Implement a lexer for structured comments.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@159223 91177308-0d34-0410-b5e6-96231b3b80d8

17 files changed, 2326 insertions, 27 deletions

diff --git a/include/clang-c/Index.h b/include/clang-c/Index.h
index b7bd8bb738..2397ae1925 100644
--- a/include/clang-c/Index.h
+++ b/include/clang-c/Index.h
@@ -3201,6 +3201,12 @@ CINDEX_LINKAGE CXSourceRange clang_Cursor_getCommentRange(CXCursor C);
 CINDEX_LINKAGE CXString clang_Cursor_getRawCommentText(CXCursor C);
 
 /**
+ * \brief Given a cursor that represents a declaration, return the associated
+ * \\brief paragraph; otherwise return the first paragraph.
+ */
+CINDEX_LINKAGE CXString clang_Cursor_getBriefCommentText(CXCursor C);
+
+/**
  * @}
  */
 
diff --git a/include/clang/AST/CommentBriefParser.h b/include/clang/AST/CommentBriefParser.h
new file mode 100644
index 0000000000..e343b94643
--- /dev/null
+++ b/include/clang/AST/CommentBriefParser.h
@@ -0,0 +1,49 @@
+//===--- CommentBriefParser.h - Dumb comment parser -------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+//  This file defines a very simple Doxygen comment parser.
+//
+//===----------------------------------------------------------------------===//
+
+
+#ifndef LLVM_CLANG_AST_BRIEF_COMMENT_PARSER_H
+#define LLVM_CLANG_AST_BRIEF_COMMENT_PARSER_H
+
+#include "clang/AST/CommentLexer.h"
+
+namespace clang {
+namespace comments {
+
+/// A very simple comment parser that extracts just the brief description or
+/// first paragraph.
+class BriefParser {
+  Lexer &L;
+
+  /// Current lookahead token.
+  Token Tok;
+
+  SourceLocation ConsumeToken() {
+    SourceLocation Loc = Tok.getLocation();
+    L.lex(Tok);
+    return Loc;
+  }
+
+public:
+  BriefParser(Lexer &L);
+
+  /// Return \\brief paragraph, if it exists; otherwise return the first
+  /// paragraph.
+  std::string Parse();
+};
+
+} // end namespace comments
+} // end namespace clang
+
+#endif
+
diff --git a/include/clang/AST/CommentLexer.h b/include/clang/AST/CommentLexer.h
new file mode 100644
index 0000000000..7f7ae62758
--- /dev/null
+++ b/include/clang/AST/CommentLexer.h
@@ -0,0 +1,352 @@
+//===--- CommentLexer.h - Lexer for structured comments ---------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+//  This file defines lexer for structured comments and supporting token class.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_AST_COMMENT_LEXER_H
+#define LLVM_CLANG_AST_COMMENT_LEXER_H
+
+#include "clang/Basic/SourceManager.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/Support/raw_ostream.h"
+
+namespace clang {
+namespace comments {
+
+class Lexer;
+
+namespace tok {
+enum TokenKind {
+  eof,
+  newline,
+  text,
+  command,
+  verbatim_block_begin,
+  verbatim_block_line,
+  verbatim_block_end,
+  verbatim_line,
+  html_tag_open,      // <tag
+  html_ident,         // attr
+  html_equals,        // =
+  html_quoted_string, // "blah\"blah" or 'blah\'blah'
+  html_greater,       // >
+  html_tag_close,     // </tag>
+
+  // Markdown tokens (not supported yet).
+  ruler,
+  md_code_line,   // Line indented at least by 4 spaces.
+  md_code_inline, // `code`
+  md_emph,        // _text_ or *text*
+  md_strong,      // __text__ or *text*
+  md_header       // ### level 3 header ###
+};
+} // end namespace tok
+
+class CommentOptions {
+public:
+  bool Markdown;
+};
+
+/// \brief Comment token.
+class Token {
+  friend class Lexer;
+
+  /// The location of the token.
+  SourceLocation Loc;
+
+  /// The actual kind of the token.
+  tok::TokenKind Kind;
+
+  /// Length of the token spelling in comment.  Can be 0 for synthenized
+  /// tokens.
+  unsigned Length;
+
+  /// Contains text value associated with a token.
+  const char *TextPtr1;
+  unsigned TextLen1;
+
+  /// Contains text value associated with a token.
+  const char *TextPtr2;
+  unsigned TextLen2;
+
+public:
+  SourceLocation getLocation() const LLVM_READONLY { return Loc; }
+  void setLocation(SourceLocation SL) { Loc = SL; }
+
+  tok::TokenKind getKind() const LLVM_READONLY { return Kind; }
+  void setKind(tok::TokenKind K) { Kind = K; }
+
+  bool is(tok::TokenKind K) const LLVM_READONLY { return Kind == K; }
+  bool isNot(tok::TokenKind K) const LLVM_READONLY { return Kind != K; }
+
+  unsigned getLength() const LLVM_READONLY { return Length; }
+  void setLength(unsigned L) { Length = L; }
+
+  StringRef getText() const LLVM_READONLY {
+    assert(is(tok::text));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setText(StringRef Text) {
+    assert(is(tok::text));
+    TextPtr1 = Text.data();
+    TextLen1 = Text.size();
+  }
+
+  StringRef getCommandName() const LLVM_READONLY {
+    assert(is(tok::command));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setCommandName(StringRef Name) {
+    assert(is(tok::command));
+    TextPtr1 = Name.data();
+    TextLen1 = Name.size();
+  }
+
+  StringRef getVerbatimBlockName() const LLVM_READONLY {
+    assert(is(tok::verbatim_block_begin) || is(tok::verbatim_block_end));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setVerbatimBlockName(StringRef Name) {
+    assert(is(tok::verbatim_block_begin) || is(tok::verbatim_block_end));
+    TextPtr1 = Name.data();
+    TextLen1 = Name.size();
+  }
+
+  StringRef getVerbatimBlockText() const LLVM_READONLY {
+    assert(is(tok::verbatim_block_line));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setVerbatimBlockText(StringRef Text) {
+    assert(is(tok::verbatim_block_line));
+    TextPtr1 = Text.data();
+    TextLen1 = Text.size();
+  }
+
+  /// Returns the name of verbatim line command.
+  StringRef getVerbatimLineName() const LLVM_READONLY {
+    assert(is(tok::verbatim_line));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setVerbatimLineName(StringRef Name) {
+    assert(is(tok::verbatim_line));
+    TextPtr1 = Name.data();
+    TextLen1 = Name.size();
+  }
+
+  StringRef getVerbatimLineText() const LLVM_READONLY {
+    assert(is(tok::verbatim_line));
+    return StringRef(TextPtr2, TextLen2);
+  }
+
+  void setVerbatimLineText(StringRef Text) {
+    assert(is(tok::verbatim_line));
+    TextPtr2 = Text.data();
+    TextLen2 = Text.size();
+  }
+
+  StringRef getHTMLTagOpenName() const LLVM_READONLY {
+    assert(is(tok::html_tag_open));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setHTMLTagOpenName(StringRef Name) {
+    assert(is(tok::html_tag_open));
+    TextPtr1 = Name.data();
+    TextLen1 = Name.size();
+  }
+
+  StringRef getHTMLIdent() const LLVM_READONLY {
+    assert(is(tok::html_ident));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setHTMLIdent(StringRef Name) {
+    assert(is(tok::html_ident));
+    TextPtr1 = Name.data();
+    TextLen1 = Name.size();
+  }
+
+  StringRef getHTMLQuotedString() const LLVM_READONLY {
+    assert(is(tok::html_quoted_string));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setHTMLQuotedString(StringRef Str) {
+    assert(is(tok::html_quoted_string));
+    TextPtr1 = Str.data();
+    TextLen1 = Str.size();
+  }
+
+  StringRef getHTMLTagCloseName() const LLVM_READONLY {
+    assert(is(tok::html_tag_close));
+    return StringRef(TextPtr1, TextLen1);
+  }
+
+  void setHTMLTagCloseName(StringRef Name) {
+    assert(is(tok::html_tag_close));
+    TextPtr1 = Name.data();
+    TextLen1 = Name.size();
+  }
+
+  void dump(const Lexer &L, const SourceManager &SM) const;
+};
+
+/// \brief Comment lexer.
+class Lexer {
+private:
+  Lexer(const Lexer&);          // DO NOT IMPLEMENT
+  void operator=(const Lexer&); // DO NOT IMPLEMENT
+
+  const char *const BufferStart;
+  const char *const BufferEnd;
+  SourceLocation FileLoc;
+  CommentOptions CommOpts;
+
+  const char *BufferPtr;
+
+  /// One past end pointer for the current comment.  For BCPL comments points
+  /// to newline or BufferEnd, for C comments points to star in '*/'.
+  const char *CommentEnd;
+
+  enum LexerCommentState {
+    LCS_BeforeComment,
+    LCS_InsideBCPLComment,
+    LCS_InsideCComment,
+    LCS_BetweenComments
+  };
+
+  /// Low-level lexer state, track if we are inside or outside of comment.
+  LexerCommentState CommentState;
+
+  enum LexerState {
+    /// Lexing normal comment text
+    LS_Normal,
+
+    /// Finished lexing verbatim block beginning command, will lex first body
+    /// line.
+    LS_VerbatimBlockFirstLine,
+
+    /// Lexing verbatim block body line-by-line, skipping line-starting
+    /// decorations.
+    LS_VerbatimBlockBody,
+
+    /// Finished lexing \verbatim <TAG \endverbatim part, lexing tag attributes.
+    LS_HTMLOpenTag
+  };
+
+  /// Current lexing mode.
+  LexerState State;
+
+  /// A verbatim-like block command eats every character (except line starting
+  /// decorations) until matching end command is seen or comment end is hit.
+  struct VerbatimBlockCommand {
+    StringRef BeginName;
+    StringRef EndName;
+  };
+
+  typedef SmallVector<VerbatimBlockCommand, 4> VerbatimBlockCommandVector;
+
+  /// Registered verbatim-like block commands.
+  VerbatimBlockCommandVector VerbatimBlockCommands;
+
+  /// If State is LS_VerbatimBlock, contains the the name of verbatim end
+  /// command, including command marker.
+  SmallString<16> VerbatimBlockEndCommandName;
+
+  bool isVerbatimBlockCommand(StringRef BeginName, StringRef &EndName) const;
+
+  /// A verbatim-like line command eats everything until a newline is seen or
+  /// comment end is hit.
+  struct VerbatimLineCommand {
+    StringRef Name;
+  };
+
+  typedef SmallVector<VerbatimLineCommand, 4> VerbatimLineCommandVector;
+
+  /// Registered verbatim-like line commands.
+  VerbatimLineCommandVector VerbatimLineCommands;
+
+  bool isVerbatimLineCommand(StringRef Name) const;
+
+  void formTokenWithChars(Token &Result, const char *TokEnd,
+                          tok::TokenKind Kind) {
+    const unsigned TokLen = TokEnd - BufferPtr;
+    Result.setLocation(getSourceLocation(BufferPtr));
+    Result.setKind(Kind);
+    Result.setLength(TokLen);
+#ifndef NDEBUG
+    Result.TextPtr1 = "<UNSET>";
+    Result.TextLen1 = 7;
+    Result.TextPtr2 = "<UNSET>";
+    Result.TextLen2 = 7;
+#endif
+    BufferPtr = TokEnd;
+  }
+
+  SourceLocation getSourceLocation(const char *Loc) const {
+    assert(Loc >= BufferStart && Loc <= BufferEnd &&
+           "Location out of range for this buffer!");
+
+    const unsigned CharNo = Loc - BufferStart;
+    return FileLoc.getLocWithOffset(CharNo);
+  }
+
+  /// Eat string matching regexp \code \s*\* \endcode.
+  void skipLineStartingDecorations();
+
+  /// Lex stuff inside comments.  CommentEnd should be set correctly.
+  void lexCommentText(Token &T);
+
+  void setupAndLexVerbatimBlock(Token &T,
+                                const char *TextBegin,
+                                char Marker, StringRef EndName);
+
+  void lexVerbatimBlockFirstLine(Token &T);
+
+  void lexVerbatimBlockBody(Token &T);
+
+  void lexVerbatimLine(Token &T, const char *TextBegin);
+
+  void setupAndLexHTMLOpenTag(Token &T);
+
+  void lexHTMLOpenTag(Token &T);
+
+  void lexHTMLCloseTag(Token &T);
+
+public:
+  Lexer(SourceLocation FileLoc, const CommentOptions &CommOpts,
+        const char *BufferStart, const char *BufferEnd);
+
+  void lex(Token &T);
+
+  StringRef getSpelling(const Token &Tok,
+                        const SourceManager &SourceMgr,
+                        bool *Invalid = NULL) const;
+
+  /// \brief Register a new verbatim block command.
+  void addVerbatimBlockCommand(StringRef BeginName, StringRef EndName);
+
+  /// \brief Register a new verbatim line command.
+  void addVerbatimLineCommand(StringRef Name);
+};
+
+} // end namespace comments
+} // end namespace clang
+
+#endif
+
diff --git a/include/clang/AST/RawCommentList.h b/include/clang/AST/RawCommentList.h
index 0965cb3a62..d670fd1428 100644
--- a/include/clang/AST/RawCommentList.h
+++ b/include/clang/AST/RawCommentList.h
@@ -15,6 +15,7 @@
 
 namespace clang {
 
+class ASTContext;
 class ASTReader;
 
 class RawComment {
@@ -27,7 +28,7 @@ public:
     CK_BCPLExcl,     ///< \code //! stuff \endcode
     CK_JavaDoc,      ///< \code /** stuff */ \endcode
     CK_Qt,           ///< \code /*! stuff */ \endcode, also used by HeaderDoc
-    CK_Merged        ///< Two or more Doxygen comments merged together
+    CK_Merged        ///< Two or more documentation comments merged together
   };
 
   RawComment() : Kind(CK_Invalid), IsAlmostTrailingComment(false) { }
@@ -53,7 +54,7 @@ public:
   /// \code /**< stuff */ \endcode
   /// \code /*!< stuff */ \endcode
   bool isTrailingComment() const LLVM_READONLY {
-    assert(isDoxygen());
+    assert(isDocumentation());
     return IsTrailingComment;
   }
 
@@ -64,13 +65,13 @@ public:
     return IsAlmostTrailingComment;
   }
 
-  /// Returns true if this comment is not a Doxygen comment.
+  /// Returns true if this comment is not a documentation comment.
   bool isOrdinary() const LLVM_READONLY {
     return (Kind == CK_OrdinaryBCPL) || (Kind == CK_OrdinaryC);
   }
 
-  /// Returns true if this comment any kind of a Doxygen comment.
-  bool isDoxygen() const LLVM_READONLY {
+  /// Returns true if this comment any kind of a documentation comment.
+  bool isDocumentation() const LLVM_READONLY {
     return !isInvalid() && !isOrdinary();
   }
 
@@ -91,11 +92,21 @@ public:
   unsigned getBeginLine(const SourceManager &SM) const;
   unsigned getEndLine(const SourceManager &SM) const;
 
+  StringRef getBriefText(const ASTContext &Context) const {
+    if (BriefTextValid)
+      return BriefText;
+
+    return extractBriefText(Context);
+  }
+
 private:
   SourceRange Range;
 
   mutable StringRef RawText;
-  mutable bool RawTextValid : 1; ///< True if RawText is valid
+  mutable StringRef BriefText;
+
+  mutable bool RawTextValid : 1;   ///< True if RawText is valid
+  mutable bool BriefTextValid : 1; ///< True if BriefText is valid
 
   unsigned Kind : 3;
 
@@ -118,6 +129,8 @@ private:
 
   StringRef getRawTextSlow(const SourceManager &SourceMgr) const;
 
+  StringRef extractBriefText(const ASTContext &Context) const;
+
   friend class ASTReader;
 };
 
diff --git a/lib/AST/ASTContext.cpp b/lib/AST/ASTContext.cpp
index 23751a56f2..ca631736d9 100644
--- a/lib/AST/ASTContext.cpp
+++ b/lib/AST/ASTContext.cpp
@@ -90,7 +90,7 @@ const RawComment *ASTContext::getRawCommentForDeclNoCache(const Decl *D) const {
 
   // First check whether we have a trailing comment.
   if (Comment != RawComments.end() &&
-      Comment->isDoxygen() && Comment->isTrailingComment() &&
+      Comment->isDocumentation() && Comment->isTrailingComment() &&
       !isa<TagDecl>(D) && !isa<NamespaceDecl>(D)) {
     std::pair<FileID, unsigned> CommentBeginDecomp
       = SourceMgr.getDecomposedLoc(Comment->getSourceRange().getBegin());
@@ -111,7 +111,7 @@ const RawComment *ASTContext::getRawCommentForDeclNoCache(const Decl *D) const {
   --Comment;
 
   // Check that we actually have a non-member Doxygen comment.
-  if (!Comment->isDoxygen() || Comment->isTrailingComment())
+  if (!Comment->isDocumentation() || Comment->isTrailingComment())
     return NULL;
 
   // Decompose the end of the comment.
diff --git a/lib/AST/CMakeLists.txt b/lib/AST/CMakeLists.txt
index d8605367a7..5dad60c490 100644
--- a/lib/AST/CMakeLists.txt
+++ b/lib/AST/CMakeLists.txt
@@ -8,6 +8,8 @@ add_clang_library(clangAST
   ASTImporter.cpp
   AttrImpl.cpp
   CXXInheritance.cpp
+  CommentBriefParser.cpp
+  CommentLexer.cpp
   Decl.cpp
   DeclarationName.cpp
   DeclBase.cpp
diff --git a/lib/AST/CommentBriefParser.cpp b/lib/AST/CommentBriefParser.cpp
new file mode 100644
index 0000000000..528fd2606f
--- /dev/null
+++ b/lib/AST/CommentBriefParser.cpp
@@ -0,0 +1,76 @@
+//===--- CommentBriefParser.cpp - Dumb comment parser ---------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/AST/CommentBriefParser.h"
+
+namespace clang {
+namespace comments {
+
+std::string BriefParser::Parse() {
+  std::string FirstParagraph;
+  std::string Brief;
+  bool InFirstParagraph = true;
+  bool InBrief = false;
+  bool BriefDone = false;
+
+  while (Tok.isNot(tok::eof)) {
+    if (Tok.is(tok::text)) {
+      if (InFirstParagraph)
+        FirstParagraph += Tok.getText();
+      if (InBrief)
+        Brief += Tok.getText();
+      ConsumeToken();
+      continue;
+    }
+
+    if (!BriefDone && Tok.is(tok::command) && Tok.getCommandName() == "brief") {
+      InBrief = true;
+      ConsumeToken();
+      continue;
+    }
+
+    if (Tok.is(tok::newline)) {
+      if (InFirstParagraph)
+        FirstParagraph += '\n';
+      if (InBrief)
+        Brief += '\n';
+      ConsumeToken();
+
+      if (Tok.is(tok::newline)) {
+        ConsumeToken();
+        // We found a paragraph end.
+        InFirstParagraph = false;
+        if (InBrief) {
+          InBrief = false;
+          BriefDone = true;
+        }
+      }
+      continue;
+    }
+
+    // We didn't handle this token, so just drop it.
+    ConsumeToken();
+  }
+
+  if (Brief.size() > 0)
+    return Brief;
+
+  return FirstParagraph;
+}
+
+BriefParser::BriefParser(Lexer &L) : L(L)
+{
+  // Get lookahead token.
+  ConsumeToken();
+}
+
+} // end namespace comments
+} // end namespace clang
+
+
diff --git a/lib/AST/CommentLexer.cpp b/lib/AST/CommentLexer.cpp
new file mode 100644
index 0000000000..e5529dad15
--- /dev/null
+++ b/lib/AST/CommentLexer.cpp
@@ -0,0 +1,676 @@
+#include "clang/AST/CommentLexer.h"
+#include "llvm/ADT/StringSwitch.h"
+#include "llvm/Support/ErrorHandling.h"
+
+namespace clang {
+namespace comments {
+
+void Token::dump(const Lexer &L, const SourceManager &SM) const {
+  llvm::errs() << "comments::Token Kind=" << Kind << " ";
+  Loc.dump(SM);
+  llvm::errs() << " " << Length << " \"" << L.getSpelling(*this, SM) << "\"\n";
+}
+
+bool Lexer::isVerbatimBlockCommand(StringRef BeginName,
+                                  StringRef &EndName) const {
+  const char *Result = llvm::StringSwitch<const char *>(BeginName)
+    .Case("code", "endcode")
+    .Case("verbatim", "endverbatim")
+    .Case("htmlonly", "endhtmlonly")
+    .Case("latexonly", "endlatexonly")
+    .Case("xmlonly", "endxmlonly")
+    .Case("manonly", "endmanonly")
+    .Case("rtfonly", "endrtfonly")
+
+    .Case("dot", "enddot")
+    .Case("msc", "endmsc")
+
+    .Case("f$", "f$") // Inline LaTeX formula
+    .Case("f[", "f]") // Displayed LaTeX formula
+    .Case("f{", "f}") // LaTeX environment
+
+    .Default(NULL);
+
+  if (Result) {
+    EndName = Result;
+    return true;
+  }
+
+  for (VerbatimBlockCommandVector::const_iterator
+           I = VerbatimBlockCommands.begin(),
+           E = VerbatimBlockCommands.end();
+       I != E; ++I)
+    if (I->BeginName == BeginName) {
+      EndName = I->EndName;
+      return true;
+    }
+
+  return false;
+}
+
+bool Lexer::isVerbatimLineCommand(StringRef Name) const {
+  bool Result = llvm::StringSwitch<bool>(Name)
+  .Case("fn", true)
+  .Case("var", true)
+  .Case("property", true)
+  .Case("typedef", true)
+
+  .Case("overload", true)
+
+  .Case("defgroup", true)
+  .Case("ingroup", true)
+  .Case("addtogroup", true)
+  .Case("weakgroup", true)
+  .Case("name", true)
+
+  .Case("section", true)
+  .Case("subsection", true)
+  .Case("subsubsection", true)
+  .Case("paragraph", true)
+
+  .Case("mainpage", true)
+  .Case("subpage", true)
+  .Case("ref", true)
+
+  .Default(false);
+
+  if (Result)
+    return true;
+
+  for (VerbatimLineCommandVector::const_iterator
+           I = VerbatimLineCommands.begin(),
+           E = VerbatimLineCommands.end();
+       I != E; ++I)
+    if (I->Name == Name)
+      return true;
+
+  return false;
+}
+
+void Lexer::skipLineStartingDecorations() {
+  // This function should be called only for C comments
+  assert(CommentState == LCS_InsideCComment);
+
+  if (BufferPtr == CommentEnd)
+    return;
+
+  switch (*BufferPtr) {
+  case ' ':
+  case '\t':
+  case '\f':
+  case '\v': {
+    const char *NewBufferPtr = BufferPtr;
+    NewBufferPtr++;
+    if (NewBufferPtr == CommentEnd)
+      return;
+
+    char C = *NewBufferPtr;
+    while (C == ' ' || C == '\t' || C == '\f' || C == '\v') {
+      NewBufferPtr++;
+      if (NewBufferPtr == CommentEnd)
+        return;
+      C = *NewBufferPtr;
+    }
+    if (C == '*')
+      BufferPtr = NewBufferPtr + 1;
+    break;
+  }
+  case '*':
+    BufferPtr++;
+    break;
+  }
+}
+
+namespace {
+const char *findNewline(const char *BufferPtr, const char *BufferEnd) {
+  for ( ; BufferPtr != BufferEnd; ++BufferPtr) {
+    const char C = *BufferPtr;
+    if (C == '\n' || C == '\r')
+      return BufferPtr;
+  }
+  return BufferEnd;
+}
+
+const char *skipNewline(const char *BufferPtr, const char *BufferEnd) {
+  if (BufferPtr == BufferEnd)
+    return BufferPtr;
+
+  if (*BufferPtr == '\n')
+    BufferPtr++;
+  else {
+    assert(*BufferPtr == '\r');
+    BufferPtr++;
+    if (BufferPtr != BufferEnd && *BufferPtr == '\n')
+      BufferPtr++;
+  }
+  return BufferPtr;
+}
+
+bool isHTMLIdentifierCharacter(char C) {
+  return (C >= 'a' && C <= 'z') ||
+         (C >= 'A' && C <= 'Z') ||
+         (C >= '0' && C <= '9');
+}
+
+const char *skipHTMLIdentifier(const char *BufferPtr, const char *BufferEnd) {
+  for ( ; BufferPtr != BufferEnd; ++BufferPtr) {
+    if (!isHTMLIdentifierCharacter(*BufferPtr))
+      return BufferPtr;
+  }
+  return BufferEnd;
+}
+
+/// Skip HTML string quoted in single or double quotes.  Escaping quotes inside
+/// string allowed.
+///
+/// Returns pointer to closing quote.
+const char *skipHTMLQuotedString(const char *BufferPtr, const char *BufferEnd)
+{
+  const char Quote = *BufferPtr;
+  assert(Quote == '\"' || Quote == '\'');
+
+  BufferPtr++;
+  for ( ; BufferPtr != BufferEnd; ++BufferPtr) {
+    const char C = *BufferPtr;
+    if (C == Quote && BufferPtr[-1] != '\\')
+      return BufferPtr;
+  }
+  return BufferEnd;
+}
+
+bool isHorizontalWhitespace(char C) {
+  return C == ' ' || C == '\t' || C == '\f' || C == '\v';
+}
+
+bool isWhitespace(char C) {
+  return C == ' ' || C == '\n' || C == '\r' ||
+         C == '\t' || C == '\f' || C == '\v';
+}
+
+const char *skipWhitespace(const char *BufferPtr, const char *BufferEnd) {
+  for ( ; BufferPtr != BufferEnd; ++BufferPtr) {
+    if (!isWhitespace(*BufferPtr))
+      return BufferPtr;
+  }
+  return BufferEnd;
+}
+
+bool isCommandNameCharacter(char C) {
+  return (C >= 'a' && C <= 'z') ||
+         (C >= 'A' && C <= 'Z') ||
+         (C >= '0' && C <= '9');
+}
+
+const char *skipCommandName(const char *BufferPtr, const char *BufferEnd) {
+  for ( ; BufferPtr != BufferEnd; ++BufferPtr) {
+    if (!isCommandNameCharacter(*BufferPtr))
+      return BufferPtr;
+  }
+  return BufferEnd;
+}
+
+/// Return the one past end pointer for BCPL comments.
+/// Handles newlines escaped with backslash or trigraph for backslahs.
+const char *findBCPLCommentEnd(const char *BufferPtr, const char *BufferEnd) {
+  const char *CurPtr = BufferPtr;
+  while (CurPtr != BufferEnd) {
+    char C = *CurPtr;
+    while (C != '\n' && C != '\r') {
+      CurPtr++;
+      if (CurPtr == BufferEnd)
+        return BufferEnd;
+      C = *CurPtr;
+    }
+    // We found a newline, check if it is escaped.
+    const char *EscapePtr = CurPtr - 1;
+    while(isHorizontalWhitespace(*EscapePtr))
+      EscapePtr--;
+
+    if (*EscapePtr == '\\' ||
+        (EscapePtr - 2 >= BufferPtr && EscapePtr[0] == '/' &&
+         EscapePtr[-1] == '?' && EscapePtr[-2] == '?')) {
+      // We found an escaped newline.
+      CurPtr = skipNewline(CurPtr, BufferEnd);
+    } else
+      return CurPtr; // Not an escaped newline.
+  }
+  return BufferEnd;
+}
+
+/// Return the one past end pointer for C comments.
+/// Very dumb, does not handle escaped newlines or trigraphs.
+const char *findCCommentEnd(const char *BufferPtr, const char *BufferEnd) {
+  for ( ; BufferPtr != BufferEnd; ++BufferPtr) {
+    if (*BufferPtr == '*') {
+      assert(BufferPtr + 1 != BufferEnd);
+      if (*(BufferPtr + 1) == '/')
+        return BufferPtr;
+    }
+  }
+  llvm_unreachable("buffer end hit before '*/' was seen");
+}
+} // unnamed namespace
+
+void Lexer::lexCommentText(Token &T) {
+  assert(CommentState == LCS_InsideBCPLComment ||
+         CommentState == LCS_InsideCComment);
+
+  switch (State) {
+  case LS_Normal:
+    break;
+  case LS_VerbatimBlockFirstLine:
+    lexVerbatimBlockFirstLine(T);
+    return;
+  case LS_VerbatimBlockBody:
+    lexVerbatimBlockBody(T);
+    return;
+  case LS_HTMLOpenTag:
+    lexHTMLOpenTag(T);
+    return;
+  }
+
+  assert(State == LS_Normal);
+
+  const char *TokenPtr = BufferPtr;
+  assert(TokenPtr < CommentEnd);
+  while (TokenPtr != CommentEnd) {
+    switch(*TokenPtr) {
+      case '\\':
+      case '@': {
+        TokenPtr++;
+        if (TokenPtr == CommentEnd) {
+          formTokenWithChars(T, TokenPtr, tok::text);
+          T.setText(StringRef(BufferPtr - T.getLength(), T.getLength()));
+          return;
+        }
+        char C = *TokenPtr;
+        switch (C) {
+        default:
+          break;
+
+        case '\\': case '@': case '&': case '$':
+        case '#':  case '<': case '>': case '%':
+        case '\"': case '.': case ':':
+          // This is one of \\ \@ \& \$ etc escape sequences.
+          TokenPtr++;
+          if (C == ':' && TokenPtr != CommentEnd && *TokenPtr == ':') {
+            // This is the \:: escape sequence.
+            TokenPtr++;
+          }
+          formTokenWithChars(T, TokenPtr, tok::text);
+          T.setText(StringRef(BufferPtr - (T.getLength() - 1),
+                              T.getLength() - 1));
+          return;
+        }
+
+        // Don't make zero-length commands.
+        if (!isCommandNameCharacter(*TokenPtr)) {
+          formTokenWithChars(T, TokenPtr, tok::text);
+          T.setText(StringRef(BufferPtr - T.getLength(), T.getLength()));
+          return;
+        }
+
+        TokenPtr = skipCommandName(TokenPtr, CommentEnd);
+        unsigned Length = TokenPtr - (BufferPtr + 1);
+
+        // Hardcoded support for lexing LaTeX formula commands
+        // \f$ \f[ \f] \f{ \f} as a single command.
+        if (Length == 1 && TokenPtr[-1] == 'f' && TokenPtr != CommentEnd) {
+          C = *TokenPtr;
+          if (C == '$' || C == '[' || C == ']' || C == '{' || C == '}') {
+            TokenPtr++;
+            Length++;
+          }
+        }
+
+        const StringRef CommandName(BufferPtr + 1, Length);
+        StringRef EndName;
+
+        if (isVerbatimBlockCommand(CommandName, EndName)) {
+          setupAndLexVerbatimBlock(T, TokenPtr, *BufferPtr, EndName);
+          return;
+        }
+        if (isVerbatimLineCommand(CommandName)) {
+          lexVerbatimLine(T, TokenPtr);
+          return;
+        }
+        formTokenWithChars(T, TokenPtr, tok::command);
+        T.setCommandName(CommandName);
+        return;
+      }
+
+      case '<': {
+        TokenPtr++;
+        if (TokenPtr == CommentEnd) {
+          formTokenWithChars(T, TokenPtr, tok::text);
+          T.setText(StringRef(BufferPtr - T.getLength(), T.getLength()));
+          return;
+        }
+        const char C = *TokenPtr;
+        if (isHTMLIdentifierCharacter(C))
+          setupAndLexHTMLOpenTag(T);
+        else if (C == '/')
+          lexHTMLCloseTag(T);
+        return;
+      }
+
+      case '\n':
+      case '\r':
+        TokenPtr = skipNewline(TokenPtr, CommentEnd);
+        formTokenWithChars(T, TokenPtr, tok::newline);
+
+        if (CommentState == LCS_InsideCComment)
+          skipLineStartingDecorations();
+        return;
+
+      default: {
+        while (true) {
+          TokenPtr++;
+          if (TokenPtr == CommentEnd)
+            break;
+          char C = *TokenPtr;
+          if(C == '\n' || C == '\r' ||
+             C == '\\' || C == '@' || C == '<')
+            break;
+        }
+        formTokenWithChars(T, TokenPtr, tok::text);
+        T.setText(StringRef(BufferPtr - T.getLength(), T.getLength()));
+        return;
+      }
+    }
+  }
+}
+
+void Lexer::setupAndLexVerbatimBlock(Token &T,
+                                     const char *TextBegin,
+                                     char Mark