WebKit内核分析之FrameLoader

参考地址：http://blog.csdn.net/dlmu2001/article/details/6168545

FrameLoader类负责一个Frame的加载，在Frame的流程中起到非常重要的作用，同很多组件都有交互，本文将分析FrameLoader类的代码

1. 概述

     顾名思义，FrameLoader是一个Frame的loader，它的作用就是为客户端提供一个下载一个Frame的一系列的接口。这里的客户指的是类的客户，比如Frame类，间接客户是上层应用，比如 qwebframe

     从它的定义看，最容易想到的是一个load接口，用来将一个frame load下来。任何一个页面都需要一个mainframe，因此一个页面的下载一般就是从load 一个 mainframe开始

     在 load frame的过程中，通过FrameLoaderClient接口将load过程的不同阶段告知客户

     FrameLoader通过 setDocumentLoader相当于把load的工作委托给了 DocumentLoader类

     FrameLoader同 DocumentLoader是has-a的关系。一般在load的时候创建 DocumentLoader。Frame调用DocumentLoader的startLoadingMainResource开始 load frame

类数据代码：

class FrameLoader {
    WTF_MAKE_NONCOPYABLE(FrameLoader);
private:
    Frame* m_frame;
    FrameLoaderClient* m_client;

    mutable PolicyChecker m_policyChecker;
    mutable HistoryController m_history;
    mutable ResourceLoadNotifier m_notifer;
    mutable SubframeLoader m_subframeLoader;
    mutable FrameLoaderStateMachine m_stateMachine;

    FrameState m_state;
    FrameLoadType m_loadType;

    // Document loaders for the three phases of frame loading. Note that while 
    // a new request is being loaded, the old document loader may still be referenced.
    // E.g. while a new request is in the "policy" state, the old document loader may
    // be consulted in particular as it makes sense to imply certain settings on the new loader.
    RefPtr<DocumentLoader> m_documentLoader;
    RefPtr<DocumentLoader> m_provisionalDocumentLoader;
    RefPtr<DocumentLoader> m_policyDocumentLoader;

    bool m_delegateIsHandlingProvisionalLoadError;

    bool m_quickRedirectComing;
    bool m_sentRedirectNotification;
    bool m_inStopAllLoaders;

    String m_outgoingReferrer;

    bool m_isExecutingJavaScriptFormAction;

    bool m_didCallImplicitClose;
    bool m_wasUnloadEventEmitted;
    bool m_pageDismissalEventBeingDispatched;
    bool m_isComplete;
    bool m_isLoadingMainResource;

    RefPtr<SerializedScriptValue> m_pendingStateObject;

    KURL m_workingURL;

    OwnPtr<IconLoader> m_iconLoader;
    bool m_mayLoadIconLater;

    bool m_needsClear;

    KURL m_submittedFormURL;

    Timer<FrameLoader> m_checkTimer;
    bool m_shouldCallCheckCompleted;
    bool m_shouldCallCheckLoadComplete;

    Frame* m_opener;
    HashSet<Frame*> m_openedFrames;

    bool m_didPerformFirstNavigation;
    bool m_loadingFromCachedPage;
    bool m_suppressOpenerInNewFrame;

    SandboxFlags m_sandboxFlags;
    SandboxFlags m_forcedSandboxFlags;

    RefPtr<FrameNetworkingContext> m_networkingContext;

    KURL m_previousUrl;
}

2. 类关系

1，Frame 和 FrameLoader 是 contain-a的关系，在 Frame的构造函数中调用 FraomeLoader的构造函数，调用时传入参数 Frame指针 和 FrameLoaderClient指针

inline Frame::Frame(Page* page, HTMLFrameOwnerElement* ownerElement, FrameLoaderClient* frameLoaderClient)
    : m_page(page)
    , m_treeNode(this, parentFromOwnerElement(ownerElement))
    , m_loader(this, frameLoaderClient)
    , m_navigationScheduler(this)
    , m_ownerElement(ownerElement)
    , m_script(this)
    , m_editor(this)
    , m_selectionController(this)
    , m_eventHandler(this)
    , m_animationController(this)
    , m_lifeSupportTimer(this, &Frame::lifeSupportTimerFired)
    , m_pageZoomFactor(parentPageZoomFactor(this))
    , m_textZoomFactor(parentTextZoomFactor(this))
    , m_pageScaleFactor(1)
#if ENABLE(ORIENTATION_EVENTS)
    , m_orientation(0)
#endif
    , m_inViewSourceMode(false)
    , m_isDisconnected(false)
    , m_excludeFromTextSearch(false)
#if ENABLE(MEDIA_STREAM)
    , m_mediaStreamFrameController(RuntimeEnabledFeatures::mediaStreamEnabled() ? adoptPtr(new MediaStreamFrameController(this)) : PassOwnPtr<MediaStreamFrameController>())
#endif
{
     ....
}

2，Frame 有可能有子Frame， 所以维护 SubFrameLoader对象 m_subframeLoader来管理子 Frame 的load。  Frame可以对应 xml document，也可以对应html document，等。跟Document相关的子resource的load不在FrameLoader的职责范围内。

3，包含一个DocumentWriter类对象 m_writer，当Frame的数据 load finish的时候，将数据传给 DocumentWriter类，进行下一步的处理（比如解码）

4，FrameLoader维护三个 DocumentLoader对象，分别对应不同的阶段。 

    RefPtr<DocumentLoader> m_documentLoader;
    RefPtr<DocumentLoader> m_provisionalDocumentLoader;
    RefPtr<DocumentLoader> m_policyDocumentLoader;

 m_policyDocumentLoader 对应于 收到用户 load 调用，进行 policy check 阶段

 m_provisionalDocumentLoader 对应于 policy check 通过之后，Frame数据还没有到来之前，它会负责 startLoadingMainResource的调用

 m_documentLoader 则是 Frame第一个数据到来以后使用的 DocumentLoader，这个时候，前一个主 Frame 的 DocumentLoader已经不能再用（user agent开始白屏，刷掉前一个页面的显示）

inline void Frame::init()
{
    m_loader.init();
}

void FrameLoader::init()
{
    // Propagate sandbox attributes to this Frameloader and its descendants.
    // This needs to be done early, so that an initial document gets correct sandbox flags in its SecurityOrigin.
    updateSandboxFlags();

    // this somewhat odd set of steps is needed to give the frame an initial empty document
    m_stateMachine.advanceTo(FrameLoaderStateMachine::CreatingInitialEmptyDocument);
    setPolicyDocumentLoader(m_client->createDocumentLoader(ResourceRequest(KURL(ParsedURLString, "")), SubstituteData()).get());
    setProvisionalDocumentLoader(m_policyDocumentLoader.get());
    setState(FrameStateProvisional);
    m_provisionalDocumentLoader->setResponse(ResourceResponse(KURL(), "text/html", 0, String(), String()));
    m_provisionalDocumentLoader->finishedLoading();
    m_documentLoader->writer()->begin(KURL(), false);
    m_documentLoader->writer()->end();
    m_frame->document()->cancelParsing();
    m_stateMachine.advanceTo(FrameLoaderStateMachine::DisplayingInitialEmptyDocument);
    m_didCallImplicitClose = true;

    m_networkingContext = m_client->createNetworkingContext();
}

从上面代码可以看出，在Frame初始化的时候，仅仅调用FrameLoader的初始化函数

然后FrameLoader初始化的时候，首先把状态机的状态标识成 FrameLoaderStateMachine::CreatingInitialEmptyDocument ，然后调用客户端实现创建一个DocumentLoader，然后设置为PolicyDocumentLoader，并且设置ProvisionalDocumentLoader和PolicyDocumentLoader为同一个loader，并且设置FrameState状态为 Provisional（初始化为commitPage状态）

然后把状态机状态更新为  FrameLoaderStateMachine::DisplayingInitialEmptyDocument  完成初始化

后面还有一次load的变迁：

void FrameLoader::transitionToCommitted(PassRefPtr<CachedPage> cachedPage)
{
    ASSERT(m_client->hasWebView());
    ASSERT(m_state == FrameStateProvisional);

    if (m_state != FrameStateProvisional)
        return;

    if (m_frame->view())
        m_frame->view()->scrollAnimator()->cancelAnimations();

    m_client->setCopiesOnScroll();
    history()->updateForCommit();

    // The call to closeURL() invokes the unload event handler, which can execute arbitrary
    // JavaScript. If the script initiates a new load, we need to abandon the current load,
    // or the two will stomp each other.
    DocumentLoader* pdl = m_provisionalDocumentLoader.get();
    if (m_documentLoader)
        closeURL();
    if (pdl != m_provisionalDocumentLoader)
        return;

    // Nothing else can interupt this commit - set the Provisional->Committed transition in stone
    if (m_documentLoader)
        m_documentLoader->stopLoadingSubresources();
    if (m_documentLoader)
        m_documentLoader->stopLoadingPlugIns();

    setDocumentLoader(m_provisionalDocumentLoader.get());
    setProvisionalDocumentLoader(0);
    setState(FrameStateCommittedPage);

    // Handle adding the URL to the back/forward list.
    DocumentLoader* dl = m_documentLoader.get();

    switch (m_loadType) {
        case FrameLoadTypeForward:
        case FrameLoadTypeBack:
        case FrameLoadTypeIndexedBackForward:
            if (m_frame->page()) {
                // If the first load within a frame is a navigation within a back/forward list that was attached
                // without any of the items being loaded then we need to update the history in a similar manner as
                // for a standard load with the exception of updating the back/forward list (<rdar://problem/8091103>).
                if (!m_stateMachine.committedFirstRealDocumentLoad())
                    history()->updateForStandardLoad(HistoryController::UpdateAllExceptBackForwardList);

                history()->updateForBackForwardNavigation();

                // For cached pages, CachedFrame::restore will take care of firing the popstate event with the history item's state object
                if (history()->currentItem() && !cachedPage)
                    m_pendingStateObject = history()->currentItem()->stateObject();

                // Create a document view for this document, or used the cached view.
                if (cachedPage) {
                    DocumentLoader* cachedDocumentLoader = cachedPage->documentLoader();
                    ASSERT(cachedDocumentLoader);
                    cachedDocumentLoader->setFrame(m_frame);
                    m_client->transitionToCommittedFromCachedFrame(cachedPage->cachedMainFrame());

                } else
                    m_client->transitionToCommittedForNewPage();
            }
            break;

        case FrameLoadTypeReload:
        case FrameLoadTypeReloadFromOrigin:
        case FrameLoadTypeSame:
        case FrameLoadTypeReplace:
            history()->updateForReload();
            m_client->transitionToCommittedForNewPage();
            break;

        case FrameLoadTypeStandard:
            history()->updateForStandardLoad();
            if (m_frame->view())
                m_frame->view()->setScrollbarsSuppressed(true);
            m_client->transitionToCommittedForNewPage();
            break;

        case FrameLoadTypeRedirectWithLockedBackForwardList:
            history()->updateForRedirectWithLockedBackForwardList();
            m_client->transitionToCommittedForNewPage();
            break;

        // FIXME Remove this check when dummy ds is removed (whatever "dummy ds" is).
        // An exception should be thrown if we're in the FrameLoadTypeUninitialized state.
        default:
            ASSERT_NOT_REACHED();
    }

    m_documentLoader->writer()->setMIMEType(dl->responseMIMEType());

    // Tell the client we've committed this URL.
    ASSERT(m_frame->view());

    if (m_stateMachine.creatingInitialEmptyDocument())
        return;

    if (!m_stateMachine.committedFirstRealDocumentLoad())
        m_stateMachine.advanceTo(FrameLoaderStateMachine::DisplayingInitialEmptyDocumentPostCommit);

    if (!m_client->hasHTMLView())
        receivedFirstData();
}

5，包含一个HistoryController对象，用于操作历史记录相关的接口，保存或者恢复Document和View相关的状态，维护前进后退队列，以实现前进后退功能，前进后退本质上是同Page对象关联的，FrameLoader通过HistoryController操作 m_backForwardController对象

6，包含一个 ResourceLoadNotifier 对象， 主要用于同 ResourceLoader及FrameLoaderClient打交道，可以理解为 ResourceLoader有事件变化或者发生的时候，通知 FrameLoader 的一个手段

7，包含一个SubframeLoader对象，当FrameLoader下载的Document有子帧需要请求的时候（比如 HTMLDocument中解析到 iframe），用于处理子帧请求

8，将FrameLoader的状态封装到 FrameLoaderStateMachine 中，这个状态同 FrameState不同。 FrameState倾向于判断Frame涉及的Document的下载状态，是出于发起状态（Provisional），还是出于已经收到响应但不全（CommitedPage），还是响应收全状态，倾向于同HTTP相关。而 FrameLoaderStateMachine 倾向于同DocumentLoader相关，用来描述FrameLoader处理DocumentLoader的节点，是处于已经创建，还是显示状态。

9，PolicyChecker主要用来对FrameLoader进行一些校验。包括三种校验： NewWindow，Navigation和Content。

    NewWindow对应于浏览器需要打开一个tab页或新窗口的时候

    Navigation 对应于一个页面请求发起的时候

    Content校验对应于收到数据之后（判断 Mime type等）

    PolicyChecker通过提供对应的接口，由FrameLoaderClient来对这些请求进行校验，以确定是否允许继续，护着需要其他的动作

3.   主要接口

void FrameLoader::init()

功能：

      FrameLoader的初始化

实现： 

void FrameLoader::init()
{
    // Propagate sandbox attributes to this Frameloader and its descendants.
    // This needs to be done early, so that an initial document gets correct sandbox flags in its SecurityOrigin.
    updateSandboxFlags();

    // this somewhat odd set of steps is needed to give the frame an initial empty document
    m_stateMachine.advanceTo(FrameLoaderStateMachine::CreatingInitialEmptyDocument);
    setPolicyDocumentLoader(m_client->createDocumentLoader(ResourceRequest(KURL(ParsedURLString, "")), SubstituteData()).get());
    setProvisionalDocumentLoader(m_policyDocumentLoader.get());
    setState(FrameStateProvisional);
    m_provisionalDocumentLoader->setResponse(ResourceResponse(KURL(), "text/html", 0, String(), String()));
    m_provisionalDocumentLoader->finishedLoading();
    m_documentLoader->writer()->begin(KURL(), false);
    m_documentLoader->writer()->end();
    m_frame->document()->cancelParsing();
    m_stateMachine.advanceTo(FrameLoaderStateMachine::DisplayingInitialEmptyDocument);
    m_didCallImplicitClose = true;

    m_networkingContext = m_client->createNetworkingContext();
}

函数调用系列：

QWebFrame::QWebFrame(QwebPage* parent,QWebFrameData *frameData)
QWebFramePrivate::init(QWebFrame* qwebframe,QWebFrameData* frameData)
Frame::init()
FrameLoader::init()

源码追踪：

1，QWebFrame::QWebFrame(QWebPage *parent, QWebFrameData *frameData)

QWebFrame::QWebFrame(QWebPage *parent, QWebFrameData *frameData)
    : QObject(parent)
    , d(new QWebFramePrivate)
{
    d->page = parent;
    d->init(this, frameData);

    if (!frameData->url.isEmpty()) {
        WebCore::ResourceRequest request(frameData->url, frameData->referrer);
        d->frame->loader()->load(request, frameData->name, false);
    }
#if ENABLE(ORIENTATION_EVENTS) && ENABLE(DEVICE_ORIENTATION)
    connect(&d->m_orientation, SIGNAL(readingChanged()), this, SLOT(_q_orientationChanged()));
    d->m_orientation.start();
#endif
}

2，QWebFramePrivate::init(QWebFrame *qframe, QWebFrameData *frameData)

void QWebFramePrivate::init(QWebFrame *qframe, QWebFrameData *frameData)
{
    q = qframe;

    allowsScrolling = frameData->allowsScrolling;
    marginWidth = frameData->marginWidth;
    marginHeight = frameData->marginHeight;
    frame = frameData->frame.get();
    frameLoaderClient = frameData->frameLoaderClient;
    frameLoaderClient->setFrame(qframe, frame);

    frame->init();
}

3，Frame::init()

inline void Frame::init()
{
    m_loader.init();
}

4， FrameLoader::init()

 

OK，源码跟踪到此

说明：

     主要做一些自身的初始化工作，比如初始化状态机，Sandbox Flags，创建DocumentLoader被设置为 PolicyDocumentLoader  和  Provisional DocumentLoader，调用 DocumentLoader  和 documentWriter等的接口进行初始化操作

FrameLoader::commitProvisionalLoad

功能：

    提交Provisional阶段下载的数据

实现：

void FrameLoader::commitProvisionalLoad()
{
    RefPtr<CachedPage> cachedPage = m_loadingFromCachedPage ? pageCache()->get(history()->provisionalItem()) : 0;
    RefPtr<DocumentLoader> pdl = m_provisionalDocumentLoader;

    LOG(PageCache, "WebCoreLoading %s: About to commit provisional load from previous URL '%s' to new URL '%s'", m_frame->tree()->uniqueName().string().utf8().data(),
        m_frame->document() ? m_frame->document()->url().string().utf8().data() : "", 
        pdl ? pdl->url().string().utf8().data() : "<no provisional DocumentLoader>");

    // Check to see if we need to cache the page we are navigating away from into the back/forward cache.
    // We are doing this here because we know for sure that a new page is about to be loaded.
    HistoryItem* item = history()->currentItem();
    if (!m_frame->tree()->parent() && PageCache::canCache(m_frame->page()) && !item->isInPageCache())
        pageCache()->add(item, m_frame->page());

    if (m_loadType != FrameLoadTypeReplace)
        closeOldDataSources();

    if (!cachedPage && !m_stateMachine.creatingInitialEmptyDocument())
        m_client->makeRepresentation(pdl.get());

    transitionToCommitted(cachedPage);

    if (pdl) {
        // Check if the destination page is allowed to access the previous page's timing information.
        RefPtr<SecurityOrigin> securityOrigin = SecurityOrigin::create(pdl->request().url());
        m_documentLoader->timing()->hasSameOriginAsPreviousDocument = securityOrigin->canRequest(m_previousUrl);
    }

    // Call clientRedirectCancelledOrFinished() here so that the frame load delegate is notified that the redirect's
    // status has changed, if there was a redirect.  The frame load delegate may have saved some state about
    // the redirect in its -webView:willPerformClientRedirectToURL:delay:fireDate:forFrame:.  Since we are
    // just about to commit a new page, there cannot possibly be a pending redirect at this point.
    if (m_sentRedirectNotification)
        clientRedirectCancelledOrFinished(false);
    
    if (cachedPage && cachedPage->document()) {
        prepareForCachedPageRestore();
        cachedPage->restore(m_frame->page());

        dispatchDidCommitLoad();

        // If we have a title let the WebView know about it. 
        StringWithDirection title = m_documentLoader->title();
        if (!title.isNull())
            m_client->dispatchDidReceiveTitle(title);

        checkCompleted();
    } else {        
        KURL url = pdl->substituteData().responseURL();
        if (url.isEmpty())
            url = pdl->url();
        if (url.isEmpty())
            url = pdl->responseURL();
        if (url.isEmpty())
            url = blankURL();

        didOpenURL(url);
    }

    LOG(Loading, "WebCoreLoading %s: Finished committing provisional load to URL %s", m_frame->tree()->uniqueName().string().utf8().data(),
        m_frame->document() ? m_frame->document()->url().string().utf8().data() : "");

    if (m_loadType == FrameLoadTypeStandard && m_documentLoader->isClientRedirect())
        history()->updateForClientRedirect();

    if (m_loadingFromCachedPage) {
        m_frame->document()->documentDidBecomeActive();
        
        // Force a layout to update view size and thereby update scrollbars.
        m_frame->view()->forceLayout();

        const ResponseVector& responses = m_documentLoader->responses();
        size_t count = responses.size();
        for (size_t i = 0; i < count; i++) {
            const ResourceResponse& response = responses[i];
            // FIXME: If the WebKit client changes or cancels the request, this is not respected.
            ResourceError error;
            unsigned long identifier;
            ResourceRequest request(response.url());
            requestFromDelegate(request, identifier, error);
            // FIXME: If we get a resource with more than 2B bytes, this code won't do the right thing.
            // However, with today's computers and networking speeds, this won't happen in practice.
            // Could be an issue with a giant local file.
            notifier()->sendRemainingDelegateMessages(m_documentLoader.get(), identifier, response, static_cast<int>(response.expectedContentLength()), 0, error);
        }
        
        pageCache()->remove(history()->currentItem());

        m_documentLoader->setPrimaryLoadComplete(true);

        // FIXME: Why only this frame and not parent frames?
        checkLoadCompleteForThisFrame();
    }
}

函数调用系列：（两种情况）

DocumentLoader::finishLoading
DocumentLoader::commitIfReady
FrameLoader::commitProvisionalLoad


ResourceLoader::didReceiveData
MainResourceLoader::addData
DocumentLoader::receiveData
DocumentLoader::commitLoad
DocumentLoader::commitIfReady
DocumentLoader::commitProvisionalLoad

源码跟踪：

情况一

1，MainResourceLoader::didFinishLoading(double finishTime)

void MainResourceLoader::didFinishLoading(double finishTime)
{
    // There is a bug in CFNetwork where callbacks can be dispatched even when loads are deferred.
    // See <rdar://problem/6304600> for more details.
#if !USE(CF)
    ASSERT(shouldLoadAsEmptyDocument(frameLoader()->activeDocumentLoader()->url()) || !defersLoading());
#endif
    
    // The additional processing can do anything including possibly removing the last
    // reference to this object.
    RefPtr<MainResourceLoader> protect(this);

#if ENABLE(OFFLINE_WEB_APPLICATIONS)
    RefPtr<DocumentLoader> dl = documentLoader();
#endif

    ASSERT(!documentLoader()->timing()->responseEnd);
    documentLoader()->timing()->responseEnd = finishTime ? finishTime : (m_timeOfLastDataReceived ? m_timeOfLastDataReceived : currentTime());
    frameLoader()->finishedLoading();
    ResourceLoader::didFinishLoading(finishTime);
    
#if ENABLE(OFFLINE_WEB_APPLICATIONS)
    dl->applicationCacheHost()->finishedLoadingMainResource();
#endif
}

2，FrameLoader::finishedLoading()

void FrameLoader::finishedLoading()
{
    // Retain because the stop may release the last reference to it.
    RefPtr<Frame> protect(m_frame);

    RefPtr<DocumentLoader> dl = activeDocumentLoader();
    dl->finishedLoading();
    if (!dl->mainDocumentError().isNull() || !dl->frameLoader())
        return;
    dl->setPrimaryLoadComplete(true);
    m_client->dispatchDidLoadMainResource(dl.get());
    checkLoadComplete();
}

3，DocumentLoader::finishedLoading()

void DocumentLoader::finishedLoading()
{
    m_gotFirstByte = true;   
    commitIfReady();
    if (FrameLoader* loader = frameLoader()) {
        loader->finishedLoadingDocument(this);
        m_writer.end();
    }
}

4， DocumentLoader::commitIfReady()

void DocumentLoader::commitIfReady()
{
    if (m_gotFirstByte && !m_committed) {
        m_committed = true;
        frameLoader()->commitProvisionalLoad();
    }
}

5， FrameLoader::commitProvisionalLoad()

OK！

情况二：

1，MainResourceLoader::didReceiveData(const char* data, int length, long long encodedDataLength, bool allAtOnce)

void MainResourceLoader::didReceiveData(const char* data, int length, long long encodedDataLength, bool allAtOnce)
{
    ASSERT(data);
    ASSERT(length != 0);

    ASSERT(!m_response.isNull());

#if USE(CFNETWORK) || PLATFORM(MAC)
    // Workaround for <rdar://problem/6060782>
    if (m_response.isNull()) {
        m_response = ResourceResponse(KURL(), "text/html", 0, String(), String());
        if (DocumentLoader* documentLoader = frameLoader()->activeDocumentLoader())
            documentLoader->setResponse(m_response);
    }
#endif

    // There is a bug in CFNetwork where callbacks can be dispatched even when loads are deferred.
    // See <rdar://problem/6304600> for more details.
#if !USE(CF)
    ASSERT(!defersLoading());
#endif
 
 #if ENABLE(OFFLINE_WEB_APPLICATIONS)
    documentLoader()->applicationCacheHost()->mainResourceDataReceived(data, length, encodedDataLength, allAtOnce);
#endif

    // The additional processing can do anything including possibly removing the last
    // reference to this object; one example of this is 3266216.
    RefPtr<MainResourceLoader> protect(this);

    m_timeOfLastDataReceived = currentTime();

    ResourceLoader::didReceiveData(data, length, encodedDataLength, allAtOnce);
}

2，ResourceLoader::didReceiveData(const char* data, int length, long long encodedDataLength, bool allAtOnce)

void ResourceLoader::didReceiveData(const char* data, int length, long long encodedDataLength, bool allAtOnce)
{
    // The following assertions are not quite valid here, since a subclass
    // might override didReceiveData in a way that invalidates them. This
    // happens with the steps listed in 3266216
    // ASSERT(con == connection);
    // ASSERT(!m_reachedTerminalState);

    // Protect this in this delegate method since the additional processing can do
    // anything including possibly derefing this; one example of this is Radar 3266216.
    RefPtr<ResourceLoader> protector(this);

    addData(data, length, allAtOnce);
    // FIXME: If we get a resource with more than 2B bytes, this code won't do the right thing.
    // However, with today's computers and networking speeds, this won't happen in practice.
    // Could be an issue with a giant local file.
    if (m_sendResourceLoadCallbacks && m_frame)
        frameLoader()->notifier()->didReceiveData(this, data, length, static_cast<int>(encodedDataLength));
}

3，MainResourceLoader::addData(const char* data, int length, bool allAtOnce)

void MainResourceLoader::addData(const char* data, int length, bool allAtOnce)
{
    ResourceLoader::addData(data, length, allAtOnce);
    documentLoader()->receivedData(data, length);
}

4，DocumentLoader::receivedData(const char* data, int length)

void DocumentLoader::receivedData(const char* data, int length)
{    
    m_gotFirstByte = true;
    if (doesProgressiveLoad(m_response.mimeType()))
        commitLoad(data, length);
}

5，DocumentLoader::commitLoad(const char* data, int length)

void DocumentLoader::commitLoad(const char* data, int length)
{
    // Both unloading the old page and parsing the new page may execute JavaScript which destroys the datasource
    // by starting a new load, so retain temporarily.
    RefPtr<Frame> protectFrame(m_frame);
    RefPtr<DocumentLoader> protectLoader(this);

    commitIfReady();
    FrameLoader* frameLoader = DocumentLoader::frameLoader();
    if (!frameLoader)
        return;
#if ENABLE(WEB_ARCHIVE)
    if (ArchiveFactory::isArchiveMimeType(response().mimeType()))
        return;
#endif
    frameLoader->client()->committedLoad(this, data, length);
}

6， DocumentLoader::commitIfReady()

void DocumentLoader::commitIfReady()
{
    if (m_gotFirstByte && !m_committed) {
        m_committed = true;
        frameLoader()->commitProvisionalLoad();
    }
}

7，FrameLoader::commitProvisionalLoad()

OK， 跟踪over！！

说明： 这个接口主要的操作是将Provisional DocumentLoader设置成 DocumentLoader，因为已经收到数据，所以FrameState也会跃迁到FrameStateCommittedPage。还有历史记录，PageCache相关的操作。另外，这个接口会间接调用 FrameLoaderClientQt::transitionToCommittedForNewPage，通过Frame::createView 创建出 FrameView来。

Frame::finishedLoading

功能：

     frame请求网络加载完成的时候调用此接口

实现：

void FrameLoader::finishedLoading()
{
    // Retain because the stop may release the last reference to it.
    RefPtr<Frame> protect(m_frame);

    RefPtr<DocumentLoader> dl = activeDocumentLoader();
    dl->finishedLoading();
    if (!dl->mainDocumentError().isNull() || !dl->frameLoader())
        return;
    dl->setPrimaryLoadComplete(true);
    m_client->dispatchDidLoadMainResource(dl.get());
    checkLoadComplete();
}

函数调用系列：

ResourceLoader::didFinishLoading
MainResourceLoader::didFinishLoading
FrameLoader::finishedLoading
FrameLoader::init()

说明：

    检查是否有网络错误，告诉DocumentLoader和DocumentWriter下载完成，以便进行后续操作（提交数据，解析）

    

FrameLoader::finishedParsing

功能：

     解析完成调用此接口

函数调用系列：

DocumentWritter::end
….
Document::finishParsing
….
Document::finishedParsing
FrameLoader::finishedParsing

源码跟踪：

1，DocumentLoader::finishedLoading()

void DocumentLoader::finishedLoading()
{
    m_gotFirstByte = true;   
    commitIfReady();
    if (FrameLoader* loader = frameLoader()) {
        loader->finishedLoadingDocument(this);
        m_writer.end();
    }
}

2， DocumentWriter::end()

void DocumentWriter::end()
{
    m_frame->loader()->didEndDocument();
    endIfNotLoadingMainResource();
}

3，DocumentWriter::endIfNotLoadingMainResource()

void DocumentWriter::endIfNotLoadingMainResource()
{
    if (m_frame->loader()->isLoadingMainResource() || !m_frame->page() || !m_frame->document())
        return;

    // http://bugs.webkit.org/show_bug.cgi?id=10854
    // The frame's last ref may be removed and it can be deleted by checkCompleted(), 
    // so we'll add a protective refcount
    RefPtr<Frame> protector(m_frame);

    // make sure nothing's left in there
    addData(0, 0, true);
    m_frame->document()->finishParsing();
}

4，Document::finishedParsing()

void Document::finishedParsing()
{
    ASSERT(!scriptableDocumentParser() || !m_parser->isParsing());
    ASSERT(!scriptableDocumentParser() || m_readyState != Loading);
    setParsing(false);
    if (!m_documentTiming.domContentLoadedEventStart)
        m_documentTiming.domContentLoadedEventStart = currentTime();
    dispatchEvent(Event::create(eventNames().DOMContentLoadedEvent, true, false));
    if (!m_documentTiming.domContentLoadedEventEnd)
        m_documentTiming.domContentLoadedEventEnd = currentTime();

    if (RefPtr<Frame> f = frame()) {
        // FrameLoader::finishedParsing() might end up calling Document::implicitClose() if all
        // resource loads are complete. HTMLObjectElements can start loading their resources from
        // post attach callbacks triggered by recalcStyle().  This means if we parse out an <object>
        // tag and then reach the end of the document without updating styles, we might not have yet
        // started the resource load and might fire the window load event too early.  To avoid this
        // we force the styles to be up to date before calling FrameLoader::finishedParsing().
        // See https://bugs.webkit.org/show_bug.cgi?id=36864 starting around comment 35.
        updateStyleIfNeeded();

        f->loader()->finishedParsing();

        InspectorInstrumentation::domContentLoadedEventFired(f.get(), url());
    }
}

5，FrameLoader::finishedParsing()

FrameLoader::load(const ResourceRequest& request,bool lockHistory)

功能：

     加载一个Frame请求，Frame请求相关的数据，封装成ResourceRequest传入

函数调用系列：

     一般由应用触发

源码跟踪：

     见前文

说明：

     这个接口调用FrameLoaderClientQt::createDocumentLoader创建出DocumentLoader，并以此DocumentLoader为Policy  DocumentLoder，进入Policy check流程