OLD | NEW |
(Empty) | |
| 1 # Copyright 2014 The Chromium Authors. All rights reserved. |
| 2 # Use of this source code is governed by a BSD-style license that can be |
| 3 # found in the LICENSE file. |
| 4 |
| 5 """ |
| 6 This recipe can be used by components like v8 to verify blink tests with a |
| 7 low false positive rate. Similar to a trybot, this recipe compares test |
| 8 failures from a build with a current component revision with test failures |
| 9 from a build with a pinned component revision. |
| 10 |
| 11 Summary of the recipe flow: |
| 12 1. Sync chromium to HEAD |
| 13 2. Sync blink to HEAD |
| 14 3. Sync component X to revision Y |
| 15 4. Run blink tests |
| 16 -> In case of failures: |
| 17 5. Sync chromium to same revision as 1 |
| 18 6. Sync blink to same revision as 2 |
| 19 7. Sync component X to pinned revision from DEPS file |
| 20 8. Run blink tests |
| 21 -> If failures in 4 don't happen in 8, then revision Y reveals a problem not |
| 22 present in the pinned revision |
| 23 |
| 24 Revision Y will be the revision property as provided by buildbot or HEAD (i.e. |
| 25 in a forced build with no revision provided). |
| 26 """ |
| 27 |
| 28 from recipe_engine.types import freeze |
| 29 |
| 30 DEPS = [ |
| 31 'build/chromium', |
| 32 'build/chromium_checkout', |
| 33 'build/chromium_tests', |
| 34 'build/test_utils', |
| 35 'depot_tools/bot_update', |
| 36 'depot_tools/gclient', |
| 37 'recipe_engine/json', |
| 38 'recipe_engine/path', |
| 39 'recipe_engine/platform', |
| 40 'recipe_engine/properties', |
| 41 'recipe_engine/python', |
| 42 'recipe_engine/raw_io', |
| 43 'recipe_engine/step', |
| 44 ] |
| 45 |
| 46 |
| 47 def V8Builder(config, bits, platform): |
| 48 chromium_configs = [] |
| 49 if config == 'Debug': |
| 50 chromium_configs.append('v8_optimize_medium') |
| 51 chromium_configs.append('v8_slow_dchecks') |
| 52 return { |
| 53 'gclient_apply_config': ['show_v8_revision'], |
| 54 'chromium_apply_config': chromium_configs, |
| 55 'chromium_config_kwargs': { |
| 56 'BUILD_CONFIG': config, |
| 57 'TARGET_BITS': bits, |
| 58 }, |
| 59 'test_args': ['--no-pixel-tests'], |
| 60 'additional_expectations': [ |
| 61 'v8', 'tools', 'blink_tests', 'TestExpectations', |
| 62 ], |
| 63 'component': {'path': 'src/v8', 'revision': '%s'}, |
| 64 'testing': {'platform': platform}, |
| 65 } |
| 66 |
| 67 |
| 68 BUILDERS = freeze({ |
| 69 'client.v8.fyi': { |
| 70 'builders': { |
| 71 'V8-Blink Win': V8Builder('Release', 32, 'win'), |
| 72 'V8-Blink Mac': V8Builder('Release', 64, 'mac'), |
| 73 'V8-Blink Linux 64': V8Builder('Release', 64, 'linux'), |
| 74 'V8-Blink Linux 64 - ignition': V8Builder('Release', 64, 'linux'), |
| 75 'V8-Blink Linux 64 (dbg)': V8Builder('Debug', 64, 'linux'), |
| 76 }, |
| 77 }, |
| 78 }) |
| 79 |
| 80 |
| 81 def determine_new_ignition_failures(caller_api, extra_args): |
| 82 tests = [ |
| 83 caller_api.chromium_tests.steps.BlinkTest( |
| 84 extra_args=extra_args + [ |
| 85 '--additional-expectations', |
| 86 caller_api.path['checkout'].join( |
| 87 'v8', 'tools', 'blink_tests', 'TestExpectationsIgnition'), |
| 88 '--additional-driver-flag', |
| 89 '--js-flags=--ignition', |
| 90 ], |
| 91 ), |
| 92 ] |
| 93 |
| 94 failing_tests = caller_api.test_utils.run_tests_with_patch(caller_api, tests) |
| 95 if not failing_tests: |
| 96 return |
| 97 |
| 98 try: |
| 99 # HACK(machenbach): Blink tests store state about failing tests. In order |
| 100 # to rerun without ignition, we need to remove the extra args from the |
| 101 # existing test object. TODO(machenbach): Remove this once ignition ships. |
| 102 failing_tests[0]._extra_args = extra_args |
| 103 caller_api.test_utils.run_tests(caller_api, failing_tests, 'without patch') |
| 104 finally: |
| 105 with caller_api.step.defer_results(): |
| 106 for t in failing_tests: |
| 107 caller_api.test_utils._summarize_retried_test(caller_api, t) |
| 108 |
| 109 |
| 110 def RunSteps(api): |
| 111 mastername = api.properties.get('mastername') |
| 112 buildername = api.properties.get('buildername') |
| 113 master_dict = BUILDERS.get(mastername, {}) |
| 114 bot_config = master_dict.get('builders', {}).get(buildername) |
| 115 |
| 116 # Sync chromium to HEAD. |
| 117 api.gclient.set_config('chromium', GIT_MODE=True) |
| 118 api.gclient.c.revisions['src'] = 'HEAD' |
| 119 api.chromium.set_config('blink', |
| 120 **bot_config.get('chromium_config_kwargs', {})) |
| 121 |
| 122 for c in bot_config.get('gclient_apply_config', []): |
| 123 api.gclient.apply_config(c) |
| 124 for c in bot_config.get('chromium_apply_config', []): |
| 125 api.chromium.apply_config(c) |
| 126 |
| 127 # Sync component to current component revision. |
| 128 component_revision = api.properties.get('revision') or 'HEAD' |
| 129 api.gclient.c.revisions[bot_config['component']['path']] = ( |
| 130 bot_config['component']['revision'] % component_revision) |
| 131 |
| 132 # Ensure we remember the chromium revision. |
| 133 api.gclient.c.got_revision_mapping['src'] = 'got_cr_revision' |
| 134 |
| 135 context = {} |
| 136 checkout_dir = api.chromium_checkout.get_checkout_dir(bot_config) |
| 137 if checkout_dir: |
| 138 context['cwd'] = checkout_dir |
| 139 |
| 140 # Run all steps in the checkout dir (consistent with chromium_tests). |
| 141 with api.step.context(context): |
| 142 # TODO(phajdan.jr): remove redundant **context below once we fix things |
| 143 # to behave the same without it. |
| 144 step_result = api.bot_update.ensure_checkout(force=True, **context) |
| 145 |
| 146 api.chromium.ensure_goma() |
| 147 |
| 148 api.chromium.c.project_generator.tool = 'mb' |
| 149 api.chromium.runhooks() |
| 150 |
| 151 api.chromium_tests.run_mb_and_compile( |
| 152 ['blink_tests'], [], |
| 153 name_suffix=' (with patch)', |
| 154 ) |
| 155 |
| 156 api.chromium.runtest('webkit_unit_tests', xvfb=True) |
| 157 |
| 158 def component_pinned_fn(_failing_steps): |
| 159 bot_update_json = step_result.json.output |
| 160 api.gclient.c.revisions['src'] = str( |
| 161 bot_update_json['properties']['got_cr_revision']) |
| 162 # Reset component revision to the pinned revision from chromium's DEPS |
| 163 # for comparison. |
| 164 del api.gclient.c.revisions[bot_config['component']['path']] |
| 165 # Update without changing got_revision. The first sync is the revision |
| 166 # that is tested. The second is just for comparison. Setting got_revision |
| 167 # again confuses the waterfall's console view. |
| 168 api.bot_update.ensure_checkout(force=True, update_presentation=False) |
| 169 |
| 170 api.chromium_tests.run_mb_and_compile( |
| 171 ['blink_tests'], [], |
| 172 name_suffix=' (without patch)', |
| 173 ) |
| 174 |
| 175 extra_args = list(bot_config.get('test_args', [])) |
| 176 if bot_config.get('additional_expectations'): |
| 177 extra_args.extend([ |
| 178 '--additional-expectations', |
| 179 api.path['checkout'].join(*bot_config['additional_expectations']), |
| 180 ]) |
| 181 |
| 182 tests = [ |
| 183 api.chromium_tests.steps.BlinkTest(extra_args=extra_args), |
| 184 ] |
| 185 |
| 186 if 'ignition' in buildername: |
| 187 determine_new_ignition_failures(api, extra_args) |
| 188 else: |
| 189 api.test_utils.determine_new_failures(api, tests, component_pinned_fn) |
| 190 |
| 191 |
| 192 def _sanitize_nonalpha(text): |
| 193 return ''.join(c if c.isalnum() else '_' for c in text) |
| 194 |
| 195 |
| 196 def GenTests(api): |
| 197 canned_test = api.test_utils.canned_test_output |
| 198 with_patch = 'webkit_tests (with patch)' |
| 199 without_patch = 'webkit_tests (without patch)' |
| 200 |
| 201 def properties(mastername, buildername): |
| 202 return ( |
| 203 api.properties.generic(mastername=mastername, |
| 204 buildername=buildername, |
| 205 revision='20123', |
| 206 path_config='kitchen') |
| 207 ) |
| 208 |
| 209 for mastername, master_config in BUILDERS.iteritems(): |
| 210 for buildername, bot_config in master_config['builders'].iteritems(): |
| 211 test_name = 'full_%s_%s' % (_sanitize_nonalpha(mastername), |
| 212 _sanitize_nonalpha(buildername)) |
| 213 tests = [] |
| 214 for (pass_first, suffix) in ((True, '_pass'), (False, '_fail')): |
| 215 test = ( |
| 216 properties(mastername, buildername) + |
| 217 api.platform( |
| 218 bot_config['testing']['platform'], |
| 219 bot_config.get( |
| 220 'chromium_config_kwargs', {}).get('TARGET_BITS', 64)) + |
| 221 api.test(test_name + suffix) + |
| 222 api.override_step_data(with_patch, canned_test(passing=pass_first)) |
| 223 ) |
| 224 if not pass_first: |
| 225 test += api.override_step_data( |
| 226 without_patch, canned_test(passing=False, minimal=True)) |
| 227 tests.append(test) |
| 228 |
| 229 for test in tests: |
| 230 yield test |
| 231 |
| 232 # This tests that if the first fails, but the second pass succeeds |
| 233 # that we fail the whole build. |
| 234 yield ( |
| 235 api.test('minimal_pass_continues') + |
| 236 properties('client.v8.fyi', 'V8-Blink Linux 64') + |
| 237 api.override_step_data(with_patch, canned_test(passing=False)) + |
| 238 api.override_step_data(without_patch, |
| 239 canned_test(passing=True, minimal=True)) |
| 240 ) |
| 241 |
| 242 |
| 243 # This tests what happens if something goes horribly wrong in |
| 244 # run-webkit-tests and we return an internal error; the step should |
| 245 # be considered a hard failure and we shouldn't try to compare the |
| 246 # lists of failing tests. |
| 247 # 255 == test_run_results.UNEXPECTED_ERROR_EXIT_STATUS in run-webkit-tests. |
| 248 yield ( |
| 249 api.test('webkit_tests_unexpected_error') + |
| 250 properties('client.v8.fyi', 'V8-Blink Linux 64') + |
| 251 api.override_step_data(with_patch, canned_test(passing=False, retcode=255)) |
| 252 ) |
| 253 |
| 254 # TODO(dpranke): crbug.com/357866 . This tests what happens if we exceed the |
| 255 # number of failures specified with --exit-after-n-crashes-or-times or |
| 256 # --exit-after-n-failures; the step should be considered a hard failure and |
| 257 # we shouldn't try to compare the lists of failing tests. |
| 258 # 130 == test_run_results.INTERRUPTED_EXIT_STATUS in run-webkit-tests. |
| 259 yield ( |
| 260 api.test('webkit_tests_interrupted') + |
| 261 properties('client.v8.fyi', 'V8-Blink Linux 64') + |
| 262 api.override_step_data(with_patch, canned_test(passing=False, retcode=130)) |
| 263 ) |
| 264 |
| 265 # This tests what happens if we don't trip the thresholds listed |
| 266 # above, but fail more tests than we can safely fit in a return code. |
| 267 # (this should be a soft failure and we can still retry w/o the patch |
| 268 # and compare the lists of failing tests). |
| 269 yield ( |
| 270 api.test('too_many_failures_for_retcode') + |
| 271 properties('client.v8.fyi', 'V8-Blink Linux 64') + |
| 272 api.override_step_data(with_patch, |
| 273 canned_test(passing=False, |
| 274 num_additional_failures=125)) + |
| 275 api.override_step_data(without_patch, |
| 276 canned_test(passing=True, minimal=True)) |
| 277 ) |
OLD | NEW |