2011-02-18 Mihai Parparita Reviewed by Ojan Vafai. Default to 500 for --exit-after-n-failures https://bugs.webkit.org/show_bug.cgi?id=54773 Make 500 the default value for --exit-after-n-failures (ORWT bots already default to this as of r75726). Similarly, change the default for --exit-after-n-crashes-or-timeouts to 20. * Scripts/webkitpy/layout_tests/run_webkit_tests.py: 2011-02-02 Martin Robinson Reviewed by Xan Lopez. [GTK] DRT needs an implementation of LayoutTestController.setSelectTrailingWhitespace https://bugs.webkit.org/show_bug.cgi?id=53603 Add support for LayoutTestController.setSelectTrailingWhitespace by calling through to DumpRenderTreeSupportGtk for this functionality. * DumpRenderTree/gtk/DumpRenderTree.cpp: (resetDefaultsToConsistentValues): Always call LayoutTestController.setSelectTrailingWhitespace to false when reseting values between tests. * DumpRenderTree/gtk/LayoutTestControllerGtk.cpp: (LayoutTestController::setSelectTrailingWhitespaceEnabled): Call LayoutTestController.setSelectTrailingWhitespace with the appropriate value. 2011-02-22 Adam Roben Don't trigger a Windows build when the top-level ChangeLog changes r79320 is an example of a change that shouldn't have built on Windows, but did, because of this bug. Fixes Windows bots build when top-level ChangeLog changes, but shouldn't Reviewed by Anders Carlsson. * Scripts/webkitpy/common/config/build.py: (_should_file_trigger_build): Modified the ChangeLog pattern to match the top-level ChangeLog, and re-sorted it in the list. * Scripts/webkitpy/common/config/build_unittest.py: (ShouldBuildTest): Added a test that ChangeLogs don't trigger builds. 2011-02-22 Sheriff Bot Unreviewed, rolling out r78691. http://trac.webkit.org/changeset/78691 https://bugs.webkit.org/show_bug.cgi?id=54956 Leopard Debug is successfully building again (Requested by aroben on #webkit). * Scripts/webkitpy/common/net/buildbot/buildbot.py: * Scripts/webkitpy/common/net/buildbot/buildbot_unittest.py: 2011-02-22 Ilya Tikhonovsky Reviewed by Pavel Feldman. Web Inspector: flakyness of inspector tests. https://bugs.webkit.org/show_bug.cgi?id=54729 As far as we have the protocol with sequence numbers we can simplify test harness support and drop out chomium specific methods. * DumpRenderTree/DumpRenderTree.gypi: * DumpRenderTree/chromium/DRTDevToolsAgent.cpp: (DRTDevToolsAgent::sendMessageToInspectorFrontend): (DRTDevToolsAgent::runtimePropertyChanged): (DRTDevToolsAgent::asyncCall): (DRTDevToolsAgent::call): * DumpRenderTree/chromium/DRTDevToolsAgent.h: * DumpRenderTree/chromium/DRTDevToolsCallArgs.cpp: Removed. * DumpRenderTree/chromium/DRTDevToolsCallArgs.h: Removed. * DumpRenderTree/chromium/DRTDevToolsClient.cpp: (DRTDevToolsClient::sendMessageToBackend): (DRTDevToolsClient::asyncCall): (DRTDevToolsClient::call): * DumpRenderTree/chromium/DRTDevToolsClient.h: 2011-02-22 Andras Becsi Reviewed by Laszlo Gombos. [Qt] Redesign the build system https://bugs.webkit.org/show_bug.cgi?id=51339 Part 2. Build WebCore as a static library, compile the WebKit API and WebKit2 API in a final step and link to WebKit2, WebCore and JSC libraries to fix linking issues resulting from stripped away symbols. * WebKitTestRunner/InjectedBundle/qt/InjectedBundle.pro: Remove superfluous includepaths. 2011-02-21 Lukasz Slachciak Reviewed by Antonio Gomes. [EFL] Restored spatial navigation option in EWebLauncher since #18662 bug is closed https://bugs.webkit.org/show_bug.cgi?id=54806 * EWebLauncher/main.c: (on_key_down): calling ewk_view_setting_spatial_navigation_set when "F12" pressed added. 2011-02-21 Dirk Pranke Reviewed by Kent Tamura. nrwt: revert behavior on 10.5 to "old-threads". It's possible that r79062 has introduced a bunch of flakiness, although I can't explain why right now. This change should restore the old behavior, and then we can look at the build logs to see if the tests stop timing out or being flaky. https://bugs.webkit.org/show_bug.cgi?id=54925 * Scripts/webkitpy/layout_tests/port/chromium_mac.py: * Scripts/webkitpy/layout_tests/port/chromium_mac_unittest.py: * Scripts/webkitpy/layout_tests/port/mac.py: * Scripts/webkitpy/layout_tests/port/mac_unittest.py: 2011-02-21 Dirk Pranke Reviewed by Dimitri Glazkov. nrwt: fix typo causing us to try and repeatedly shut down the servers https://bugs.webkit.org/show_bug.cgi?id=54904 * Scripts/webkitpy/layout_tests/layout_package/worker_mixin.py: 2011-02-21 Alejandro G. Castro Reviewed by Csaba Osztrogonác. REGRESSION(r79157): plugins/get-url-with-javascript-destroying-plugin.html make the next test crash https://bugs.webkit.org/show_bug.cgi?id=54863 * GNUmakefile.am: 2011-02-20 Dirk Pranke Reviewed by Ojan Vafai. test-webkitpy: stop skipping a bunch of tests on win32 https://bugs.webkit.org/show_bug.cgi?id=54788 * Scripts/webkitpy/common/net/testoutputset_unittest.py: * Scripts/webkitpy/test/main.py: 2011-02-20 Dirk Pranke Reviewed by Ojan Vafai. Fix various bugs related to handling of dir separators and real vs. fake filesystems that were causing test-webkitpy to fail tests on win32. These bugs have a variety of causes but are lumped together because they should be easy to review. https://bugs.webkit.org/show_bug.cgi?id=54700 * Scripts/webkitpy/common/net/testoutput.py: Here we were using os.path.sep but not escaping it prior to using in a regexp, which caused bad things on win32. * Scripts/webkitpy/common/net/testoutput_unittest.py: Same here. * Scripts/webkitpy/layout_tests/port/chromium_unittest.py: Here we were using the native filesystem instead of the mock filesystem, making some tests unnecessarily platform dependent. Also we weren't setting the default configuration for the TestChromiumWinPort. * Scripts/webkitpy/layout_tests/update_webgl_conformance_tests.py: Here we *shouldn't* use os.path.join(), just so the code becomes less platform dependent. * Scripts/webkitpy/style/checker.py: Here we should be using os.path.sep instead of "/"; the tests require the native filesystem. * Scripts/webkitpy/style/checkers/cpp_unittest.py: Here the underlying code is using os.path.sep, so we probably should as well. * Scripts/webkitpy/style/test_expectations_unittest.py: Here we should be using the filesystem hanging of the port object (which is the mocked filesystem) rather than the native one. 2011-02-20 Anders Carlsson Reviewed by Maciej Stachowiak. Crash when a plug-in requests a javascript: url that destroys the plug-in https://bugs.webkit.org/show_bug.cgi?id=54837 Add new plug-in test. * DumpRenderTree/DumpRenderTree.gypi: * DumpRenderTree/DumpRenderTree.xcodeproj/project.pbxproj: * DumpRenderTree/TestNetscapePlugIn/PluginTest.cpp: (PluginTest::NPN_GetURL): * DumpRenderTree/TestNetscapePlugIn/PluginTest.h: * DumpRenderTree/TestNetscapePlugIn/Tests/GetURLWithJavaScriptURLDestroyingPlugin.cpp: Added. (GetURLWithJavaScriptURLDestroyingPlugin::GetURLWithJavaScriptURLDestroyingPlugin): (GetURLWithJavaScriptURLDestroyingPlugin::NPP_New): * DumpRenderTree/TestNetscapePlugIn/win/TestNetscapePlugin.vcproj: * DumpRenderTree/qt/TestNetscapePlugin/TestNetscapePlugin.pro: 2011-02-19 Patrick Gansterer Reviewed by Andreas Kling. Make WinCE a core builder https://bugs.webkit.org/show_bug.cgi?id=54801 * Scripts/webkitpy/common/net/buildbot/buildbot.py: * Scripts/webkitpy/common/net/buildbot/buildbot_unittest.py: 2011-02-19 Lukasz Slachciak Reviewed by Andreas Kling. Fixed problem with launching EWebLauncher with "run-launcher --efl". Incorrect path referring to non-existing .libs directory ws set. https://webkit.org/b/54778 * Scripts/webkitdirs.pm: 2011-02-18 David Levin Reviewed by Eric Seidel. check-webkit-style: Misses brace style error when the line with the { has a } in it. https://bugs.webkit.org/show_bug.cgi?id=54769 * Scripts/webkitpy/style/checkers/cpp.py: Fixed the check for the close brace to only look after the last open brace, so that the open brace in this line "} else {" will still be able to trigger the error. * Scripts/webkitpy/style/checkers/cpp_unittest.py: Added related unit tests. 2011-02-18 Zan Dobersek Reviewed by Martin Robinson. [Gtk] Re-enable meter tag support https://bugs.webkit.org/show_bug.cgi?id=54762 Meter tag is supported on the Gtk port, enable it by default. * Scripts/build-webkit: 2011-02-18 David Levin Reviewed by Eric Seidel. check-webkit-style falsely complains about WebKitGTK+ public headers https://bugs.webkit.org/show_bug.cgi?id=54650 * Scripts/webkitpy/style/checker.py: Add some exceptions for the gtk files. * Scripts/webkitpy/style/checkers/cpp.py: Restrict the checks for WEBKIT_API to the chromium directory (and improved the checks slightly). * Scripts/webkitpy/style/checkers/cpp_unittest.py: Added corresponding tests. 2011-02-18 Dirk Pranke Reviewed by James Robinson. new-run-webkit-tests: enable multiple processes by default on mac. This change removes the artificial restrictions we placed on the # of child processes we used with the old threading model, and switches to the new message-based model and multiple processes, where available. If multiple processes are not available (Leopard / Python 2.5), then we use the 'inline' model instead of the 'old-threads' model on the 'mac' port or one process on the 'chromium-mac' port. We need additional testing to see if the new 'threads' model works reliably and is worth supporting, or if we should just live with things being slightly slow. https://bugs.webkit.org/show_bug.cgi?id=54596 * Scripts/webkitpy/layout_tests/port/chromium_mac.py: * Scripts/webkitpy/layout_tests/port/chromium_mac_unittest.py: * Scripts/webkitpy/layout_tests/port/mac.py: * Scripts/webkitpy/layout_tests/port/mac_unittest.py: * Scripts/webkitpy/layout_tests/run_webkit_tests_unittest.py: 2011-02-18 Andrew Wilson Unreviewed, rolling out r79047. http://trac.webkit.org/changeset/79047 https://bugs.webkit.org/show_bug.cgi?id=54596 Broke canary bots - please watch the canaries the next time you land this * Scripts/webkitpy/layout_tests/port/chromium_mac.py: * Scripts/webkitpy/layout_tests/port/mac.py: 2011-02-18 Emil A Eklund Reviewed by Darin Adler. Crash in EventHandler::sendContextMenuEventForKey https://bugs.webkit.org/show_bug.cgi?id=54495 Add support for the context menu key (VK_APPS) to EventSender for platforms that support the key. * DumpRenderTree/chromium/EventSender.cpp: (EventSender::keyDown): * DumpRenderTree/gtk/EventSender.cpp: (keyDownCallback): * DumpRenderTree/qt/EventSenderQt.cpp: (EventSender::keyDown): * DumpRenderTree/win/EventSender.cpp: (keyDownCallback): 2011-02-18 Dirk Pranke Reviewed by Ojan Vafai. new-run-webkit-tests: enable multiple processes by default on mac https://bugs.webkit.org/show_bug.cgi?id=54596 * Scripts/webkitpy/layout_tests/port/chromium_mac.py: * Scripts/webkitpy/layout_tests/port/mac.py: 2011-02-18 Dirk Pranke Reviewed by Adam Roben. Skip pretty patch unit tests if ruby isn't installed. https://bugs.webkit.org/show_bug.cgi?id=54699 * Scripts/webkitpy/common/prettypatch_unittest.py: 2011-02-18 Dirk Pranke Reviewed by Adam Roben. Skip mac.test_skipped_file_paths() when running on win32; the test has hardcoded '/' directory separators and since it's a port-specific test and we have coverage on other ports it's not worth it to make the test more generic. https://bugs.webkit.org/show_bug.cgi?id=54698 * Scripts/webkitpy/layout_tests/port/mac_unittest.py: 2011-02-18 Dirk Pranke Reviewed by Adam Roben. Fix webkitpy.layout_tests.port.config_unittest to work on Win32. https://bugs.webkit.org/show_bug.cgi?id=54703 * Scripts/webkitpy/layout_tests/port/config_unittest.py: 2011-02-18 Dirk Pranke Reviewed by Adam Roben. Change the default port we're mocking out on win32 when we run mock_drt_unittest. Normally we'd default to the port-specific default, but that would be "win", which doesn't currently work. https://bugs.webkit.org/show_bug.cgi?id=54708 * Scripts/webkitpy/layout_tests/port/mock_drt_unittest.py: 2011-02-18 Kenneth Russell Reviewed by James Robinson. [chromium] Clean up initialization of minimum timer interval https://bugs.webkit.org/show_bug.cgi?id=54772 Use new GetForegroundTabTimerInterval function in webkit_support to reset the minimum timer interval before each test. * DumpRenderTree/chromium/WebViewHost.cpp: (WebViewHost::reset): 2011-02-18 Dirk Pranke Reviewed by Adam Roben. r78494 fixed the chromium_win unittest failing under win32, but not cygwin. This fixes cygwin as well. https://bugs.webkit.org/show_bug.cgi?id=54652 * Scripts/webkitpy/layout_tests/port/chromium_win.py: 2011-02-18 Dirk Pranke Reviewed by Ojan Vafai, Mihai Parparita. new-run-webkit-tests: r78522 made it impossible to create a 'chromium-win' or 'chromium-mac' port; they would always get the version tacked on to the end. It turns out that we actually need to be able to create these ports in order for rebaseline-chromium-webkit-tests to work correctly given the broken coverage of our layout bots. As soon as we get SL and Win 7 bots running reliably on the canaries, we can get rid of this change. Note that we only need to change the chromium_mac and win ports because linux doesn't yet use versions and that script doesn't work with non-chromium ports. Also, this change fixes a bug where we were using "platform/chromium-win-win7" and "platform/chromium-mac-snowleopard" as the baseline dirs for Win 7 and SL; we should've been using the generic directories instead since they are the newest versions of the os's. https://bugs.webkit.org/show_bug.cgi?id=54691 * Scripts/webkitpy/layout_tests/port/chromium_mac.py: * Scripts/webkitpy/layout_tests/port/chromium_mac_unittest.py: * Scripts/webkitpy/layout_tests/port/chromium_win.py: * Scripts/webkitpy/layout_tests/port/chromium_win_unittest.py: * Scripts/webkitpy/layout_tests/rebaseline_chromium_webkit_tests.py: 2011-02-16 MORITA Hajime Reviewed by Kent Tamura. [Chromium][DRT] some PostDelayedTask() accesses possibly dangling this pointer. http://webkit.org/b/53899 Replaced static invocation wrappers with MethodTask objects, which are removed on WebViewHost destructor. * DumpRenderTree/chromium/WebViewHost.cpp: (WebViewHost::requestCheckingOfText): (WebViewHost::scheduleAnimation): (WebViewHost::closeWidgetSoon): * DumpRenderTree/chromium/WebViewHost.h: (WebViewHost::HostMethodTask::HostMethodTask): (WebViewHost::HostMethodTask::runIfValid): (WebViewHost::taskList): 2011-02-07 Hayato Ito Reviewed by Ojan Vafai. Get rid of code which writes test results from test_type's compare_output() method. Writing test results now happens in single_test_runner, using newly introduced test_result_writer module. Subsequent patches will eliminate test_type/* classes. https://bugs.webkit.org/show_bug.cgi?id=53908 * Scripts/webkitpy/layout_tests/layout_package/single_test_runner.py: * Scripts/webkitpy/layout_tests/layout_package/test_result_writer.py: Added. * Scripts/webkitpy/layout_tests/test_types/image_diff.py: * Scripts/webkitpy/layout_tests/test_types/test_type_base.py: * Scripts/webkitpy/layout_tests/test_types/text_diff.py: 2011-02-17 Dirk Pranke Reviewed by Adam Roben. Add blacklist of remaining webkitpy/* modules that fail on Win32, so that test-webkitpy will at least run cleanly. https://bugs.webkit.org/show_bug.cgi?id=54709 * Scripts/webkitpy/layout_tests/test/main.py: 2011-02-17 Simon Fraser Reviewed by Sam Weinig. Set a preference to ensure that scrollbars are in a predictable state for DRT. * DumpRenderTree/mac/DumpRenderTree.mm: (resetDefaultsToConsistentValues): 2011-02-17 Dirk Pranke Reviewed by Adam Roben. test-webkitpy doesn't work with the multiprocessing module on win32 or cygwin, so we skip it for now. https://bugs.webkit.org/show_bug.cgi?id=54520 * Scripts/webkitpy/layout_tests/layout_package/manager_worker_broker_unittest.py: * Scripts/webkitpy/layout_tests/layout_package/run_webkit_tests_unittest.py: 2011-02-17 Dirk Pranke Reviewed by Adam Roben. Fix version, baseline_search_path for chromium-win-win7. Also fix chromium_win_unittests, and fix port_testcase.make_port() to correctly use the options passed into it and only set results_directory() if necessary. https://bugs.webkit.org/show_bug.cgi?id=54652 * Scripts/webkitpy/layout_tests/port/chromium_win.py: * Scripts/webkitpy/layout_tests/port/chromium_win_unittest.py: * Scripts/webkitpy/layout_tests/port/port_testcase.py: 2011-02-17 Dirk Pranke Reviewed by Ojan Vafai. Fix a hang in new-run-webkit-tests when run with --dry-run --worker-model=processes https://bugs.webkit.org/show_bug.cgi?id=54595 * Scripts/webkitpy/layout_tests/layout_package/test_runner2.py: * Scripts/webkitpy/layout_tests/run_webkit_tests_unittest.py: 2011-02-17 Kevin Ollivier [wx] Build fixes after recent changes. * DumpRenderTree/wx/LayoutTestControllerWx.cpp: (LayoutTestController::setMinimumTimerInterval): * wx/build/settings.py: 2011-02-17 Csaba Osztrogonác Reviewed by Andreas Kling. [Qt] ImageDiff sometimes hangs https://bugs.webkit.org/show_bug.cgi?id=54641 * DumpRenderTree/qt/ImageDiff.cpp: Clear actualImage and baselineImage after all comparison. (main): 2011-02-17 Peter Varga Reviewed by Csaba Osztrogonác. The run-sunspider --v8-suite doesn't work https://bugs.webkit.org/show_bug.cgi?id=54660 * Scripts/run-sunspider: 2011-02-17 Gabor Rapcsanyi Reviewed by Csaba Osztrogonác. [NRWT] Fix environment setup on Qt port. https://bugs.webkit.org/show_bug.cgi?id=54549 * Scripts/webkitpy/layout_tests/port/qt.py: 2011-02-16 Yael Aharon Reviewed by Andreas Kling. [Qt] Enable tiled backing store by default in MiniBrowser. https://bugs.webkit.org/show_bug.cgi?id=54581 Instead of a flag to turn it on, a new flag is defined to turn tiled backing store off. * MiniBrowser/qt/MiniBrowserApplication.cpp: (MiniBrowserApplication::handleUserOptions): * MiniBrowser/qt/MiniBrowserApplication.h: (WindowOptions::WindowOptions): 2011-02-16 Eric Seidel Reviewed by Mihai Parparita. flaky test archive uploads are always 22 bytes long https://bugs.webkit.org/show_bug.cgi?id=54593 Speculative fix for the 22-bytes bug. * Scripts/webkitpy/tool/bot/flakytestreporter.py: 2011-02-16 Anna Cavender Reviewed by Mihai Parparita. http/tests/media should be served over HTTP (not from local file) https://bugs.webkit.org/show_bug.cgi?id=54028 Removed special file:// treatment of http/tests/media * Scripts/old-run-webkit-tests: * Scripts/webkitpy/layout_tests/port/base.py: 2011-02-16 Mihai Parparita Reviewed by Ojan Vafai. ChangeLog.parse_latest_entry_from_file does not handle rolled over ChangeLogs https://bugs.webkit.org/show_bug.cgi?id=54609 r78737 rolled over ChangeLogs, and webkit-patch complains when landing patches such as this one that are the first to add an entry to the ChangeLog (since they don't see a second date line to know where the entry ends). * Scripts/webkitpy/common/checkout/changelog.py: * Scripts/webkitpy/common/checkout/changelog_unittest.py: == Rolled over to ChangeLog-2011-02-16 ==