_default_poller',
        default=False,
        action='store_const',
        const=True,
        help='Pass --force_default_poller to run_tests.py instances.')
    argp.add_argument(
        '--dry_run',
        default=False,
        action='store_const',
        const=True,
        help='Only print what would be run.')
    argp.add_argument(
        '--filter_pr_tests',
        default=False,
        action='store_const',
        const=True,
        help='Filters out tests irrelevant to pull request changes.')
    argp.add_argument(
        '--base_branch',
        default='origin/master',
        type=str,
        help='Branch that pull request is requesting to merge into')
    argp.add_argument(
        '--inner_jobs',
        default=_DEFAULT_INNER_JOBS,
        type=int,
        help='Number of jobs in each run_tests.py instance')
    argp.add_argument(
        '-n',
        '--runs_per_test',
        default=1,
        type=_runs_per_test_type,
        help='How many times to run each tests. >1 runs implies ' +
        'omitting passing test from the output & reports.')
    argp.add_argument(
        '--max_time',
        default=-1,
        type=int,
        help='Maximum amount of time to run tests for' +
        '(other tests will be skipped)')
    argp.add_argument(
        '--internal_ci',
        default=False,
        action='store_const',
        const=True,
        help=
        '(Deprecated, has no effect) Put reports into subdirectories to improve presentation of '
        'results by Kokoro.')
    argp.add_argument(
        '--bq_result_table',
        default='',
        type=str,
        nargs='?',
        help='Upload test results to a specified BQ table.')
    argp.add_argument(
        '--extra_args',
        default='',
        type=str,
        nargs=argparse.REMAINDER,
        help='Extra test args passed to each sub-script.')
    args = argp.parse_args()

    extra_args = []
    if args.build_only:
        extra_args.append('--build_only')
    if args.force_default_poller:
        extra_args.append('--force_default_poller')
    if args.runs_per_test > 1:
        extra_args.append('-n')
        extra_args.append('%s' % args.runs_per_test)
        extra_args.append('--quiet_success')
    if args.max_time > 0:
        extra_args.extend(('--max_time', '%d' % args.max_time))
    if args.bq_result_table:
        extra_args.append('--bq_result_table')
        extra_args.append('%s' % args.bq_result_table)
        extra_args.append('--measure_cpu_costs')
    if args.extra_args:
        extra_args.extend(args.extra_args)

    all_jobs = _create_test_jobs(extra_args=extra_args, inner_jobs=args.inner_jobs) + \
               _create_portability_test_jobs(extra_args=extra_args, inner_jobs=args.inner_jobs)

    jobs = []
    for job in all_jobs:
        if not args.filter or all(
                filter in job.labels for filter in args.filter):
            if not any(exclude_label in job.labels
                       for exclude_label in args.exclude):
                jobs.append(job)

    if not jobs:
        jobset.message(
            'FAILED', 'No test suites match given criteria.', do_newline=True)
        sys.exit(1)

    print('IMPORTANT: The changes you are testing need to be locally committed')
    print('because only the committed changes in the current branch will be')
    print('copied to the docker environment or into subworkspaces.')

    skipped_jobs = []

    if args.filter_pr_tests:
        print('Looking for irrelevant tests to skip...')
        relevant_jobs = filter_tests(jobs, args.base_branch)
        if len(relevant_jobs) == len(jobs):
            print('No tests will be skipped.')
        else:
            print('These tests will be skipped:')
            skipped_jobs = list(set(jobs) - set(relevant_jobs))
            # Sort by shortnames to make printing of skipped tests consistent
            skipped_jobs.sort(key=lambda job: job.shortname)
            for job in list(skipped_jobs):
                print('  %s' % job.shortname)
        jobs = relevant_jobs

    print('Will run these tests:')
    for job in jobs:
        print('  %s: "%s"' % (job.shortname, ' '.join(job.cmdline)))
    print('')

    if args.dry_run:
        print('--dry_run was used, exiting')
        sys.exit(1)

    jobset.message('START', 'Running test matrix.', do_newline=True)
    num_failures, resultset = jobset.run(
        jobs, newline_on_success=True, travis=True, maxjobs=args.jobs)
    # Merge skipped tests into results to show skipped tests on report.xml
    if skipped_jobs:
        ignored_num_skipped_failures, skipped_results = jobset.run(
            skipped_jobs, skip_jobs=True)
        resultset.update(skipped_results)
    report_utils.render_junit_xml_report(
        resultset,
        _report_filename(_MATRIX_REPORT_NAME),
        suite_name=_MATRIX_REPORT_NAME, multi_target=True)

    if num_failures == 0:
        jobset.message(
            'SUCCESS',
            'All run_tests.py instances finished successfully.',
            do_newline=True)
    else:
        jobset.message(
            'FAILED',
            'Some run_tests.py instances have failed.',
            do_newline=True)
        sys.exit(1)