dom/base/nsDocumentEncoder.cpp
author Mike Hommey <mh+mozilla@glandium.org>
Thu, 17 Jan 2019 13:18:37 +0000
changeset 454358 c28522aaf3918c5f556c65432f8bf14a65bff65a
parent 452446 f0a91d36587266d7454a450c6044d573664fbed5
child 463734 19f3dfdf76da0c37c018683808add3d19f192e56
permissions -rw-r--r--
Bug 1520149 - Also disable the machine outliner on Android when LTO is enabled. r=dmajor Differential Revision: https://phabricator.services.mozilla.com/D16790

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

/*
 * Object that can be used to serialize selections, ranges, or nodes
 * to strings in a gazillion different ways.
 */

#include "nsIDocumentEncoder.h"

#include "nscore.h"
#include "nsIFactory.h"
#include "nsISupports.h"
#include "mozilla/dom/Document.h"
#include "nsIHTMLDocument.h"
#include "nsCOMPtr.h"
#include "nsIContentSerializer.h"
#include "mozilla/Encoding.h"
#include "nsIOutputStream.h"
#include "nsRange.h"
#include "nsGkAtoms.h"
#include "nsIContent.h"
#include "nsIScriptContext.h"
#include "nsIScriptGlobalObject.h"
#include "nsIScriptSecurityManager.h"
#include "mozilla/dom/Selection.h"
#include "nsITransferable.h"  // for kUnicodeMime
#include "nsContentUtils.h"
#include "nsElementTable.h"
#include "nsNodeUtils.h"
#include "nsUnicharUtils.h"
#include "nsReadableUtils.h"
#include "nsTArray.h"
#include "nsIFrame.h"
#include "nsStringBuffer.h"
#include "mozilla/dom/Comment.h"
#include "mozilla/dom/DocumentType.h"
#include "mozilla/dom/Element.h"
#include "mozilla/dom/ProcessingInstruction.h"
#include "mozilla/dom/ShadowRoot.h"
#include "mozilla/dom/Text.h"
#include "nsLayoutUtils.h"
#include "mozilla/ScopeExit.h"

using namespace mozilla;
using namespace mozilla::dom;

enum nsRangeIterationDirection { kDirectionOut = -1, kDirectionIn = 1 };

class nsDocumentEncoder : public nsIDocumentEncoder {
 public:
  nsDocumentEncoder();

  NS_DECL_CYCLE_COLLECTING_ISUPPORTS
  NS_DECL_CYCLE_COLLECTION_CLASS(nsDocumentEncoder)
  NS_DECL_NSIDOCUMENTENCODER

 protected:
  virtual ~nsDocumentEncoder();

  void Initialize(bool aClearCachedSerializer = true);
  nsresult SerializeNodeStart(nsINode* aNode, int32_t aStartOffset,
                              int32_t aEndOffset, nsAString& aStr,
                              nsINode* aOriginalNode = nullptr);
  nsresult SerializeToStringRecursive(nsINode* aNode, nsAString& aStr,
                                      bool aDontSerializeRoot,
                                      uint32_t aMaxLength = 0);
  nsresult SerializeNodeEnd(nsINode* aNode, nsAString& aStr);
  // This serializes the content of aNode.
  nsresult SerializeToStringIterative(nsINode* aNode, nsAString& aStr);
  nsresult SerializeRangeToString(nsRange* aRange, nsAString& aOutputString);
  nsresult SerializeRangeNodes(nsRange* aRange, nsINode* aNode,
                               nsAString& aString, int32_t aDepth);
  nsresult SerializeRangeContextStart(const nsTArray<nsINode*>& aAncestorArray,
                                      nsAString& aString);
  nsresult SerializeRangeContextEnd(nsAString& aString);

  virtual int32_t GetImmediateContextCount(
      const nsTArray<nsINode*>& aAncestorArray) {
    return -1;
  }

  nsresult FlushText(nsAString& aString, bool aForce);

  bool IsVisibleNode(nsINode* aNode) {
    MOZ_ASSERT(aNode, "null node");

    if (mFlags & SkipInvisibleContent) {
      // Treat the visibility of the ShadowRoot as if it were
      // the host content.
      //
      // FIXME(emilio): I suspect instead of this a bunch of the GetParent()
      // calls here should be doing GetFlattenedTreeParent, then this condition
      // should be unreachable...
      if (ShadowRoot* shadowRoot = ShadowRoot::FromNode(aNode)) {
        aNode = shadowRoot->GetHost();
      }

      if (aNode->IsContent()) {
        nsIFrame* frame = aNode->AsContent()->GetPrimaryFrame();
        if (!frame) {
          if (aNode->IsElement() && aNode->AsElement()->IsDisplayContents()) {
            return true;
          }
          if (aNode->IsText()) {
            // We have already checked that our parent is visible.
            //
            // FIXME(emilio): Text not assigned to a <slot> in Shadow DOM should
            // probably return false...
            return true;
          }
          if (aNode->IsHTMLElement(nsGkAtoms::rp)) {
            // Ruby parentheses are part of ruby structure, hence
            // shouldn't be stripped out even if it is not displayed.
            return true;
          }
          return false;
        }
        bool isVisible = frame->StyleVisibility()->IsVisible();
        if (!isVisible && aNode->IsText()) return false;
      }
    }
    return true;
  }

  virtual bool IncludeInContext(nsINode* aNode);

  void Clear();

  class MOZ_STACK_CLASS AutoReleaseDocumentIfNeeded final {
   public:
    explicit AutoReleaseDocumentIfNeeded(nsDocumentEncoder* aEncoder)
        : mEncoder(aEncoder) {}

    ~AutoReleaseDocumentIfNeeded() {
      if (mEncoder->mFlags & RequiresReinitAfterOutput) {
        mEncoder->Clear();
      }
    }

   private:
    nsDocumentEncoder* mEncoder;
  };

  nsCOMPtr<Document> mDocument;
  RefPtr<Selection> mSelection;
  RefPtr<nsRange> mRange;
  nsCOMPtr<nsINode> mNode;
  nsCOMPtr<nsIOutputStream> mStream;
  nsCOMPtr<nsIContentSerializer> mSerializer;
  UniquePtr<Encoder> mUnicodeEncoder;
  nsCOMPtr<nsINode> mCommonParent;
  nsCOMPtr<nsIDocumentEncoderNodeFixup> mNodeFixup;

  nsString mMimeType;
  const Encoding* mEncoding;
  uint32_t mFlags;
  uint32_t mWrapColumn;
  uint32_t mStartDepth;
  uint32_t mEndDepth;
  int32_t mStartRootIndex;
  int32_t mEndRootIndex;
  AutoTArray<nsINode*, 8> mCommonAncestors;
  AutoTArray<nsIContent*, 8> mStartNodes;
  AutoTArray<int32_t, 8> mStartOffsets;
  AutoTArray<nsIContent*, 8> mEndNodes;
  AutoTArray<int32_t, 8> mEndOffsets;
  AutoTArray<AutoTArray<nsINode*, 8>, 8> mRangeContexts;
  // Whether the serializer cares about being notified to scan elements to
  // keep track of whether they are preformatted.  This stores the out
  // argument of nsIContentSerializer::Init().
  bool mNeedsPreformatScanning;
  bool mHaltRangeHint;
  // Used when context has already been serialized for
  // table cell selections (where parent is <tr>)
  bool mDisableContextSerialize;
  bool mIsCopying;  // Set to true only while copying
  bool mNodeIsContainer;
  bool mIsPlainText;
  nsStringBuffer* mCachedBuffer;
};

NS_IMPL_CYCLE_COLLECTING_ADDREF(nsDocumentEncoder)
NS_IMPL_CYCLE_COLLECTING_RELEASE_WITH_LAST_RELEASE(nsDocumentEncoder, Clear())

NS_INTERFACE_MAP_BEGIN_CYCLE_COLLECTION(nsDocumentEncoder)
  NS_INTERFACE_MAP_ENTRY(nsIDocumentEncoder)
  NS_INTERFACE_MAP_ENTRY(nsISupports)
NS_INTERFACE_MAP_END

NS_IMPL_CYCLE_COLLECTION(nsDocumentEncoder, mDocument, mSelection, mRange,
                         mNode, mSerializer, mCommonParent)

nsDocumentEncoder::nsDocumentEncoder()
    : mEncoding(nullptr), mIsCopying(false), mCachedBuffer(nullptr) {
  Initialize();
  mMimeType.AssignLiteral("text/plain");
}

void nsDocumentEncoder::Initialize(bool aClearCachedSerializer) {
  mFlags = 0;
  mWrapColumn = 72;
  mStartDepth = 0;
  mEndDepth = 0;
  mStartRootIndex = 0;
  mEndRootIndex = 0;
  mNeedsPreformatScanning = false;
  mHaltRangeHint = false;
  mDisableContextSerialize = false;
  mNodeIsContainer = false;
  mIsPlainText = false;
  if (aClearCachedSerializer) {
    mSerializer = nullptr;
  }
}

nsDocumentEncoder::~nsDocumentEncoder() {
  if (mCachedBuffer) {
    mCachedBuffer->Release();
  }
}

NS_IMETHODIMP
nsDocumentEncoder::Init(Document* aDocument, const nsAString& aMimeType,
                        uint32_t aFlags) {
  return NativeInit(aDocument, aMimeType, aFlags);
}

NS_IMETHODIMP
nsDocumentEncoder::NativeInit(Document* aDocument, const nsAString& aMimeType,
                              uint32_t aFlags) {
  if (!aDocument) return NS_ERROR_INVALID_ARG;

  Initialize(!mMimeType.Equals(aMimeType));

  mDocument = aDocument;

  mMimeType = aMimeType;

  mFlags = aFlags;
  mIsCopying = false;

  return NS_OK;
}

NS_IMETHODIMP
nsDocumentEncoder::SetWrapColumn(uint32_t aWC) {
  mWrapColumn = aWC;
  return NS_OK;
}

NS_IMETHODIMP
nsDocumentEncoder::SetSelection(Selection* aSelection) {
  mSelection = aSelection;
  return NS_OK;
}

NS_IMETHODIMP
nsDocumentEncoder::SetRange(nsRange* aRange) {
  mRange = aRange;
  return NS_OK;
}

NS_IMETHODIMP
nsDocumentEncoder::SetNode(nsINode* aNode) {
  mNodeIsContainer = false;
  mNode = aNode;
  return NS_OK;
}

NS_IMETHODIMP
nsDocumentEncoder::SetContainerNode(nsINode* aContainer) {
  mNodeIsContainer = true;
  mNode = aContainer;
  return NS_OK;
}

NS_IMETHODIMP
nsDocumentEncoder::SetCharset(const nsACString& aCharset) {
  const Encoding* encoding = Encoding::ForLabel(aCharset);
  if (!encoding) {
    return NS_ERROR_UCONV_NOCONV;
  }
  mEncoding = encoding->OutputEncoding();
  return NS_OK;
}

NS_IMETHODIMP
nsDocumentEncoder::GetMimeType(nsAString& aMimeType) {
  aMimeType = mMimeType;
  return NS_OK;
}

bool nsDocumentEncoder::IncludeInContext(nsINode* aNode) { return false; }

nsresult nsDocumentEncoder::SerializeNodeStart(nsINode* aNode,
                                               int32_t aStartOffset,
                                               int32_t aEndOffset,
                                               nsAString& aStr,
                                               nsINode* aOriginalNode) {
  if (mNeedsPreformatScanning && aNode->IsElement()) {
    mSerializer->ScanElementForPreformat(aNode->AsElement());
  }

  if (!IsVisibleNode(aNode)) return NS_OK;

  nsINode* node = nullptr;
  nsCOMPtr<nsINode> fixedNodeKungfuDeathGrip;

  // Caller didn't do fixup, so we'll do it ourselves
  if (!aOriginalNode) {
    aOriginalNode = aNode;
    if (mNodeFixup) {
      bool dummy;
      mNodeFixup->FixupNode(aNode, &dummy,
                            getter_AddRefs(fixedNodeKungfuDeathGrip));
      node = fixedNodeKungfuDeathGrip;
    }
  }

  // Either there was no fixed-up node,
  // or the caller did fixup themselves and aNode is already fixed
  if (!node) node = aNode;

  if (node->IsElement()) {
    if ((mFlags & (nsIDocumentEncoder::OutputPreformatted |
                   nsIDocumentEncoder::OutputDropInvisibleBreak)) &&
        nsLayoutUtils::IsInvisibleBreak(node)) {
      return NS_OK;
    }
    Element* originalElement = Element::FromNodeOrNull(aOriginalNode);
    mSerializer->AppendElementStart(node->AsElement(), originalElement, aStr);
    return NS_OK;
  }

  switch (node->NodeType()) {
    case nsINode::TEXT_NODE: {
      mSerializer->AppendText(static_cast<nsIContent*>(node), aStartOffset,
                              aEndOffset, aStr);
      break;
    }
    case nsINode::CDATA_SECTION_NODE: {
      mSerializer->AppendCDATASection(static_cast<nsIContent*>(node),
                                      aStartOffset, aEndOffset, aStr);
      break;
    }
    case nsINode::PROCESSING_INSTRUCTION_NODE: {
      mSerializer->AppendProcessingInstruction(
          static_cast<ProcessingInstruction*>(node), aStartOffset, aEndOffset,
          aStr);
      break;
    }
    case nsINode::COMMENT_NODE: {
      mSerializer->AppendComment(static_cast<Comment*>(node), aStartOffset,
                                 aEndOffset, aStr);
      break;
    }
    case nsINode::DOCUMENT_TYPE_NODE: {
      mSerializer->AppendDoctype(static_cast<DocumentType*>(node), aStr);
      break;
    }
  }

  return NS_OK;
}

nsresult nsDocumentEncoder::SerializeNodeEnd(nsINode* aNode, nsAString& aStr) {
  if (mNeedsPreformatScanning && aNode->IsElement()) {
    mSerializer->ForgetElementForPreformat(aNode->AsElement());
  }

  if (!IsVisibleNode(aNode)) return NS_OK;

  if (aNode->IsElement()) {
    mSerializer->AppendElementEnd(aNode->AsElement(), aStr);
  }
  return NS_OK;
}

nsresult nsDocumentEncoder::SerializeToStringRecursive(nsINode* aNode,
                                                       nsAString& aStr,
                                                       bool aDontSerializeRoot,
                                                       uint32_t aMaxLength) {
  if (aMaxLength > 0 && aStr.Length() >= aMaxLength) {
    return NS_OK;
  }

  if (!IsVisibleNode(aNode)) return NS_OK;

  nsresult rv = NS_OK;
  bool serializeClonedChildren = false;
  nsINode* maybeFixedNode = nullptr;

  // Keep the node from FixupNode alive.
  nsCOMPtr<nsINode> fixedNodeKungfuDeathGrip;
  if (mNodeFixup) {
    mNodeFixup->FixupNode(aNode, &serializeClonedChildren,
                          getter_AddRefs(fixedNodeKungfuDeathGrip));
    maybeFixedNode = fixedNodeKungfuDeathGrip;
  }

  if (!maybeFixedNode) maybeFixedNode = aNode;

  if ((mFlags & SkipInvisibleContent) &&
      !(mFlags & OutputNonTextContentAsPlaceholder)) {
    if (aNode->IsContent()) {
      if (nsIFrame* frame = aNode->AsContent()->GetPrimaryFrame()) {
        if (!frame->IsSelectable(nullptr)) {
          aDontSerializeRoot = true;
        }
      }
    }
  }

  if (!aDontSerializeRoot) {
    int32_t endOffset = -1;
    if (aMaxLength > 0) {
      MOZ_ASSERT(aMaxLength >= aStr.Length());
      endOffset = aMaxLength - aStr.Length();
    }
    rv = SerializeNodeStart(maybeFixedNode, 0, endOffset, aStr, aNode);
    NS_ENSURE_SUCCESS(rv, rv);
  }

  nsINode* node = serializeClonedChildren ? maybeFixedNode : aNode;

  for (nsINode* child = nsNodeUtils::GetFirstChildOfTemplateOrNode(node); child;
       child = child->GetNextSibling()) {
    rv = SerializeToStringRecursive(child, aStr, false, aMaxLength);
    NS_ENSURE_SUCCESS(rv, rv);
  }

  if (!aDontSerializeRoot) {
    rv = SerializeNodeEnd(maybeFixedNode, aStr);
    NS_ENSURE_SUCCESS(rv, rv);
  }

  return FlushText(aStr, false);
}

nsresult nsDocumentEncoder::SerializeToStringIterative(nsINode* aNode,
                                                       nsAString& aStr) {
  nsresult rv;

  nsINode* node = nsNodeUtils::GetFirstChildOfTemplateOrNode(aNode);
  while (node) {
    nsINode* current = node;
    rv = SerializeNodeStart(current, 0, -1, aStr, current);
    NS_ENSURE_SUCCESS(rv, rv);
    node = nsNodeUtils::GetFirstChildOfTemplateOrNode(current);
    while (!node && current && current != aNode) {
      rv = SerializeNodeEnd(current, aStr);
      NS_ENSURE_SUCCESS(rv, rv);
      // Check if we have siblings.
      node = current->GetNextSibling();
      if (!node) {
        // Perhaps parent node has siblings.
        current = current->GetParentNode();

        // Handle template element. If the parent is a template's content,
        // then adjust the parent to be the template element.
        if (current && current != aNode && current->IsDocumentFragment()) {
          nsIContent* host = current->AsDocumentFragment()->GetHost();
          if (host && host->IsHTMLElement(nsGkAtoms::_template)) {
            current = host;
          }
        }
      }
    }
  }

  return NS_OK;
}

static nsresult ConvertAndWrite(const nsAString& aString,
                                nsIOutputStream* aStream, Encoder* aEncoder,
                                bool aIsPlainText) {
  NS_ENSURE_ARG_POINTER(aStream);
  NS_ENSURE_ARG_POINTER(aEncoder);

  if (!aString.Length()) {
    return NS_OK;
  }

  uint8_t buffer[4096];
  auto src = MakeSpan(aString);
  auto bufferSpan = MakeSpan(buffer);
  // Reserve space for terminator
  auto dst = bufferSpan.To(bufferSpan.Length() - 1);
  for (;;) {
    uint32_t result;
    size_t read;
    size_t written;
    bool hadErrors;
    if (aIsPlainText) {
      Tie(result, read, written) =
          aEncoder->EncodeFromUTF16WithoutReplacement(src, dst, false);
      if (result != kInputEmpty && result != kOutputFull) {
        // There's always room for one byte in the case of
        // an unmappable character, because otherwise
        // we'd have gotten `kOutputFull`.
        dst[written++] = '?';
      }
    } else {
      Tie(result, read, written, hadErrors) =
          aEncoder->EncodeFromUTF16(src, dst, false);
    }
    Unused << hadErrors;
    src = src.From(read);
    // Sadly, we still have test cases that implement nsIOutputStream in JS, so
    // the buffer needs to be zero-terminated for XPConnect to do its thing.
    // See bug 170416.
    bufferSpan[written] = 0;
    uint32_t streamWritten;
    nsresult rv = aStream->Write(reinterpret_cast<char*>(dst.Elements()),
                                 written, &streamWritten);
    if (NS_FAILED(rv)) {
      return rv;
    }
    if (result == kInputEmpty) {
      return NS_OK;
    }
  }
}

nsresult nsDocumentEncoder::FlushText(nsAString& aString, bool aForce) {
  if (!mStream) return NS_OK;

  nsresult rv = NS_OK;

  if (aString.Length() > 1024 || aForce) {
    rv = ConvertAndWrite(aString, mStream, mUnicodeEncoder.get(), mIsPlainText);

    aString.Truncate();
  }

  return rv;
}

static bool IsTextNode(nsINode* aNode) { return aNode && aNode->IsText(); }

nsresult nsDocumentEncoder::SerializeRangeNodes(nsRange* aRange, nsINode* aNode,
                                                nsAString& aString,
                                                int32_t aDepth) {
  nsCOMPtr<nsIContent> content = do_QueryInterface(aNode);
  NS_ENSURE_TRUE(content, NS_ERROR_FAILURE);

  if (!IsVisibleNode(aNode)) return NS_OK;

  nsresult rv = NS_OK;

  // get start and end nodes for this recursion level
  nsCOMPtr<nsIContent> startNode, endNode;
  {
    int32_t start = mStartRootIndex - aDepth;
    if (start >= 0 && (uint32_t)start <= mStartNodes.Length())
      startNode = mStartNodes[start];

    int32_t end = mEndRootIndex - aDepth;
    if (end >= 0 && (uint32_t)end <= mEndNodes.Length())
      endNode = mEndNodes[end];
  }

  if (startNode != content && endNode != content) {
    // node is completely contained in range.  Serialize the whole subtree
    // rooted by this node.
    rv = SerializeToStringRecursive(aNode, aString, false);
    NS_ENSURE_SUCCESS(rv, rv);
  } else {
    // due to implementation it is impossible for text node to be both start and
    // end of range.  We would have handled that case without getting here.
    // XXXsmaug What does this all mean?
    if (IsTextNode(aNode)) {
      if (startNode == content) {
        int32_t startOffset = aRange->StartOffset();
        rv = SerializeNodeStart(aNode, startOffset, -1, aString);
        NS_ENSURE_SUCCESS(rv, rv);
      } else {
        int32_t endOffset = aRange->EndOffset();
        rv = SerializeNodeStart(aNode, 0, endOffset, aString);
        NS_ENSURE_SUCCESS(rv, rv);
      }
    } else {
      if (aNode != mCommonParent) {
        if (IncludeInContext(aNode)) {
          // halt the incrementing of mStartDepth/mEndDepth.  This is
          // so paste client will include this node in paste.
          mHaltRangeHint = true;
        }
        if ((startNode == content) && !mHaltRangeHint) mStartDepth++;
        if ((endNode == content) && !mHaltRangeHint) mEndDepth++;

        // serialize the start of this node
        rv = SerializeNodeStart(aNode, 0, -1, aString);
        NS_ENSURE_SUCCESS(rv, rv);
      }

      // do some calculations that will tell us which children of this
      // node are in the range.
      int32_t startOffset = 0, endOffset = -1;
      if (startNode == content && mStartRootIndex >= aDepth)
        startOffset = mStartOffsets[mStartRootIndex - aDepth];
      if (endNode == content && mEndRootIndex >= aDepth)
        endOffset = mEndOffsets[mEndRootIndex - aDepth];
      // generated content will cause offset values of -1 to be returned.
      uint32_t childCount = content->GetChildCount();

      if (startOffset == -1) startOffset = 0;
      if (endOffset == -1)
        endOffset = childCount;
      else {
        // if we are at the "tip" of the selection, endOffset is fine.
        // otherwise, we need to add one.  This is because of the semantics
        // of the offset list created by GetAncestorsAndOffsets().  The
        // intermediate points on the list use the endOffset of the
        // location of the ancestor, rather than just past it.  So we need
        // to add one here in order to include it in the children we serialize.
        if (aNode != aRange->GetEndContainer()) {
          endOffset++;
        }
      }

      if (endOffset) {
        // serialize the children of this node that are in the range
        nsIContent* childAsNode = content->GetFirstChild();
        int32_t j = 0;

        for (; j < startOffset && childAsNode; ++j) {
          childAsNode = childAsNode->GetNextSibling();
        }

        NS_ENSURE_TRUE(!!childAsNode, NS_ERROR_FAILURE);
        MOZ_ASSERT(j == startOffset);

        for (; childAsNode && j < endOffset; ++j) {
          if ((j == startOffset) || (j == endOffset - 1)) {
            rv = SerializeRangeNodes(aRange, childAsNode, aString, aDepth + 1);
          } else {
            rv = SerializeToStringRecursive(childAsNode, aString, false);
          }

          NS_ENSURE_SUCCESS(rv, rv);

          childAsNode = childAsNode->GetNextSibling();
        }
      }

      // serialize the end of this node
      if (aNode != mCommonParent) {
        rv = SerializeNodeEnd(aNode, aString);
        NS_ENSURE_SUCCESS(rv, rv);
      }
    }
  }
  return NS_OK;
}

nsresult nsDocumentEncoder::SerializeRangeContextStart(
    const nsTArray<nsINode*>& aAncestorArray, nsAString& aString) {
  if (mDisableContextSerialize) {
    return NS_OK;
  }

  AutoTArray<nsINode*, 8>* serializedContext = mRangeContexts.AppendElement();

  int32_t i = aAncestorArray.Length(), j;
  nsresult rv = NS_OK;

  // currently only for table-related elements; see Bug 137450
  j = GetImmediateContextCount(aAncestorArray);

  while (i > 0) {
    nsINode* node = aAncestorArray.ElementAt(--i);

    if (!node) break;

    // Either a general inclusion or as immediate context
    if (IncludeInContext(node) || i < j) {
      rv = SerializeNodeStart(node, 0, -1, aString);
      serializedContext->AppendElement(node);
      if (NS_FAILED(rv)) break;
    }
  }

  return rv;
}

nsresult nsDocumentEncoder::SerializeRangeContextEnd(nsAString& aString) {
  if (mDisableContextSerialize) {
    return NS_OK;
  }

  MOZ_RELEASE_ASSERT(!mRangeContexts.IsEmpty(),
                     "Tried to end context without starting one.");
  AutoTArray<nsINode*, 8>& serializedContext = mRangeContexts.LastElement();

  nsresult rv = NS_OK;
  for (nsINode* node : Reversed(serializedContext)) {
    rv = SerializeNodeEnd(node, aString);

    if (NS_FAILED(rv)) break;
  }

  mRangeContexts.RemoveLastElement();
  return rv;
}

nsresult nsDocumentEncoder::SerializeRangeToString(nsRange* aRange,
                                                   nsAString& aOutputString) {
  if (!aRange || aRange->Collapsed()) return NS_OK;

  mCommonParent = aRange->GetCommonAncestor();

  if (!mCommonParent) return NS_OK;

  nsINode* startContainer = aRange->GetStartContainer();
  NS_ENSURE_TRUE(startContainer, NS_ERROR_FAILURE);
  int32_t startOffset = aRange->StartOffset();

  nsINode* endContainer = aRange->GetEndContainer();
  NS_ENSURE_TRUE(endContainer, NS_ERROR_FAILURE);
  int32_t endOffset = aRange->EndOffset();

  mStartDepth = mEndDepth = 0;
  mCommonAncestors.Clear();
  mStartNodes.Clear();
  mStartOffsets.Clear();
  mEndNodes.Clear();
  mEndOffsets.Clear();

  nsContentUtils::GetAncestors(mCommonParent, mCommonAncestors);
  nsContentUtils::GetAncestorsAndOffsets(startContainer, startOffset,
                                         &mStartNodes, &mStartOffsets);
  nsContentUtils::GetAncestorsAndOffsets(endContainer, endOffset, &mEndNodes,
                                         &mEndOffsets);

  nsCOMPtr<nsIContent> commonContent = do_QueryInterface(mCommonParent);
  mStartRootIndex = mStartNodes.IndexOf(commonContent);
  mEndRootIndex = mEndNodes.IndexOf(commonContent);

  nsresult rv = NS_OK;

  rv = SerializeRangeContextStart(mCommonAncestors, aOutputString);
  NS_ENSURE_SUCCESS(rv, rv);

  if (startContainer == endContainer && IsTextNode(startContainer)) {
    if (mFlags & SkipInvisibleContent) {
      // Check that the parent is visible if we don't a frame.
      // IsVisibleNode() will do it when there's a frame.
      nsCOMPtr<nsIContent> content = do_QueryInterface(startContainer);
      if (content && !content->GetPrimaryFrame()) {
        nsIContent* parent = content->GetParent();
        if (!parent || !IsVisibleNode(parent)) return NS_OK;
      }
    }
    rv = SerializeNodeStart(startContainer, startOffset, endOffset,
                            aOutputString);
    NS_ENSURE_SUCCESS(rv, rv);
  } else {
    rv = SerializeRangeNodes(aRange, mCommonParent, aOutputString, 0);
    NS_ENSURE_SUCCESS(rv, rv);
  }
  rv = SerializeRangeContextEnd(aOutputString);
  NS_ENSURE_SUCCESS(rv, rv);

  return rv;
}

void nsDocumentEncoder::Clear() {
  mDocument = nullptr;
  mSelection = nullptr;
  mRange = nullptr;
  mNode = nullptr;
  mCommonParent = nullptr;
  mNodeFixup = nullptr;

  Initialize(false);
}

NS_IMETHODIMP
nsDocumentEncoder::EncodeToString(nsAString& aOutputString) {
  return EncodeToStringWithMaxLength(0, aOutputString);
}

static bool ParentIsTR(nsIContent* aContent) {
  mozilla::dom::Element* parent = aContent->GetParentElement();
  if (!parent) {
    return false;
  }
  return parent->IsHTMLElement(nsGkAtoms::tr);
}

NS_IMETHODIMP
nsDocumentEncoder::EncodeToStringWithMaxLength(uint32_t aMaxLength,
                                               nsAString& aOutputString) {
  MOZ_ASSERT(mRangeContexts.IsEmpty(), "Re-entrant call to nsDocumentEncoder.");
  auto rangeContextGuard = MakeScopeExit([&] { mRangeContexts.Clear(); });

  if (!mDocument) return NS_ERROR_NOT_INITIALIZED;

  AutoReleaseDocumentIfNeeded autoReleaseDocument(this);

  aOutputString.Truncate();

  nsString output;
  static const size_t bufferSize = 2048;
  if (!mCachedBuffer) {
    mCachedBuffer = nsStringBuffer::Alloc(bufferSize).take();
    if (NS_WARN_IF(!mCachedBuffer)) {
      return NS_ERROR_OUT_OF_MEMORY;
    }
  }
  NS_ASSERTION(
      !mCachedBuffer->IsReadonly(),
      "nsIDocumentEncoder shouldn't keep reference to non-readonly buffer!");
  static_cast<char16_t*>(mCachedBuffer->Data())[0] = char16_t(0);
  mCachedBuffer->ToString(0, output, true);
  // output owns the buffer now!
  mCachedBuffer = nullptr;

  if (!mSerializer) {
    nsAutoCString progId(NS_CONTENTSERIALIZER_CONTRACTID_PREFIX);
    AppendUTF16toUTF8(mMimeType, progId);

    mSerializer = do_CreateInstance(progId.get());
    NS_ENSURE_TRUE(mSerializer, NS_ERROR_NOT_IMPLEMENTED);
  }

  nsresult rv = NS_OK;

  bool rewriteEncodingDeclaration =
      !(mSelection || mRange || mNode) &&
      !(mFlags & OutputDontRewriteEncodingDeclaration);
  mSerializer->Init(mFlags, mWrapColumn, mEncoding, mIsCopying,
                    rewriteEncodingDeclaration, &mNeedsPreformatScanning);

  if (mSelection) {
    uint32_t count = mSelection->RangeCount();

    nsCOMPtr<nsINode> node, prevNode;
    uint32_t firstRangeStartDepth = 0;
    for (uint32_t i = 0; i < count; ++i) {
      RefPtr<nsRange> range = mSelection->GetRangeAt(i);

      // Bug 236546: newlines not added when copying table cells into clipboard
      // Each selected cell shows up as a range containing a row with a single
      // cell get the row, compare it to previous row and emit </tr><tr> as
      // needed Bug 137450: Problem copying/pasting a table from a web page to
      // Excel. Each separate block of <tr></tr> produced above will be wrapped
      // by the immediate context. This assumes that you can't select cells that
      // are multiple selections from two tables simultaneously.
      node = range->GetStartContainer();
      NS_ENSURE_TRUE(node, NS_ERROR_FAILURE);
      if (node != prevNode) {
        if (prevNode) {
          rv = SerializeNodeEnd(prevNode, output);
          NS_ENSURE_SUCCESS(rv, rv);
        }
        nsCOMPtr<nsIContent> content = do_QueryInterface(node);
        if (content && content->IsHTMLElement(nsGkAtoms::tr) &&
            !ParentIsTR(content)) {
          if (!prevNode) {
            // Went from a non-<tr> to a <tr>
            mCommonAncestors.Clear();
            nsContentUtils::GetAncestors(node->GetParentNode(),
                                         mCommonAncestors);
            rv = SerializeRangeContextStart(mCommonAncestors, output);
            NS_ENSURE_SUCCESS(rv, rv);
            // Don't let SerializeRangeToString serialize the context again
            mDisableContextSerialize = true;
          }

          rv = SerializeNodeStart(node, 0, -1, output);
          NS_ENSURE_SUCCESS(rv, rv);
          prevNode = node;
        } else if (prevNode) {
          // Went from a <tr> to a non-<tr>
          mDisableContextSerialize = false;
          rv = SerializeRangeContextEnd(output);
          NS_ENSURE_SUCCESS(rv, rv);
          prevNode = nullptr;
        }
      }

      rv = SerializeRangeToString(range, output);
      NS_ENSURE_SUCCESS(rv, rv);
      if (i == 0) {
        firstRangeStartDepth = mStartDepth;
      }
    }
    mStartDepth = firstRangeStartDepth;

    if (prevNode) {
      rv = SerializeNodeEnd(prevNode, output);
      NS_ENSURE_SUCCESS(rv, rv);
      mDisableContextSerialize = false;
      rv = SerializeRangeContextEnd(output);
      NS_ENSURE_SUCCESS(rv, rv);
    }

    // Just to be safe
    mDisableContextSerialize = false;

    mSelection = nullptr;
  } else if (mRange) {
    rv = SerializeRangeToString(mRange, output);

    mRange = nullptr;
  } else if (mNode) {
    if (!mNodeFixup && !(mFlags & SkipInvisibleContent) && !mStream &&
        mNodeIsContainer) {
      rv = SerializeToStringIterative(mNode, output);
    } else {
      rv = SerializeToStringRecursive(mNode, output, mNodeIsContainer);
    }
    mNode = nullptr;
  } else {
    rv = mSerializer->AppendDocumentStart(mDocument, output);

    if (NS_SUCCEEDED(rv)) {
      rv = SerializeToStringRecursive(mDocument, output, false, aMaxLength);
    }
  }

  NS_ENSURE_SUCCESS(rv, rv);
  rv = mSerializer->Flush(output);

  mCachedBuffer = nsStringBuffer::FromString(output);
  // We have to be careful how we set aOutputString, because we don't
  // want it to end up sharing mCachedBuffer if we plan to reuse it.
  bool setOutput = false;
  // Try to cache the buffer.
  if (mCachedBuffer) {
    if (mCachedBuffer->StorageSize() == bufferSize &&
        !mCachedBuffer->IsReadonly()) {
      mCachedBuffer->AddRef();
    } else {
      if (NS_SUCCEEDED(rv)) {
        mCachedBuffer->ToString(output.Length(), aOutputString);
        setOutput = true;
      }
      mCachedBuffer = nullptr;
    }
  }

  if (!setOutput && NS_SUCCEEDED(rv)) {
    aOutputString.Append(output.get(), output.Length());
  }

  return rv;
}

NS_IMETHODIMP
nsDocumentEncoder::EncodeToStream(nsIOutputStream* aStream) {
  MOZ_ASSERT(mRangeContexts.IsEmpty(), "Re-entrant call to nsDocumentEncoder.");
  auto rangeContextGuard = MakeScopeExit([&] { mRangeContexts.Clear(); });
  nsresult rv = NS_OK;

  if (!mDocument) return NS_ERROR_NOT_INITIALIZED;

  if (!mEncoding) {
    return NS_ERROR_UCONV_NOCONV;
  }

  mUnicodeEncoder = mEncoding->NewEncoder();

  mIsPlainText = (mMimeType.LowerCaseEqualsLiteral("text/plain"));

  mStream = aStream;

  nsAutoString buf;

  rv = EncodeToString(buf);

  // Force a flush of the last chunk of data.
  FlushText(buf, true);

  mStream = nullptr;
  mUnicodeEncoder = nullptr;

  return rv;
}

NS_IMETHODIMP
nsDocumentEncoder::EncodeToStringWithContext(nsAString& aContextString,
                                             nsAString& aInfoString,
                                             nsAString& aEncodedString) {
  return NS_ERROR_NOT_IMPLEMENTED;
}

NS_IMETHODIMP
nsDocumentEncoder::SetNodeFixup(nsIDocumentEncoderNodeFixup* aFixup) {
  mNodeFixup = aFixup;
  return NS_OK;
}

bool do_getDocumentTypeSupportedForEncoding(const char* aContentType) {
  if (!nsCRT::strcmp(aContentType, "text/xml") ||
      !nsCRT::strcmp(aContentType, "application/xml") ||
      !nsCRT::strcmp(aContentType, "application/xhtml+xml") ||
      !nsCRT::strcmp(aContentType, "image/svg+xml") ||
      !nsCRT::strcmp(aContentType, "text/html") ||
      !nsCRT::strcmp(aContentType, "text/plain")) {
    return true;
  }
  return false;
}

already_AddRefed<nsIDocumentEncoder> do_createDocumentEncoder(
    const char* aContentType) {
  if (do_getDocumentTypeSupportedForEncoding(aContentType)) {
    return do_AddRef(new nsDocumentEncoder);
  }
  return nullptr;
}

class nsHTMLCopyEncoder : public nsDocumentEncoder {
 public:
  nsHTMLCopyEncoder();
  virtual ~nsHTMLCopyEncoder();

  NS_IMETHOD Init(Document* aDocument, const nsAString& aMimeType,
                  uint32_t aFlags) override;

  // overridden methods from nsDocumentEncoder
  MOZ_CAN_RUN_SCRIPT_BOUNDARY
  NS_IMETHOD SetSelection(Selection* aSelection) override;
  NS_IMETHOD EncodeToStringWithContext(nsAString& aContextString,
                                       nsAString& aInfoString,
                                       nsAString& aEncodedString) override;
  NS_IMETHOD EncodeToString(nsAString& aOutputString) override;

 protected:
  enum Endpoint { kStart, kEnd };

  nsresult PromoteRange(nsRange* inRange);
  nsresult PromoteAncestorChain(nsCOMPtr<nsINode>* ioNode,
                                int32_t* ioStartOffset, int32_t* ioEndOffset);
  nsresult GetPromotedPoint(Endpoint aWhere, nsINode* aNode, int32_t aOffset,
                            nsCOMPtr<nsINode>* outNode, int32_t* outOffset,
                            nsINode* aCommon);
  nsCOMPtr<nsINode> GetChildAt(nsINode* aParent, int32_t aOffset);
  bool IsMozBR(Element* aNode);
  nsresult GetNodeLocation(nsINode* inChild, nsCOMPtr<nsINode>* outParent,
                           int32_t* outOffset);
  bool IsRoot(nsINode* aNode);
  bool IsFirstNode(nsINode* aNode);
  bool IsLastNode(nsINode* aNode);
  virtual bool IncludeInContext(nsINode* aNode) override;
  virtual int32_t GetImmediateContextCount(
      const nsTArray<nsINode*>& aAncestorArray) override;

  bool mIsTextWidget;
};

nsHTMLCopyEncoder::nsHTMLCopyEncoder() { mIsTextWidget = false; }

nsHTMLCopyEncoder::~nsHTMLCopyEncoder() {}

NS_IMETHODIMP
nsHTMLCopyEncoder::Init(Document* aDocument, const nsAString& aMimeType,
                        uint32_t aFlags) {
  if (!aDocument) return NS_ERROR_INVALID_ARG;

  mIsTextWidget = false;
  Initialize();

  mIsCopying = true;
  mDocument = aDocument;

  // Hack, hack! Traditionally, the caller passes text/unicode, which is
  // treated as "guess text/html or text/plain" in this context. (It has a
  // different meaning in other contexts. Sigh.) From now on, "text/plain"
  // means forcing text/plain instead of guessing.
  if (aMimeType.EqualsLiteral("text/plain")) {
    mMimeType.AssignLiteral("text/plain");
  } else {
    mMimeType.AssignLiteral("text/html");
  }

  // Make all links absolute when copying
  // (see related bugs #57296, #41924, #58646, #32768)
  mFlags = aFlags | OutputAbsoluteLinks;

  if (!mDocument->IsScriptEnabled()) mFlags |= OutputNoScriptContent;

  return NS_OK;
}

NS_IMETHODIMP
nsHTMLCopyEncoder::SetSelection(Selection* aSelection) {
  // check for text widgets: we need to recognize these so that
  // we don't tweak the selection to be outside of the magic
  // div that ender-lite text widgets are embedded in.

  if (!aSelection) return NS_ERROR_NULL_POINTER;

  uint32_t rangeCount = aSelection->RangeCount();

  // if selection is uninitialized return
  if (!rangeCount) {
    return NS_ERROR_FAILURE;
  }

  // we'll just use the common parent of the first range.  Implicit assumption
  // here that multi-range selections are table cell selections, in which case
  // the common parent is somewhere in the table and we don't really care where.
  //
  // FIXME(emilio, bug 1455894): This assumption is already wrong, and will
  // probably be more wrong in a Shadow DOM world...
  //
  // We should be able to write this as "Find the common ancestor of the
  // selection, then go through the flattened tree and serialize the selected
  // nodes", effectively serializing the composed tree.
  RefPtr<nsRange> range = aSelection->GetRangeAt(0);
  nsINode* commonParent = range->GetCommonAncestor();

  for (nsCOMPtr<nsIContent> selContent(do_QueryInterface(commonParent));
       selContent; selContent = selContent->GetParent()) {
    // checking for selection inside a plaintext form widget
    if (selContent->IsAnyOfHTMLElements(nsGkAtoms::input,
                                        nsGkAtoms::textarea)) {
      mIsTextWidget = true;
      break;
    }
#if defined(MOZ_THUNDERBIRD) || defined(MOZ_SUITE)
    else if (selContent->IsHTMLElement(nsGkAtoms::body)) {
      // Currently, setting mIsTextWidget to 'true' will result in the selection
      // being encoded/copied as pre-formatted plain text.
      // This is fine for copying pre-formatted plain text with Firefox, it is
      // already not correct for copying pre-formatted "rich" text (bold,
      // colour) with Firefox. As long as the serialisers aren't fixed, copying
      // pre-formatted text in Firefox is broken. If we set mIsTextWidget,
      // pre-formatted plain text is copied, but pre-formatted "rich" text loses
      // the "rich" formatting. If we don't set mIsTextWidget, "rich" text
      // attributes aren't lost, but white-space is lost.
      // So far the story for Firefox.
      //
      // Thunderbird has two *conflicting* requirements.
      // Case 1:
      // When selecting and copying text, even pre-formatted text, as a quote
      // to be placed into a reply, we *always* expect HTML to be copied.
      // Case 2:
      // When copying text in a so-called "plain text" message, that is
      // one where the body carries style "white-space:pre-wrap", the text
      // should be copied as pre-formatted plain text.
      //
      // Therefore the following code checks for "pre-wrap" on the body.
      // This is a terrible hack.
      //
      // The proper fix would be this:
      // For case 1:
      // Communicate the fact that HTML is required to EncodeToString(),
      // bug 1141786.
      // For case 2:
      // Wait for Firefox to get fixed to detect pre-formatting correctly,
      // bug 1174452.
      nsAutoString styleVal;
      if (selContent->IsElement() &&
          selContent->AsElement()->GetAttr(kNameSpaceID_None, nsGkAtoms::style,
                                           styleVal) &&
          styleVal.Find(NS_LITERAL_STRING("pre-wrap")) != kNotFound) {
        mIsTextWidget = true;
        break;
      }
    }
#endif
  }

  // normalize selection if we are not in a widget
  if (mIsTextWidget) {
    mSelection = aSelection;
    mMimeType.AssignLiteral("text/plain");
    return NS_OK;
  }

  // XXX We should try to get rid of the Selection object here.
  // XXX bug 1245883

  // also consider ourselves in a text widget if we can't find an html document
  nsCOMPtr<nsIHTMLDocument> htmlDoc = do_QueryInterface(mDocument);
  if (!(htmlDoc && mDocument->IsHTMLDocument())) {
    mIsTextWidget = true;
    mSelection = aSelection;
    // mMimeType is set to text/plain when encoding starts.
    return NS_OK;
  }

  // there's no Clone() for selection! fix...
  // nsresult rv = aSelection->Clone(getter_AddRefs(mSelection);
  // NS_ENSURE_SUCCESS(rv, rv);
  mSelection = new Selection();

  // loop thru the ranges in the selection
  for (uint32_t rangeIdx = 0; rangeIdx < rangeCount; ++rangeIdx) {
    range = aSelection->GetRangeAt(rangeIdx);
    NS_ENSURE_TRUE(range, NS_ERROR_FAILURE);
    RefPtr<nsRange> myRange = range->CloneRange();
    MOZ_ASSERT(myRange);

    // adjust range to include any ancestors who's children are entirely
    // selected
    nsresult rv = PromoteRange(myRange);
    NS_ENSURE_SUCCESS(rv, rv);

    ErrorResult result;
    mSelection->AddRangeInternal(*myRange, mDocument, result);
    rv = result.StealNSResult();
    NS_ENSURE_SUCCESS(rv, rv);
  }

  return NS_OK;
}

NS_IMETHODIMP
nsHTMLCopyEncoder::EncodeToString(nsAString& aOutputString) {
  if (mIsTextWidget) {
    mMimeType.AssignLiteral("text/plain");
  }
  return nsDocumentEncoder::EncodeToString(aOutputString);
}

NS_IMETHODIMP
nsHTMLCopyEncoder::EncodeToStringWithContext(nsAString& aContextString,
                                             nsAString& aInfoString,
                                             nsAString& aEncodedString) {
  nsresult rv = EncodeToString(aEncodedString);
  NS_ENSURE_SUCCESS(rv, rv);

  // do not encode any context info or range hints if we are in a text widget.
  if (mIsTextWidget) return NS_OK;

  // now encode common ancestors into aContextString.  Note that the common
  // ancestors will be for the last range in the selection in the case of
  // multirange selections. encoding ancestors every range in a multirange
  // selection in a way that could be understood by the paste code would be a
  // lot more work to do.  As a practical matter, selections are single range,
  // and the ones that aren't are table cell selections where all the cells are
  // in the same table.

  // leaf of ancestors might be text node.  If so discard it.
  int32_t count = mCommonAncestors.Length();
  int32_t i;
  nsCOMPtr<nsINode> node;
  if (count > 0) node = mCommonAncestors.ElementAt(0);

  if (node && IsTextNode(node)) {
    mCommonAncestors.RemoveElementAt(0);
    // don't forget to adjust range depth info
    if (mStartDepth) mStartDepth--;
    if (mEndDepth) mEndDepth--;
    // and the count
    count--;
  }

  i = count;
  while (i > 0) {
    node = mCommonAncestors.ElementAt(--i);
    SerializeNodeStart(node, 0, -1, aContextString);
  }
  // i = 0; guaranteed by above
  while (i < count) {
    node = mCommonAncestors.ElementAt(i++);
    SerializeNodeEnd(node, aContextString);
  }

  // encode range info : the start and end depth of the selection, where the
  // depth is distance down in the parent hierarchy.  Later we will need to add
  // leading/trailing whitespace info to this.
  nsAutoString infoString;
  infoString.AppendInt(mStartDepth);
  infoString.Append(char16_t(','));
  infoString.AppendInt(mEndDepth);
  aInfoString = infoString;

  return NS_OK;
}

bool nsHTMLCopyEncoder::IncludeInContext(nsINode* aNode) {
  nsCOMPtr<nsIContent> content(do_QueryInterface(aNode));

  if (!content) return false;

  return content->IsAnyOfHTMLElements(
      nsGkAtoms::b, nsGkAtoms::i, nsGkAtoms::u, nsGkAtoms::a, nsGkAtoms::tt,
      nsGkAtoms::s, nsGkAtoms::big, nsGkAtoms::small, nsGkAtoms::strike,
      nsGkAtoms::em, nsGkAtoms::strong, nsGkAtoms::dfn, nsGkAtoms::code,
      nsGkAtoms::cite, nsGkAtoms::var, nsGkAtoms::abbr, nsGkAtoms::font,
      nsGkAtoms::script, nsGkAtoms::span, nsGkAtoms::pre, nsGkAtoms::h1,
      nsGkAtoms::h2, nsGkAtoms::h3, nsGkAtoms::h4, nsGkAtoms::h5,
      nsGkAtoms::h6);
}

nsresult nsHTMLCopyEncoder::PromoteRange(nsRange* inRange) {
  if (!inRange->IsPositioned()) {
    return NS_ERROR_UNEXPECTED;
  }
  nsCOMPtr<nsINode> startNode = inRange->GetStartContainer();
  uint32_t startOffset = inRange->StartOffset();
  nsCOMPtr<nsINode> endNode = inRange->GetEndContainer();
  uint32_t endOffset = inRange->EndOffset();
  nsCOMPtr<nsINode> common = inRange->GetCommonAncestor();

  nsCOMPtr<nsINode> opStartNode;
  nsCOMPtr<nsINode> opEndNode;
  int32_t opStartOffset, opEndOffset;

  // examine range endpoints.
  nsresult rv =
      GetPromotedPoint(kStart, startNode, static_cast<int32_t>(startOffset),
                       address_of(opStartNode), &opStartOffset, common);
  NS_ENSURE_SUCCESS(rv, rv);
  rv = GetPromotedPoint(kEnd, endNode, static_cast<int32_t>(endOffset),
                        address_of(opEndNode), &opEndOffset, common);
  NS_ENSURE_SUCCESS(rv, rv);

  // if both range endpoints are at the common ancestor, check for possible
  // inclusion of ancestors
  if (opStartNode == common && opEndNode == common) {
    rv = PromoteAncestorChain(address_of(opStartNode), &opStartOffset,
                              &opEndOffset);
    NS_ENSURE_SUCCESS(rv, rv);
    opEndNode = opStartNode;
  }

  // set the range to the new values
  ErrorResult err;
  inRange->SetStart(*opStartNode, static_cast<uint32_t>(opStartOffset), err);
  if (NS_WARN_IF(err.Failed())) {
    return err.StealNSResult();
  }
  inRange->SetEnd(*opEndNode, static_cast<uint32_t>(opEndOffset), err);
  if (NS_WARN_IF(err.Failed())) {
    return err.StealNSResult();
  }
  return NS_OK;
}

// PromoteAncestorChain will promote a range represented by
// [{*ioNode,*ioStartOffset} , {*ioNode,*ioEndOffset}] The promotion is
// different from that found in getPromotedPoint: it will only promote one
// endpoint if it can promote the other.  Thus, instead of having a
// startnode/endNode, there is just the one ioNode.
nsresult nsHTMLCopyEncoder::PromoteAncestorChain(nsCOMPtr<nsINode>* ioNode,
                                                 int32_t* ioStartOffset,
                                                 int32_t* ioEndOffset) {
  if (!ioNode || !ioStartOffset || !ioEndOffset) return NS_ERROR_NULL_POINTER;

  nsresult rv = NS_OK;
  bool done = false;

  nsCOMPtr<nsINode> frontNode, endNode, parent;
  int32_t frontOffset, endOffset;

  // save the editable state of the ioNode, so we don't promote an ancestor if
  // it has different editable state
  nsCOMPtr<nsINode> node = *ioNode;
  bool isEditable = node->IsEditable();

  // loop for as long as we can promote both endpoints
  while (!done) {
    node = *ioNode;
    parent = node->GetParentNode();
    if (!parent) {
      done = true;
    } else {
      // passing parent as last param to GetPromotedPoint() allows it to promote
      // only one level up the hierarchy.
      rv = GetPromotedPoint(kStart, *ioNode, *ioStartOffset,
                            address_of(frontNode), &frontOffset, parent);
      NS_ENSURE_SUCCESS(rv, rv);
      // then we make the same attempt with the endpoint
      rv = GetPromotedPoint(kEnd, *ioNode, *ioEndOffset, address_of(endNode),
                            &endOffset, parent);
      NS_ENSURE_SUCCESS(rv, rv);

      // if both endpoints were promoted one level and isEditable is the same as
      // the original node, keep looping - otherwise we are done.
      if ((frontNode != parent) || (endNode != parent) ||
          (frontNode->IsEditable() != isEditable))
        done = true;
      else {
        *ioNode = frontNode;
        *ioStartOffset = frontOffset;
        *ioEndOffset = endOffset;
      }
    }
  }
  return rv;
}

nsresult nsHTMLCopyEncoder::GetPromotedPoint(Endpoint aWhere, nsINode* aNode,
                                             int32_t aOffset,
                                             nsCOMPtr<nsINode>* outNode,
                                             int32_t* outOffset,
                                             nsINode* common) {
  nsresult rv = NS_OK;
  nsCOMPtr<nsINode> node = aNode;
  nsCOMPtr<nsINode> parent = aNode;
  int32_t offset = aOffset;
  bool bResetPromotion = false;

  // default values
  *outNode = node;
  *outOffset = offset;

  if (common == node) return NS_OK;

  if (aWhere == kStart) {
    // some special casing for text nodes
    if (auto nodeAsText = aNode->GetAsText()) {
      // if not at beginning of text node, we are done
      if (offset > 0) {
        // unless everything before us in just whitespace.  NOTE: we need a more
        // general solution that truly detects all cases of non-significant
        // whitesace with no false alarms.
        nsAutoString text;
        nodeAsText->SubstringData(0, offset, text, IgnoreErrors());
        text.CompressWhitespace();
        if (!text.IsEmpty()) return NS_OK;
        bResetPromotion = true;
      }
      // else
      rv = GetNodeLocation(aNode, address_of(parent), &offset);
      NS_ENSURE_SUCCESS(rv, rv);
    } else {
      node = GetChildAt(parent, offset);
    }
    if (!node) node = parent;

    // finding the real start for this point.  look up the tree for as long as
    // we are the first node in the container, and as long as we haven't hit the
    // body node.
    if (!IsRoot(node) && (parent != common)) {
      rv = GetNodeLocation(node, address_of(parent), &offset);
      NS_ENSURE_SUCCESS(rv, rv);
      if (offset == -1) return NS_OK;  // we hit generated content; STOP
      while ((IsFirstNode(node)) && (!IsRoot(parent)) && (parent != common)) {
        if (bResetPromotion) {
          nsCOMPtr<nsIContent> content = do_QueryInterface(parent);
          if (content && content->IsHTMLElement()) {
            if (nsHTMLElement::IsBlock(
                    nsHTMLTags::AtomTagToId(content->NodeInfo()->NameAtom()))) {
              bResetPromotion = false;
            }
          }
        }

        node = parent;
        rv = GetNodeLocation(node, address_of(parent), &offset);
        NS_ENSURE_SUCCESS(rv, rv);
        if (offset == -1)  // we hit generated content; STOP
        {
          // back up a bit
          parent = node;
          offset = 0;
          break;
        }
      }
      if (bResetPromotion) {
        *outNode = aNode;
        *outOffset = aOffset;
      } else {
        *outNode = parent;
        *outOffset = offset;
      }
      return rv;
    }
  }

  if (aWhere == kEnd) {
    // some special casing for text nodes
    if (auto nodeAsText = aNode->GetAsText()) {
      // if not at end of text node, we are done
      uint32_t len = aNode->Length();
      if (offset < (int32_t)len) {
        // unless everything after us in just whitespace.  NOTE: we need a more
        // general solution that truly detects all cases of non-significant
        // whitespace with no false alarms.
        nsAutoString text;
        nodeAsText->SubstringData(offset, len - offset, text, IgnoreErrors());
        text.CompressWhitespace();
        if (!text.IsEmpty()) return NS_OK;
        bResetPromotion = true;
      }
      rv = GetNodeLocation(aNode, address_of(parent), &offset);
      NS_ENSURE_SUCCESS(rv, rv);
    } else {
      if (offset) offset--;  // we want node _before_ offset
      node = GetChildAt(parent, offset);
    }
    if (!node) node = parent;

    // finding the real end for this point.  look up the tree for as long as we
    // are the last node in the container, and as long as we haven't hit the
    // body node.
    if (!IsRoot(node) && (parent != common)) {
      rv = GetNodeLocation(node, address_of(parent), &offset);
      NS_ENSURE_SUCCESS(rv, rv);
      if (offset == -1) return NS_OK;  // we hit generated content; STOP
      while ((IsLastNode(node)) && (!IsRoot(parent)) && (parent != common)) {
        if (bResetPromotion) {
          nsCOMPtr<nsIContent> content = do_QueryInterface(parent);
          if (content && content->IsHTMLElement()) {
            if (nsHTMLElement::IsBlock(
                    nsHTMLTags::AtomTagToId(content->NodeInfo()->NameAtom()))) {
              bResetPromotion = false;
            }
          }
        }

        node = parent;
        rv = GetNodeLocation(node, address_of(parent), &offset);
        NS_ENSURE_SUCCESS(rv, rv);
        if (offset == -1)  // we hit generated content; STOP
        {
          // back up a bit
          parent = node;
          offset = 0;
          break;
        }
      }
      if (bResetPromotion) {
        *outNode = aNode;
        *outOffset = aOffset;
      } else {
        *outNode = parent;
        offset++;  // add one since this in an endpoint - want to be AFTER node.
        *outOffset = offset;
      }
      return rv;
    }
  }

  return rv;
}

nsCOMPtr<nsINode> nsHTMLCopyEncoder::GetChildAt(nsINode* aParent,
                                                int32_t aOffset) {
  nsCOMPtr<nsINode> resultNode;

  if (!aParent) return resultNode;

  nsCOMPtr<nsIContent> content = do_QueryInterface(aParent);
  MOZ_ASSERT(content, "null content in nsHTMLCopyEncoder::GetChildAt");

  resultNode = content->GetChildAt_Deprecated(aOffset);

  return resultNode;
}

bool nsHTMLCopyEncoder::IsMozBR(Element* aElement) {
  return aElement->IsHTMLElement(nsGkAtoms::br) &&
         aElement->AttrValueIs(kNameSpaceID_None, nsGkAtoms::type,
                               NS_LITERAL_STRING("_moz"), eIgnoreCase);
}

nsresult nsHTMLCopyEncoder::GetNodeLocation(nsINode* inChild,
                                            nsCOMPtr<nsINode>* outParent,
                                            int32_t* outOffset) {
  NS_ASSERTION((inChild && outParent && outOffset), "bad args");
  if (inChild && outParent && outOffset) {
    nsCOMPtr<nsIContent> child = do_QueryInterface(inChild);
    if (!child) {
      return NS_ERROR_NULL_POINTER;
    }

    nsIContent* parent = child->GetParent();
    if (!parent) {
      return NS_ERROR_NULL_POINTER;
    }

    *outParent = parent;
    *outOffset = parent->ComputeIndexOf(child);
    return NS_OK;
  }
  return NS_ERROR_NULL_POINTER;
}

bool nsHTMLCopyEncoder::IsRoot(nsINode* aNode) {
  nsCOMPtr<nsIContent> content = do_QueryInterface(aNode);
  if (!content) {
    return false;
  }

  if (mIsTextWidget) {
    return content->IsHTMLElement(nsGkAtoms::div);
  }

  return content->IsAnyOfHTMLElements(nsGkAtoms::body, nsGkAtoms::td,
                                      nsGkAtoms::th);
}

bool nsHTMLCopyEncoder::IsFirstNode(nsINode* aNode) {
  // need to check if any nodes before us are really visible.
  // Mike wrote something for me along these lines in nsSelectionController,
  // but I don't think it's ready for use yet - revisit.
  // HACK: for now, simply consider all whitespace text nodes to be
  // invisible formatting nodes.
  for (nsIContent* sibling = aNode->GetPreviousSibling(); sibling;
       sibling = sibling->GetPreviousSibling()) {
    if (!sibling->TextIsOnlyWhitespace()) {
      return false;
    }
  }

  return true;
}

bool nsHTMLCopyEncoder::IsLastNode(nsINode* aNode) {
  // need to check if any nodes after us are really visible.
  // Mike wrote something for me along these lines in nsSelectionController,
  // but I don't think it's ready for use yet - revisit.
  // HACK: for now, simply consider all whitespace text nodes to be
  // invisible formatting nodes.
  for (nsIContent* sibling = aNode->GetNextSibling(); sibling;
       sibling = sibling->GetNextSibling()) {
    if (sibling->IsElement() && IsMozBR(sibling->AsElement())) {
      // we ignore trailing moz BRs.
      continue;
    }
    if (!sibling->TextIsOnlyWhitespace()) {
      return false;
    }
  }

  return true;
}

already_AddRefed<nsIDocumentEncoder> do_createHTMLCopyEncoder() {
  return do_AddRef(new nsHTMLCopyEncoder);
}

int32_t nsHTMLCopyEncoder::GetImmediateContextCount(
    const nsTArray<nsINode*>& aAncestorArray) {
  int32_t i = aAncestorArray.Length(), j = 0;
  while (j < i) {
    nsINode* node = aAncestorArray.ElementAt(j);
    if (!node) {
      break;
    }
    nsCOMPtr<nsIContent> content(do_QueryInterface(node));
    if (!content || !content->IsAnyOfHTMLElements(
                        nsGkAtoms::tr, nsGkAtoms::thead, nsGkAtoms::tbody,
                        nsGkAtoms::tfoot, nsGkAtoms::table)) {
      break;
    }
    ++j;
  }
  return j;
}