ChangeLog

2017-07-28  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: fix return value of close window command
        https://bugs.webkit.org/show_bug.cgi?id=174861

        Reviewed by Brian Burg.

        We are currently returning null, but we should return the list of window handles, and try to close the session
        if there aren't more window handles.

        10.2 Close Window
        https://w3c.github.io/webdriver/webdriver-spec.html#close-window

        3. If there are no more open top-level browsing contexts, then try to close the session.
        4. Return the result of running the remote end steps for the Get Window Handles command.

        * Session.cpp:
        (WebDriver::Session::closeAllToplevelBrowsingContexts): Helper function to close the given toplevel browsing
        context and the next one if there are more.
        (WebDriver::Session::close): Call closeAllToplevelBrowsingContexts() to delete all toplevel browsing contexts of
        the session.
        (WebDriver::Session::closeTopLevelBrowsingContext): Close the given toplevel browsing context and call
        getWindowHandles() when done.
        (WebDriver::Session::closeWindow): Call closeTopLevelBrowsingContext() passing the current toplevel browsing context.
        (WebDriver::Session::getWindowHandles): Remove the early return, this command doesn't depend on a current
        toplevel browsing context.
        * Session.h:
        * SessionHost.h:
        * WebDriverService.cpp:
        (WebDriver::WebDriverService::run): Disconnect the server when main loop quits.
        (WebDriver::WebDriverService::deleteSession): Do not fail if the given session is not active.
        (WebDriver::WebDriverService::closeWindow): Remove the session if the closed window was the last one.
        * WebDriverService.h: Remove unused quit() method.
        * glib/SessionHostGlib.cpp:
        (WebDriver::SessionHost::isConnected): Return whether host is connected to a browser instance.
        (WebDriver::SessionHost::dbusConnectionClosedCallback): Delete m_browser.

2017-08-14  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: handle click events on option elements
        https://bugs.webkit.org/show_bug.cgi?id=174710
        <rdar://problem/33459305>

        Reviewed by Brian Burg.

        Option elements are considered as a special case by the specification. When clicking an option element, we
        should get its container and use it when scrolling into view and calculating in-view center point instead of the
        option element itself. Then, we should not emulate a click, but change the selected status of the option element
        like if it were done by a user action, firing the corresponding events. Now we check whether the element is an
        option to call selectOptionElement() or performMouseInteraction().

        This fixes more than 20 selenium tests.

        * CommandResult.cpp:
        (WebDriver::CommandResult::CommandResult): Handle ElementNotSelectable protocol error.
        (WebDriver::CommandResult::httpStatusCode const): Add ElementNotSelectable.
        (WebDriver::CommandResult::errorString const): Ditto.
        * CommandResult.h:
        * Session.cpp:
        (WebDriver::Session::selectOptionElement): Ask automation to select the given option element.
        (WebDriver::Session::elementClick): Call selectOptionElement() or performMouseInteraction() depending on whether
        the element is an option or not.
        * Session.h:

2017-08-11  Carlos Alberto Lopez Perez  <clopez@igalia.com>

        Fix build warning in WebDriverService.h
        https://bugs.webkit.org/show_bug.cgi?id=166682

        Unreviewed build fix.

        * WebDriverService.h: Fixes the warning: class 'Capabilities' was previously declared as a struct [-Wmismatched-tags]

2017-08-08  Michael Catanzaro  <mcatanzaro@igalia.com>

        Unreviewed, fix Ubuntu LTS build
        https://bugs.webkit.org/show_bug.cgi?id=174490

        * glib/SessionHostGlib.cpp:

2017-08-08  Michael Catanzaro  <mcatanzaro@igalia.com>

        [CMake] Properly test if compiler supports compiler flags
        https://bugs.webkit.org/show_bug.cgi?id=174490

        Reviewed by Konstantin Tokarev.

        * WebDriverService.cpp:
        (WebDriver::WebDriverService::run):
        * glib/SessionHostGlib.cpp:

2017-08-07  Carlos Garcia Campos  <cgarcia@igalia.com>

        Web Automation: setUserInputForCurrentJavaScriptPrompt should fail if current dialog is not a prompt
        https://bugs.webkit.org/show_bug.cgi?id=175261

        Reviewed by Brian Burg.

        * CommandResult.cpp:
        (WebDriver::CommandResult::CommandResult): Handle ElementNotInteractable protocol error.

2017-08-07  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: implement unhandled prompt behavior
        https://bugs.webkit.org/show_bug.cgi?id=175184

        Reviewed by Brian Burg.

        Handle user prompts before running some of the commands according to the specification.

        * Capabilities.h: Add UnhandledPromptBehavior capability.
        * CommandResult.cpp:
        (WebDriver::CommandResult::httpStatusCode const): Add UnexpectedAlertOpen error.
        (WebDriver::CommandResult::errorString const): Ditto.
        * CommandResult.h:
        (WebDriver::CommandResult::setAdditonalErrorData): New method to set an additional data object that will be sent
        as part of the result error message.
        (WebDriver::CommandResult::additionalErrorData const): Return the additional data object.
        * Session.cpp:
        (WebDriver::Session::handleUserPrompts): Check if there's an active JavaScript dialog and deal with it depeding
        on the unhandled prompt behavior.
        (WebDriver::Session::reportUnexpectedAlertOpen): Generate an error message with UnexpectedAlertOpen error and
        including the alert text as additional error data.
        (WebDriver::Session::go): Handle user prompts before running the command.
        (WebDriver::Session::getCurrentURL): Ditto.
        (WebDriver::Session::back): Ditto.
        (WebDriver::Session::forward): Ditto.
        (WebDriver::Session::refresh): Ditto.
        (WebDriver::Session::getTitle): Ditto.
        (WebDriver::Session::closeWindow): Ditto.
        (WebDriver::Session::switchToFrame): Ditto.
        (WebDriver::Session::switchToParentFrame): Ditto.
        (WebDriver::Session::isElementSelected): Ditto.
        (WebDriver::Session::getElementText): Ditto.
        (WebDriver::Session::getElementTagName): Ditto.
        (WebDriver::Session::getElementRect): Ditto.
        (WebDriver::Session::isElementEnabled): Ditto.
        (WebDriver::Session::isElementDisplayed): Ditto.
        (WebDriver::Session::getElementAttribute): Ditto.
        (WebDriver::Session::elementSendKeys): Ditto.
        (WebDriver::Session::elementSubmit): Ditto.
        (WebDriver::Session::executeScript): Ditto.
        * Session.h:
        * WebDriverService.cpp:
        (WebDriver::WebDriverService::sendResponse const): Send data object as part of the result error message if present.
        (WebDriver::deserializeUnhandledPromptBehavior):
        (WebDriver::WebDriverService::parseCapabilities const):
        (WebDriver::WebDriverService::validatedCapabilities const):
        (WebDriver::WebDriverService::newSession):

2017-08-07  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: implement user prompt commands
        https://bugs.webkit.org/show_bug.cgi?id=174614

        Reviewed by Brian Burg.

        * CommandResult.cpp:
        (WebDriver::CommandResult::CommandResult): Handle NoJavaScriptDialog protocol error.
        (WebDriver::CommandResult::httpStatusCode const): Add NoSuchAlert.
        (WebDriver::CommandResult::errorString const): Ditto.
        * CommandResult.h:
        * Session.cpp:
        (WebDriver::Session::dismissAlert):
        (WebDriver::Session::acceptAlert):
        (WebDriver::Session::getAlertText):
        (WebDriver::Session::sendAlertText):
        * Session.h:
        * WebDriverService.cpp:
        (WebDriver::WebDriverService::dismissAlert):
        (WebDriver::WebDriverService::acceptAlert):
        (WebDriver::WebDriverService::getAlertText):
        (WebDriver::WebDriverService::sendAlertText):
        * WebDriverService.h:

2017-08-05  Carlos Garcia Campos  <cgarcia@igalia.com>

        [GTK][WPE] Add API to provide browser information required by automation
        https://bugs.webkit.org/show_bug.cgi?id=175130

        Reviewed by Brian Burg.

        * Session.cpp:
        (WebDriver::Session::createTopLevelBrowsingContext): Check if startAutomationSession and complete the command
        with error in that case.
        * SessionHost.h:
        * glib/SessionHostGlib.cpp:
        (WebDriver::SessionHost::matchCapabilities): Match the capabilities that are known only after the browser has
        been launched.
        (WebDriver::SessionHost::startAutomationSession): Handle the StartAutomationSession response, extracting the
        capabilities and calling matchCapabilities() to match them.
        (WebDriver::SessionHost::setTargetList): Return early if the session was rejected before due to invalid
        capabilities.

2017-08-05  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: Implement page load strategy
        https://bugs.webkit.org/show_bug.cgi?id=175183

        Reviewed by Brian Burg.

        Validate and parse page load strategy when processing capabilities.

        * Capabilities.h:
        * Session.cpp:
        (WebDriver::Session::pageLoadStrategyString const): Helper to get the page load strategy as a String to be
        passed to Automation.
        (WebDriver::Session::go): Pass page load strategy if present.
        (WebDriver::Session::back): Ditto.
        (WebDriver::Session::forward): Ditto.
        (WebDriver::Session::refresh): Ditto.
        (WebDriver::Session::waitForNavigationToComplete): Ditto.
        * Session.h:
        * WebDriverService.cpp:
        (WebDriver::deserializePageLoadStrategy):
        (WebDriver::WebDriverService::parseCapabilities const):
        (WebDriver::WebDriverService::validatedCapabilities const):
        (WebDriver::WebDriverService::newSession):

2017-08-05  Carlos Garcia Campos  <cgarcia@igalia.com>

        Unreviewed. Try to fix build with clang after r220315.

        * WebDriverService.cpp:
        (WebDriver::WebDriverService::validatedCapabilities const):
        (WebDriver::WebDriverService::mergeCapabilities const):

2017-08-05  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: properly handle capabilities and process firstMatch too
        https://bugs.webkit.org/show_bug.cgi?id=174618

        Reviewed by Brian Burg.

        Implement processing of capabilities following the spec. This patch adds validation, merging and matching of
        capabilities.

        7.2 Processing Capabilities.
        https://w3c.github.io/webdriver/webdriver-spec.html#processing-capabilities

        * Capabilities.h: Make all capabilities optional and move Timeouts struct here.
        * Session.h:
        * WebDriverService.cpp:
        (WebDriver::deserializeTimeouts): Helper to extract timeouts from JSON object.
        (WebDriver::WebDriverService::parseCapabilities const): At this point capabilities have already been validated,
        so we just need to get them without checking the value type.
        (WebDriver::WebDriverService::validatedCapabilities const): Validate the given capabilities, ensuring types of
        values are the expected one.
        (WebDriver::WebDriverService::mergeCapabilities const): Merge the alwaysMatch and firstMatch capabilities into a
        single object ensuring that the same capability is not in both.
        (WebDriver::WebDriverService::matchCapabilities const): Try to match the merged capabilities againt the platform
        expected capabilities.
        (WebDriver::WebDriverService::processCapabilities const): Validate, merge and match the capabilities.
        (WebDriver::WebDriverService::newSession): Use processCapabilities(). Also initialize the timeouts from
        capabilities and add all capabilities to the command result.
        (WebDriver::WebDriverService::setTimeouts): Use deserializeTimeouts().
        * WebDriverService.h:
        * glib/SessionHostGlib.cpp:
        (WebDriver::SessionHost::launchBrowser): Updated to properly access the capabilities that are now optional.
        (WebDriver::SessionHost::startAutomationSession): Add FIXME.
        * gtk/WebDriverServiceGtk.cpp:
        (WebDriver::WebDriverService::platformCapabilities): Return the Capabilities with the known required ones filled.
        (WebDriver::WebDriverService::platformValidateCapability const): Validate webkitgtk:browserOptions.
        (WebDriver::WebDriverService::platformMatchCapability const): This does nothing for now.
        (WebDriver::WebDriverService::platformCompareBrowserVersions): Compare the given browser versions.
        (WebDriver::WebDriverService::platformParseCapabilities const): Updated now that capabilites have already been
        validated before.

2017-08-05  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: use in-view center point for clicks instead of bounding box center point
        https://bugs.webkit.org/show_bug.cgi?id=174863

        Reviewed by Simon Fraser.

        The center of the element bounding box is not always part of the element, like in multiline links, for example.

        11.1 Element Interactability.
        https://www.w3.org/TR/webdriver/#dfn-in-view-center-point

        * CommandResult.cpp:
        (WebDriver::CommandResult::httpStatusCode): Add ElementClickIntercepted and ElementNotInteractable errors.
        (WebDriver::CommandResult::errorString): Ditto.
        * CommandResult.h: Ditto.
        * Session.cpp:
        (WebDriver::Session::computeElementLayout): Get the in-view center point and isObscured from the result too.
        (WebDriver::Session::getElementRect): Ignore in-view center point and isObscured.
        (WebDriver::Session::elementClick): Fail in case the element is not interactable or is obscured.
        * Session.h:

2017-08-01  Michael Catanzaro  <mcatanzaro@igalia.com>

        [CMake] WebKitFS.cmake depends on options set in Option cmake files that are included later
        https://bugs.webkit.org/show_bug.cgi?id=174855

        Reviewed by Carlos Garcia Campos.

        Don't create derived sources directory here anymore.

        * PlatformGTK.cmake:

2017-07-26  Carlos Garcia Campos  <cgarcia@igalia.com>

        Unreviewed. Fix GTK distcheck.

        Ensure WebDriver derived sources directory is created, WebKitFS.cmake is useless for this.

        * PlatformGTK.cmake:

2017-07-24  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: rename m_browsingContext as m_currentBrowsingContext in Session
        https://bugs.webkit.org/show_bug.cgi?id=174783

        Reviewed by Brian Burg.

        We have m_toplevelBrowsingContext and m_browsingContext, which is confusing. m_browsingContext is actually the
        current browsing context, and the spec also refers to it as the current browsing context, so better use
        m_currentBrowsingContext.

        * Session.cpp:
        (WebDriver::Session::switchToTopLevelBrowsingContext):
        (WebDriver::Session::switchToBrowsingContext):
        (WebDriver::Session::switchToFrame):
        (WebDriver::Session::switchToParentFrame):
        (WebDriver::Session::computeElementLayout):
        (WebDriver::Session::findElements):
        (WebDriver::Session::isElementSelected):
        (WebDriver::Session::getElementText):
        (WebDriver::Session::getElementTagName):
        (WebDriver::Session::isElementEnabled):
        (WebDriver::Session::isElementDisplayed):
        (WebDriver::Session::getElementAttribute):
        (WebDriver::Session::waitForNavigationToComplete):
        (WebDriver::Session::elementClear):
        (WebDriver::Session::elementSendKeys):
        (WebDriver::Session::elementSubmit):
        (WebDriver::Session::executeScript):
        * Session.h:

2017-07-20  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: implement page load timeout
        https://bugs.webkit.org/show_bug.cgi?id=174672

        Reviewed by Brian Burg.

        Handle timeout errors and pass the page load timeout to waitForNavigationToComplete and all other navigation
        commands. Also fix the setTimeouts command that was still using the legacy name of the page load timeout,
        instead of the one in the spec.

        8. Sessions
        https://www.w3.org/TR/webdriver/#dfn-session-page-load-timeout

        * CommandResult.cpp:
        (WebDriver::CommandResult::CommandResult):
        (WebDriver::CommandResult::httpStatusCode):
        (WebDriver::CommandResult::errorString):
        * CommandResult.h:
        * Session.cpp:
        (WebDriver::Session::go):
        (WebDriver::Session::back):
        (WebDriver::Session::forward):
        (WebDriver::Session::refresh):
        (WebDriver::Session::waitForNavigationToComplete):
        * WebDriverService.cpp:
        (WebDriver::WebDriverService::setTimeouts):

2017-07-21  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: wait until navigation is complete before running new commands and after a click
        https://bugs.webkit.org/show_bug.cgi?id=174670

        Reviewed by Brian Burg.

        We are already waiting for navigation to complete after navigation commands (go, back, forward, refresh), but
        the spec says we should always wait before executing a new command and also after a click. This is causing test
        testShouldBeAbleToNavigateBackInTheBrowserHistoryInPresenceOfIframes to sometimes fail, because it does .click()
        + .title and expects the title to tbe the one of the page loaded by the click. Since the load happens very fast,
        the test usually passes, but in a real case with a slower load, the title of the previous page will be returned
        most of the times.

        6.3 Processing Model. Step 7. Wait for navigation to complete. If this returns an error return its value and
        jump to step 1 in this overall algorithm, otherwise continue.
        https://www.w3.org/TR/webdriver/#processing-model

        14.1 Element Click. Step 10. If the click causes navigation: 1. Run the post-navigation checks and return its
        value if it is an error. 2. Try to wait for navigation to complete.
        https://www.w3.org/TR/webdriver/#element-click

        * Session.cpp:
        (WebDriver::Session::waitForNavigationToComplete): Send waitForNavigationToComplete message to the browser to
        wait for any pending navigation of current browsing context to complete.
        (WebDriver::Session::elementClick): Call waitForNavigationToComplete() after the click.
        * Session.h:
        * WebDriverService.cpp:
        (WebDriver::WebDriverService::go): Wait for navigations to complete before running the command.
        (WebDriver::WebDriverService::getCurrentURL): Ditto.
        (WebDriver::WebDriverService::back): Ditto.
        (WebDriver::WebDriverService::forward): Ditto.
        (WebDriver::WebDriverService::refresh): Ditto.
        (WebDriver::WebDriverService::getTitle): Ditto.
        (WebDriver::WebDriverService::switchToFrame): Ditto.
        (WebDriver::WebDriverService::switchToParentFrame): Ditto.
        (WebDriver::WebDriverService::findElement): Ditto.
        (WebDriver::WebDriverService::findElements): Ditto.
        (WebDriver::WebDriverService::executeScript): Ditto.
        (WebDriver::WebDriverService::executeAsyncScript): Ditto.

2017-07-21  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: correctly handle main frame handles
        https://bugs.webkit.org/show_bug.cgi?id=174668

        Reviewed by Brian Burg.

        When I switched to use std::optional instead of empty strings for the browsing contexts in WebDriver, I forgot
        that automation uses empty string for frames to refer to the main frame. We should handle that case, because we
        are currently considering empty strings as valid browsing context. It's not a big deal because Automation
        converts back the empty string received to the main frame, though. We should also ensure we close the current
        browsing context when switching to a new top level browsing context. This patch adds to helper private methods
        to switch browsing contexts that deal with the special cases.

        * Session.cpp:
        (WebDriver::Session::close):
        (WebDriver::Session::switchToTopLevelBrowsingContext):
        (WebDriver::Session::switchToBrowsingContext):
        (WebDriver::Session::createTopLevelBrowsingContext):
        (WebDriver::Session::go):
        (WebDriver::Session::back):
        (WebDriver::Session::forward):
        (WebDriver::Session::refresh):
        (WebDriver::Session::switchToWindow):
        (WebDriver::Session::switchToFrame):
        (WebDriver::Session::switchToParentFrame):
        * Session.h:

2017-07-18  Carlos Garcia Campos  <cgarcia@igalia.com>

        WebDriver: handle invalid selector errors
        https://bugs.webkit.org/show_bug.cgi?id=174619

        Reviewed by Brian Burg.

        Add InvalidSelector error and handle it in case of protocol server error.

        * CommandResult.cpp:
        (WebDriver::CommandResult::CommandResult):
        (WebDriver::CommandResult::httpStatusCode):
        (WebDriver::CommandResult::errorString):
        * CommandResult.h:

2017-07-18  Carlos Alberto Lopez Perez  <clopez@igalia.com>

        [GTK] Fix build with Clang after r219605.
        https://bugs.webkit.org/show_bug.cgi?id=166682

        Unreviewed build fix.

        Clang-3.8 complains with the following error:
        non-constant-expression cannot be narrowed from type 'gboolean' (aka 'int') to 'bool' in initializer list [-Wc++11-narrowing]

        * glib/SessionHostGlib.cpp: insert an explicit cast to silence this issue.

2017-07-13  Carlos Garcia Campos  <cgarcia@igalia.com>

        Add initial implementation of WebDriver process to run the HTTP server
        https://bugs.webkit.org/show_bug.cgi?id=166682

        Reviewed by Brian Burg.

        Add WebDriver process that runs the HTTP server and implements an initial set of commands. Most of the code is
        cross-platform, only the HTTP server implementation, the code to launch the browser and the communication with
        the remote inspector requires platform specific code. This patch includes the GTK port implementation, using
        libsoup for the HTTP server, and GLib for launching the browser and communicating with the remote
        inspector. This implementation follows the w3c spec (https://www.w3.org/TR/webdriver) as close as possible, but
        using the official selenium python tests as reference.

        * CMakeLists.txt: Added.
        * Capabilities.h: Added.
        * CommandResult.cpp: Added.
        * CommandResult.h: Added.
        * HTTPServer.cpp: Added.
        * HTTPServer.h: Added.
        * PlatformGTK.cmake: Added.
        * Session.cpp: Added.
        * Session.h: Added.
        * SessionHost.cpp: Added.
        * SessionHost.h: Added.
        * WebDriverMain.cpp: Added.
        * WebDriverService.cpp: Added.
        * WebDriverService.h: Added.
        * config.h: Added.
        * glib/SessionHostGlib.cpp: Added.
        * gtk/WebDriverServiceGtk.cpp: Added.
        * soup/HTTPServerSoup.cpp: Added.