vendor/clang/clang-trunk-r256633

15 files changed, 699 insertions, 379 deletions

diff --git a/utils/TableGen/ClangAttrEmitter.cpp b/utils/TableGen/ClangAttrEmitter.cpp
index f2aa4002b99f..f70bff2c3b74 100644
--- a/utils/TableGen/ClangAttrEmitter.cpp
+++ b/utils/TableGen/ClangAttrEmitter.cpp
@@ -54,7 +54,7 @@ public:
   const std::string &nameSpace() const { return NS; }
   bool knownToGCC() const { return K; }
 };
-} // namespace
+} // end anonymous namespace
 
 static std::vector<FlattenedSpelling>
 GetFlattenedSpellings(const Record &Attr) {
@@ -166,17 +166,18 @@ namespace {
     std::string lowerName, upperName;
     StringRef attrName;
     bool isOpt;
+    bool Fake;
 
   public:
     Argument(const Record &Arg, StringRef Attr)
       : lowerName(Arg.getValueAsString("Name")), upperName(lowerName),
-        attrName(Attr), isOpt(false) {
+        attrName(Attr), isOpt(false), Fake(false) {
       if (!lowerName.empty()) {
         lowerName[0] = std::tolower(lowerName[0]);
         upperName[0] = std::toupper(upperName[0]);
       }
     }
-    virtual ~Argument() {}
+    virtual ~Argument() = default;
 
     StringRef getLowerName() const { return lowerName; }
     StringRef getUpperName() const { return upperName; }
@@ -185,6 +186,9 @@ namespace {
     bool isOptional() const { return isOpt; }
     void setOptional(bool set) { isOpt = set; }
 
+    bool isFake() const { return Fake; }
+    void setFake(bool fake) { Fake = fake; }
+
     // These functions print the argument contents formatted in different ways.
     virtual void writeAccessors(raw_ostream &OS) const = 0;
     virtual void writeAccessorDefinitions(raw_ostream &OS) const {}
@@ -275,6 +279,8 @@ namespace {
         OS << "    OS << \" \";\n";
         OS << "    dumpBareDeclRef(SA->get" << getUpperName() << "());\n"; 
       } else if (type == "IdentifierInfo *") {
+        if (isOptional())
+          OS << "    if (SA->get" << getUpperName() << "())\n  ";
         OS << "    OS << \" \" << SA->get" << getUpperName()
            << "()->getName();\n";
       } else if (type == "TypeSourceInfo *") {
@@ -348,7 +354,7 @@ namespace {
          << "Length])";
     }
     void writeCtorDefaultInitializers(raw_ostream &OS) const override {
-      OS << getLowerName() << "Length(0)," << getLowerName() << "(0)";
+      OS << getLowerName() << "Length(0)," << getLowerName() << "(nullptr)";
     }
     void writeCtorParameters(raw_ostream &OS) const override {
       OS << "llvm::StringRef " << getUpperName();
@@ -491,7 +497,7 @@ namespace {
       // The aligned attribute argument expression is optional.
       OS << "    if (is" << getLowerName() << "Expr && "
          << getLowerName() << "Expr)\n";
-      OS << "      " << getLowerName() << "Expr->printPretty(OS, 0, Policy);\n";
+      OS << "      " << getLowerName() << "Expr->printPretty(OS, nullptr, Policy);\n";
       OS << "    OS << \"";
     }
     void writeDump(raw_ostream &OS) const override {
@@ -613,7 +619,7 @@ namespace {
     std::vector<std::string> uniques;
     std::set<std::string> unique_set(enums.begin(), enums.end());
     for (const auto &i : enums) {
-      std::set<std::string>::iterator set_i = unique_set.find(i);
+      auto set_i = unique_set.find(i);
       if (set_i != unique_set.end()) {
         uniques.push_back(i);
         unique_set.erase(set_i);
@@ -659,8 +665,7 @@ namespace {
       OS << type << " " << getUpperName();
     }
     void writeDeclarations(raw_ostream &OS) const override {
-      std::vector<std::string>::const_iterator i = uniques.begin(),
-                                               e = uniques.end();
+      auto i = uniques.cbegin(), e = uniques.cend();
       // The last one needs to not have a comma.
       --e;
 
@@ -765,8 +770,7 @@ namespace {
     bool isVariadicEnumArg() const override { return true; }
     
     void writeDeclarations(raw_ostream &OS) const override {
-      std::vector<std::string>::const_iterator i = uniques.begin(),
-                                               e = uniques.end();
+      auto i = uniques.cbegin(), e = uniques.cend();
       // The last one needs to not have a comma.
       --e;
 
@@ -952,7 +956,7 @@ namespace {
     }
 
     void writeTemplateInstantiation(raw_ostream &OS) const override {
-      OS << "      " << getType() << " *tempInst" << getUpperName()
+      OS << "      auto *tempInst" << getUpperName()
          << " = new (C, 16) " << getType()
          << "[A->" << getLowerName() << "_size()];\n";
       OS << "      {\n";
@@ -1018,7 +1022,7 @@ namespace {
           getType(), "SA->get" + std::string(getUpperName()) + "Loc()");
     }
   };
-}
+} // end anonymous namespace
 
 static std::unique_ptr<Argument>
 createArgument(const Record &Arg, StringRef Attr,
@@ -1078,6 +1082,9 @@ createArgument(const Record &Arg, StringRef Attr,
   if (Ptr && Arg.getValueAsBit("Optional"))
     Ptr->setOptional(true);
 
+  if (Ptr && Arg.getValueAsBit("Fake"))
+    Ptr->setFake(true);
+
   return Ptr;
 }
 
@@ -1180,28 +1187,39 @@ writePrettyPrintFunction(Record &R,
     if (Variety == "Pragma") {
       OS << " \";\n";
       OS << "    printPrettyPragma(OS, Policy);\n";
+      OS << "    OS << \"\\n\";";
       OS << "    break;\n";
       OS << "  }\n";
       continue;
     }
 
+    // Fake arguments aren't part of the parsed form and should not be
+    // pretty-printed.
+    bool hasNonFakeArgs = false;
+    for (const auto &arg : Args) {
+      if (arg->isFake()) continue;
+      hasNonFakeArgs = true;
+    }
+
     // FIXME: always printing the parenthesis isn't the correct behavior for
     // attributes which have optional arguments that were not provided. For
     // instance: __attribute__((aligned)) will be pretty printed as
     // __attribute__((aligned())). The logic should check whether there is only
     // a single argument, and if it is optional, whether it has been provided.
-    if (!Args.empty())
+    if (hasNonFakeArgs)
       OS << "(";
     if (Spelling == "availability") {
       writeAvailabilityValue(OS);
     } else {
-      for (auto I = Args.begin(), E = Args.end(); I != E; ++ I) {
-        if (I != Args.begin()) OS << ", ";
-        (*I)->writeValue(OS);
+      unsigned index = 0;
+      for (const auto &arg : Args) {
+        if (arg->isFake()) continue;
+        if (index++) OS << ", ";
+        arg->writeValue(OS);
       }
     }
 
-    if (!Args.empty())
+    if (hasNonFakeArgs)
       OS << ")";
     OS << Suffix + "\";\n";
 
@@ -1473,10 +1491,19 @@ void EmitClangAttrClass(RecordKeeper &Records, raw_ostream &OS) {
     std::vector<std::unique_ptr<Argument>> Args;
     Args.reserve(ArgRecords.size());
 
+    bool HasOptArg = false;
+    bool HasFakeArg = false;
     for (const auto *ArgRecord : ArgRecords) {
       Args.emplace_back(createArgument(*ArgRecord, R.getName()));
       Args.back()->writeDeclarations(OS);
       OS << "\n\n";
+
+      // For these purposes, fake takes priority over optional.
+      if (Args.back()->isFake()) {
+        HasFakeArg = true;
+      } else if (Args.back()->isOptional()) {
+        HasOptArg = true;
+      }
     }
 
     OS << "\npublic:\n";
@@ -1495,69 +1522,53 @@ void EmitClangAttrClass(RecordKeeper &Records, raw_ostream &OS) {
     if (!ElideSpelling)
       OS << CreateSemanticSpellings(Spellings, SemanticToSyntacticMap);
 
-    OS << "  static " << R.getName() << "Attr *CreateImplicit(";
-    OS << "ASTContext &Ctx";
-    if (!ElideSpelling)
-      OS << ", Spelling S";
-    for (auto const &ai : Args) {
-      OS << ", ";
-      ai->writeCtorParameters(OS);
-    }
-    OS << ", SourceRange Loc = SourceRange()";
-    OS << ") {\n";
-    OS << "    " << R.getName() << "Attr *A = new (Ctx) " << R.getName();
-    OS << "Attr(Loc, Ctx, ";
-    for (auto const &ai : Args) {
-      ai->writeImplicitCtorArgs(OS);
-      OS << ", ";
-    }
-    OS << (ElideSpelling ? "0" : "S") << ");\n";
-    OS << "    A->setImplicit(true);\n";
-    OS << "    return A;\n  }\n\n";
-
-    OS << "  " << R.getName() << "Attr(SourceRange R, ASTContext &Ctx\n";
-    
-    bool HasOpt = false;
-    for (auto const &ai : Args) {
-      OS << "              , ";
-      ai->writeCtorParameters(OS);
-      OS << "\n";
-      if (ai->isOptional())
-        HasOpt = true;
-    }
-
-    OS << "              , ";
-    OS << "unsigned SI\n";
+    // Emit CreateImplicit factory methods.
+    auto emitCreateImplicit = [&](bool emitFake) {
+      OS << "  static " << R.getName() << "Attr *CreateImplicit(";
+      OS << "ASTContext &Ctx";
+      if (!ElideSpelling)
+        OS << ", Spelling S";
+      for (auto const &ai : Args) {
+        if (ai->isFake() && !emitFake) continue;
+        OS << ", ";
+        ai->writeCtorParameters(OS);
+      }
+      OS << ", SourceRange Loc = SourceRange()";
+      OS << ") {\n";
+      OS << "    auto *A = new (Ctx) " << R.getName();
+      OS << "Attr(Loc, Ctx, ";
+      for (auto const &ai : Args) {
+        if (ai->isFake() && !emitFake) continue;
+        ai->writeImplicitCtorArgs(OS);
+        OS << ", ";
+      }
+      OS << (ElideSpelling ? "0" : "S") << ");\n";
+      OS << "    A->setImplicit(true);\n";
+      OS << "    return A;\n  }\n\n";
+    };
 
-    OS << "             )\n";
-    OS << "    : " << SuperName << "(attr::" << R.getName() << ", R, SI, "
-       << R.getValueAsBit("LateParsed") << ", "
-       << R.getValueAsBit("DuplicatesAllowedWhileMerging") << ")\n";
+    // Emit a CreateImplicit that takes all the arguments.
+    emitCreateImplicit(true);
 
-    for (auto const &ai : Args) {
-      OS << "              , ";
-      ai->writeCtorInitializers(OS);
-      OS << "\n";
+    // Emit a CreateImplicit that takes all the non-fake arguments.
+    if (HasFakeArg) {
+      emitCreateImplicit(false);
     }
 
-    OS << "  {\n";
-  
-    for (auto const &ai : Args) {
-      ai->writeCtorBody(OS);
-      OS << "\n";
-    }
-    OS << "  }\n\n";
+    // Emit constructors.
+    auto emitCtor = [&](bool emitOpt, bool emitFake) {
+      auto shouldEmitArg = [=](const std::unique_ptr<Argument> &arg) {
+        if (arg->isFake()) return emitFake;
+        if (arg->isOptional()) return emitOpt;
+        return true;
+      };
 
-    // If there are optional arguments, write out a constructor that elides the
-    // optional arguments as well.
-    if (HasOpt) {
       OS << "  " << R.getName() << "Attr(SourceRange R, ASTContext &Ctx\n";
       for (auto const &ai : Args) {
-        if (!ai->isOptional()) {
-          OS << "              , ";
-          ai->writeCtorParameters(OS);
-          OS << "\n";
-        }
+        if (!shouldEmitArg(ai)) continue;
+        OS << "              , ";
+        ai->writeCtorParameters(OS);
+        OS << "\n";
       }
 
       OS << "              , ";
@@ -1570,19 +1581,37 @@ void EmitClangAttrClass(RecordKeeper &Records, raw_ostream &OS) {
 
       for (auto const &ai : Args) {
         OS << "              , ";
-        ai->writeCtorDefaultInitializers(OS);
+        if (!shouldEmitArg(ai)) {
+          ai->writeCtorDefaultInitializers(OS);
+        } else {
+          ai->writeCtorInitializers(OS);
+        }
         OS << "\n";
       }
 
       OS << "  {\n";
   
       for (auto const &ai : Args) {
-        if (!ai->isOptional()) {
-          ai->writeCtorBody(OS);
-          OS << "\n";
-        }
+        if (!shouldEmitArg(ai)) continue;
+        ai->writeCtorBody(OS);
+        OS << "\n";
       }
       OS << "  }\n\n";
+
+    };
+
+    // Emit a constructor that includes all the arguments.
+    // This is necessary for cloning.
+    emitCtor(true, true);
+
+    // Emit a constructor that takes all the non-fake arguments.
+    if (HasFakeArg) {
+      emitCtor(true, false);
+    }
+ 
+    // Emit a constructor that takes all the non-fake, non-optional arguments.
+    if (HasOptArg) {
+      emitCtor(false, false);
     }
 
     OS << "  " << R.getName() << "Attr *clone(ASTContext &C) const;\n";
@@ -1604,6 +1633,9 @@ void EmitClangAttrClass(RecordKeeper &Records, raw_ostream &OS) {
       ai->writeAccessors(OS);
       OS << "\n\n";
 
+      // Don't write conversion routines for fake arguments.
+      if (ai->isFake()) continue;
+
       if (ai->isEnumArg())
         static_cast<const EnumArgument *>(ai.get())->writeConversion(OS);
       else if (ai->isVariadicEnumArg())
@@ -1620,7 +1652,7 @@ void EmitClangAttrClass(RecordKeeper &Records, raw_ostream &OS) {
     OS << "};\n\n";
   }
 
-  OS << "#endif\n";
+  OS << "#endif // LLVM_CLANG_ATTR_CLASSES_INC\n";
 }
 
 // Emits the class method definitions for attributes.
@@ -1695,7 +1727,7 @@ void EmitClangAttrImpl(RecordKeeper &Records, raw_ostream &OS) {
 
 static void EmitAttrList(raw_ostream &OS, StringRef Class,
                          const std::vector<Record*> &AttrList) {
-  std::vector<Record*>::const_iterator i = AttrList.begin(), e = AttrList.end();
+  auto i = AttrList.cbegin(), e = AttrList.cend();
 
   if (i != e) {
     // Move the end iterator back to emit the last attribute.
@@ -1847,7 +1879,7 @@ void EmitClangAttrPCHWrite(RecordKeeper &Records, raw_ostream &OS) {
     OS << "  case attr::" << R.getName() << ": {\n";
     Args = R.getValueAsListOfDefs("Args");
     if (R.isSubClassOf(InhClass) || !Args.empty())
-      OS << "    const " << R.getName() << "Attr *SA = cast<" << R.getName()
+      OS << "    const auto *SA = cast<" << R.getName()
          << "Attr>(A);\n";
     if (R.isSubClassOf(InhClass))
       OS << "    Record.push_back(SA->isInherited());\n";
@@ -1862,6 +1894,65 @@ void EmitClangAttrPCHWrite(RecordKeeper &Records, raw_ostream &OS) {
   OS << "  }\n";
 }
 
+// Generate a conditional expression to check if the current target satisfies
+// the conditions for a TargetSpecificAttr record, and append the code for
+// those checks to the Test string. If the FnName string pointer is non-null,
+// append a unique suffix to distinguish this set of target checks from other
+// TargetSpecificAttr records.
+static void GenerateTargetSpecificAttrChecks(const Record *R,
+                                             std::vector<std::string> &Arches,
+                                             std::string &Test,
+                                             std::string *FnName) {
+  // It is assumed that there will be an llvm::Triple object
+  // named "T" and a TargetInfo object named "Target" within
+  // scope that can be used to determine whether the attribute exists in
+  // a given target.
+  Test += "(";
+
+  for (auto I = Arches.begin(), E = Arches.end(); I != E; ++I) {
+    std::string Part = *I;
+    Test += "T.getArch() == llvm::Triple::" + Part;
+    if (I + 1 != E)
+      Test += " || ";
+    if (FnName)
+      *FnName += Part;
+  }
+  Test += ")";
+
+  // If the attribute is specific to particular OSes, check those.
+  if (!R->isValueUnset("OSes")) {
+    // We know that there was at least one arch test, so we need to and in the
+    // OS tests.
+    Test += " && (";
+    std::vector<std::string> OSes = R->getValueAsListOfStrings("OSes");
+    for (auto I = OSes.begin(), E = OSes.end(); I != E; ++I) {
+      std::string Part = *I;
+
+      Test += "T.getOS() == llvm::Triple::" + Part;
+      if (I + 1 != E)
+        Test += " || ";
+      if (FnName)
+        *FnName += Part;
+    }
+    Test += ")";
+  }
+
+  // If one or more CXX ABIs are specified, check those as well.
+  if (!R->isValueUnset("CXXABIs")) {
+    Test += " && (";
+    std::vector<std::string> CXXABIs = R->getValueAsListOfStrings("CXXABIs");
+    for (auto I = CXXABIs.begin(), E = CXXABIs.end(); I != E; ++I) {
+      std::string Part = *I;
+      Test += "Target.getCXXABI().getKind() == TargetCXXABI::" + Part;
+      if (I + 1 != E)
+        Test += " || ";
+      if (FnName)
+        *FnName += Part;
+    }
+    Test += ")";
+  }
+}
+
 static void GenerateHasAttrSpellingStringSwitch(
     const std::vector<Record *> &Attrs, raw_ostream &OS,
     const std::string &Variety = "", const std::string &Scope = "") {
@@ -1887,37 +1978,12 @@ static void GenerateHasAttrSpellingStringSwitch(
       }
     }
 
-    // It is assumed that there will be an llvm::Triple object named T within
-    // scope that can be used to determine whether the attribute exists in
-    // a given target.
     std::string Test;
     if (Attr->isSubClassOf("TargetSpecificAttr")) {
       const Record *R = Attr->getValueAsDef("Target");
       std::vector<std::string> Arches = R->getValueAsListOfStrings("Arches");
+      GenerateTargetSpecificAttrChecks(R, Arches, Test, nullptr);
 
-      Test += "(";
-      for (auto AI = Arches.begin(), AE = Arches.end(); AI != AE; ++AI) {
-        std::string Part = *AI;
-        Test += "T.getArch() == llvm::Triple::" + Part;
-        if (AI + 1 != AE)
-          Test += " || ";
-      }
-      Test += ")";
-
-      std::vector<std::string> OSes;
-      if (!R->isValueUnset("OSes")) {
-        Test += " && (";
-        std::vector<std::string> OSes = R->getValueAsListOfStrings("OSes");
-        for (auto AI = OSes.begin(), AE = OSes.end(); AI != AE; ++AI) {
-          std::string Part = *AI;
-
-          Test += "T.getOS() == llvm::Triple::" + Part;
-          if (AI + 1 != AE)
-            Test += " || ";
-        }
-        Test += ")";
-      }
-      
       // If this is the C++11 variety, also add in the LangOpts test.
       if (Variety == "CXX11")
         Test += " && LangOpts.CPlusPlus11";
@@ -1964,6 +2030,7 @@ void EmitClangAttrHasAttrImpl(RecordKeeper &Records, raw_ostream &OS) {
     }
   }
 
+  OS << "const llvm::Triple &T = Target.getTriple();\n";
   OS << "switch (Syntax) {\n";
   OS << "case AttrSyntax::GNU:\n";
   OS << "  return llvm::StringSwitch<int>(Name)\n";
@@ -1976,9 +2043,7 @@ void EmitClangAttrHasAttrImpl(RecordKeeper &Records, raw_ostream &OS) {
   GenerateHasAttrSpellingStringSwitch(Pragma, OS, "Pragma");
   OS << "case AttrSyntax::CXX: {\n";
   // C++11-style attributes are further split out based on the Scope.
-  for (std::map<std::string, std::vector<Record *>>::iterator I = CXX.begin(),
-                                                              E = CXX.end();
-       I != E; ++I) {
+  for (auto I = CXX.cbegin(), E = CXX.cend(); I != E; ++I) {
     if (I != CXX.begin())
       OS << " else ";
     if (I->first.empty())
@@ -2049,7 +2114,7 @@ void EmitClangAttrASTVisitor(RecordKeeper &Records, raw_ostream &OS) {
     OS << "  bool Visit"
        << R.getName() << "Attr(" << R.getName() << "Attr *A) {\n"
        << "    return true; \n"
-       << "  };\n";
+       << "  }\n";
   }
   OS << "\n#else // ATTR_VISITOR_DECLS_ONLY\n\n";
 
@@ -2123,12 +2188,12 @@ void EmitClangAttrTemplateInstantiate(RecordKeeper &Records, raw_ostream &OS) {
     bool ShouldClone = R.getValueAsBit("Clone");
 
     if (!ShouldClone) {
-      OS << "      return NULL;\n";
+      OS << "      return nullptr;\n";
       OS << "    }\n";
       continue;
     }
 
-    OS << "      const " << R.getName() << "Attr *A = cast<"
+    OS << "      const auto *A = cast<"
        << R.getName() << "Attr>(At);\n";
     bool TDependent = R.getValueAsBit("TemplateDependent");
 
@@ -2157,7 +2222,7 @@ void EmitClangAttrTemplateInstantiate(RecordKeeper &Records, raw_ostream &OS) {
   }
   OS << "  } // end switch\n"
      << "  llvm_unreachable(\"Unknown attribute!\");\n"
-     << "  return 0;\n"
+     << "  return nullptr;\n"
      << "}\n\n"
      << "} // end namespace sema\n"
      << "} // end namespace clang\n";
@@ -2328,6 +2393,7 @@ static std::string GetSubjectWithSuffix(const Record *R) {
     return "Decl";
   return B + "Decl";
 }
+
 static std::string GenerateCustomAppertainsTo(const Record &Subject,
                                               raw_ostream &OS) {
   std::string FnName = "is" + Subject.getName();
@@ -2335,7 +2401,7 @@ static std::string GenerateCustomAppertainsTo(const Record &Subject,
   // If this code has already been generated, simply return the previous
   // instance of it.
   static std::set<std::string> CustomSubjectSet;
-  std::set<std::string>::iterator I = CustomSubjectSet.find(FnName);
+  auto I = CustomSubjectSet.find(FnName);
   if (I != CustomSubjectSet.end())
     return *I;
 
@@ -2349,7 +2415,7 @@ static std::string GenerateCustomAppertainsTo(const Record &Subject,
   }
 
   OS << "static bool " << FnName << "(const Decl *D) {\n";
-  OS << "  if (const " << GetSubjectWithSuffix(Base) << " *S = dyn_cast<";
+  OS << "  if (const auto *S = dyn_cast<";
   OS << GetSubjectWithSuffix(Base);
   OS << ">(D))\n";
   OS << "    return " << Subject.getValueAsString("CheckCode") << ";\n";
@@ -2449,7 +2515,7 @@ static std::string GenerateLangOptRequirements(const Record &R,
   // If this code has already been generated, simply return the previous
   // instance of it.
   static std::set<std::string> CustomLangOptsSet;
-  std::set<std::string>::iterator I = CustomLangOptsSet.find(FnName);
+  auto I = CustomLangOptsSet.find(FnName);
   if (I != CustomLangOptsSet.end())
     return *I;
 
@@ -2466,7 +2532,7 @@ static std::string GenerateLangOptRequirements(const Record &R,
 }
 
 static void GenerateDefaultTargetRequirements(raw_ostream &OS) {
-  OS << "static bool defaultTargetRequirements(const llvm::Triple &) {\n";
+  OS << "static bool defaultTargetRequirements(const TargetInfo &) {\n";
   OS << "  return true;\n";
   OS << "}\n\n";
 }
@@ -2505,47 +2571,19 @@ static std::string GenerateTargetRequirements(const Record &Attr,
     }
   }
 
-  std::string FnName = "isTarget", Test = "(";
-  for (auto I = Arches.begin(), E = Arches.end(); I != E; ++I) {
-    std::string Part = *I;
-    Test += "Arch == llvm::Triple::" + Part;
-    if (I + 1 != E)
-      Test += " || ";
-    FnName += Part;
-  }
-  Test += ")";
-
-  // If the target also requires OS testing, generate those tests as well.
-  bool UsesOS = false;
-  if (!R->isValueUnset("OSes")) {
-    UsesOS = true;
-    
-    // We know that there was at least one arch test, so we need to and in the
-    // OS tests.
-    Test += " && (";
-    std::vector<std::string> OSes = R->getValueAsListOfStrings("OSes");
-    for (auto I = OSes.begin(), E = OSes.end(); I != E; ++I) {
-      std::string Part = *I;
-
-      Test += "OS == llvm::Triple::" + Part;
-      if (I + 1 != E)
-        Test += " || ";
-      FnName += Part;
-    }
-    Test += ")";
-  }
+  std::string FnName = "isTarget";
+  std::string Test;
+  GenerateTargetSpecificAttrChecks(R, Arches, Test, &FnName);
 
   // If this code has already been generated, simply return the previous
   // instance of it.
   static std::set<std::string> CustomTargetSet;
-  std::set<std::string>::iterator I = CustomTargetSet.find(FnName);
+  auto I = CustomTargetSet.find(FnName);
   if (I != CustomTargetSet.end())
     return *I;
 
-  OS << "static bool " << FnName << "(const llvm::Triple &T) {\n";
-  OS << "  llvm::Triple::ArchType Arch = T.getArch();\n";
-  if (UsesOS)
-    OS << "  llvm::Triple::OSType OS = T.getOS();\n";
+  OS << "static bool " << FnName << "(const TargetInfo &Target) {\n";
+  OS << "  const llvm::Triple &T = Target.getTriple();\n";
   OS << "  return " << Test << ";\n";
   OS << "}\n\n";
 
@@ -2758,13 +2796,13 @@ void EmitClangAttrDump(RecordKeeper &Records, raw_ostream &OS) {
 
     Args = R.getValueAsListOfDefs("Args");
     if (!Args.empty()) {
-      OS << "    const " << R.getName() << "Attr *SA = cast<" << R.getName()
+      OS << "    const auto *SA = cast<" << R.getName()
          << "Attr>(A);\n";
       for (const auto *Arg : Args)
         createArgument(*Arg, R.getName())->writeDump(OS);
 
-      for (auto AI = Args.begin(), AE = Args.end(); AI != AE; ++AI)
-        createArgument(**AI, R.getName())->writeDumpChildren(OS);
+      for (const auto *AI : Args)
+        createArgument(*AI, R.getName())->writeDumpChildren(OS);
     }
     OS <<
       "    break;\n"
diff --git a/utils/TableGen/ClangCommentCommandInfoEmitter.cpp b/utils/TableGen/ClangCommentCommandInfoEmitter.cpp
index 3349030466fc..3522cd472d9d 100644
--- a/utils/TableGen/ClangCommentCommandInfoEmitter.cpp
+++ b/utils/TableGen/ClangCommentCommandInfoEmitter.cpp
@@ -72,7 +72,7 @@ void EmitClangCommentCommandInfo(RecordKeeper &Records, raw_ostream &OS) {
   OS << "const CommandInfo *CommandTraits::getBuiltinCommandInfo(\n"
      << "                                         StringRef Name) {\n";
   StringMatcher("Name", Matches, OS).Emit();
-  OS << "  return NULL;\n"
+  OS << "  return nullptr;\n"
      << "}\n\n";
 }
 
@@ -123,4 +123,3 @@ void EmitClangCommentCommandList(RecordKeeper &Records, raw_ostream &OS) {
   }
 }
 } // end namespace clang
-
diff --git a/utils/TableGen/NeonEmitter.cpp b/utils/TableGen/NeonEmitter.cpp
index 7644ae2c04da..6e7bc9057fd7 100644
--- a/utils/TableGen/NeonEmitter.cpp
+++ b/utils/TableGen/NeonEmitter.cpp
@@ -36,6 +36,7 @@
 #include "llvm/TableGen/SetTheory.h"
 #include "llvm/TableGen/TableGenBackend.h"
 #include <algorithm>
+#include <deque>
 #include <map>
 #include <sstream>
 #include <string>
@@ -393,7 +394,7 @@ public:
 
   /// Return true if the prototype has a scalar argument.
   /// This does not return true for the "splat" code ('a').
-  bool protoHasScalar();
+  bool protoHasScalar() const;
 
   /// Return the index that parameter PIndex will sit at
   /// in a generated function call. This is often just PIndex,
@@ -431,9 +432,9 @@ public:
   /// Return the name, mangled with type information.
   /// If ForceClassS is true, use ClassS (u32/s32) instead
   /// of the intrinsic's own type class.
-  std::string getMangledName(bool ForceClassS = false);
+  std::string getMangledName(bool ForceClassS = false) const;
   /// Return the type code for a builtin function call.
-  std::string getInstTypeCode(Type T, ClassKind CK);
+  std::string getInstTypeCode(Type T, ClassKind CK) const;
   /// Return the type string for a BUILTIN() macro in Builtins.def.
   std::string getBuiltinTypeStr();
 
@@ -444,7 +445,7 @@ public:
   void indexBody();
 
 private:
-  std::string mangleName(std::string Name, ClassKind CK);
+  std::string mangleName(std::string Name, ClassKind CK) const;
 
   void initVariables();
   std::string replaceParamsIn(std::string S);
@@ -494,7 +495,7 @@ private:
 class NeonEmitter {
   RecordKeeper &Records;
   DenseMap<Record *, ClassKind> ClassMap;
-  std::map<std::string, std::vector<Intrinsic *>> IntrinsicMap;
+  std::map<std::string, std::deque<Intrinsic>> IntrinsicMap;
   unsigned UniqueNumber;
 
   void createIntrinsic(Record *R, SmallVectorImpl<Intrinsic *> &Out);
@@ -507,7 +508,7 @@ class NeonEmitter {
 public:
   /// Called by Intrinsic - this attempts to get an intrinsic that takes
   /// the given types as arguments.
-  Intrinsic *getIntrinsic(StringRef Name, ArrayRef<Type> Types);
+  Intrinsic &getIntrinsic(StringRef Name, ArrayRef<Type> Types);
 
   /// Called by Intrinsic - returns a globally-unique number.
   unsigned getUniqueNumber() { return UniqueNumber++; }
@@ -836,10 +837,6 @@ void Type::applyModifier(char Mod) {
     Float = true;
     break;
   case 'f':
-    // Special case - if we're half-precision, a floating
-    // point argument needs to be 128-bits (double size).
-    if (isHalf())
-      Bitwidth = 128;
     Float = true;
     ElementBitwidth = 32;
     break;
@@ -954,7 +951,7 @@ void Type::applyModifier(char Mod) {
 // Intrinsic implementation
 //===----------------------------------------------------------------------===//
 
-std::string Intrinsic::getInstTypeCode(Type T, ClassKind CK) {
+std::string Intrinsic::getInstTypeCode(Type T, ClassKind CK) const {
   char typeCode = '\0';
   bool printNumber = true;
 
@@ -992,6 +989,10 @@ std::string Intrinsic::getInstTypeCode(Type T, ClassKind CK) {
   return S;
 }
 
+static bool isFloatingPointProtoModifier(char Mod) {
+  return Mod == 'F' || Mod == 'f';
+}
+
 std::string Intrinsic::getBuiltinTypeStr() {
   ClassKind LocalCK = getClassKind(true);
   std::string S;
@@ -1013,7 +1014,7 @@ std::string Intrinsic::getBuiltinTypeStr() {
     if (!RetT.isScalar() && !RetT.isSigned())
       RetT.makeSigned();
 
-    bool ForcedVectorFloatingType = Proto[0] == 'F' || Proto[0] == 'f';
+    bool ForcedVectorFloatingType = isFloatingPointProtoModifier(Proto[0]);
     if (LocalCK == ClassB && !RetT.isScalar() && !ForcedVectorFloatingType)
       // Cast to vector of 8-bit elements.
       RetT.makeInteger(8, true);
@@ -1026,7 +1027,7 @@ std::string Intrinsic::getBuiltinTypeStr() {
     if (T.isPoly())
       T.makeInteger(T.getElementSizeInBits(), false);
 
-    bool ForcedFloatingType = Proto[I + 1] == 'F' || Proto[I + 1] == 'f';
+    bool ForcedFloatingType = isFloatingPointProtoModifier(Proto[I + 1]);
     if (LocalCK == ClassB && !T.isScalar() && !ForcedFloatingType)
       T.makeInteger(8, true);
     // Halves always get converted to 8-bit elements.
@@ -1049,7 +1050,7 @@ std::string Intrinsic::getBuiltinTypeStr() {
   return S;
 }
 
-std::string Intrinsic::getMangledName(bool ForceClassS) {
+std::string Intrinsic::getMangledName(bool ForceClassS) const {
   // Check if the prototype has a scalar operand with the type of the vector
   // elements.  If not, bitcasting the args will take care of arg checking.
   // The actual signedness etc. will be taken care of with special enums.
@@ -1060,12 +1061,12 @@ std::string Intrinsic::getMangledName(bool ForceClassS) {
   return mangleName(Name, ForceClassS ? ClassS : LocalCK);
 }
 
-std::string Intrinsic::mangleName(std::string Name, ClassKind LocalCK) {
+std::string Intrinsic::mangleName(std::string Name, ClassKind LocalCK) const {
   std::string typeCode = getInstTypeCode(BaseType, LocalCK);
   std::string S = Name;
 
-  if (Name == "vcvt_f32_f16" || Name == "vcvt_f32_f64" ||
-      Name == "vcvt_f64_f32")
+  if (Name == "vcvt_f16_f32" || Name == "vcvt_f32_f16" ||
+      Name == "vcvt_f32_f64" || Name == "vcvt_f64_f32")
     return Name;
 
   if (typeCode.size() > 0) {
@@ -1278,7 +1279,7 @@ void Intrinsic::emitShadowedArgs() {
 
 // We don't check 'a' in this function, because for builtin function the
 // argument matching to 'a' uses a vector type splatted from a scalar type.
-bool Intrinsic::protoHasScalar() {
+bool Intrinsic::protoHasScalar() const {
   return (Proto.find('s') != std::string::npos ||
           Proto.find('z') != std::string::npos ||
           Proto.find('r') != std::string::npos ||
@@ -1363,7 +1364,7 @@ void Intrinsic::emitBodyAsBuiltinCall() {
   // Extra constant integer to hold type class enum for this function, e.g. s8
   if (getClassKind(true) == ClassB) {
     Type ThisTy = getReturnType();
-    if (Proto[0] == 'v' || Proto[0] == 'f' || Proto[0] == 'F')
+    if (Proto[0] == 'v' || isFloatingPointProtoModifier(Proto[0]))
       ThisTy = getParamType(0);
     if (ThisTy.isPointer())
       ThisTy = getParamType(1);
@@ -1491,15 +1492,14 @@ std::pair<Type, std::string> Intrinsic::DagEmitter::emitDagCall(DagInit *DI) {
     N = SI->getAsUnquotedString();
   else
     N = emitDagArg(DI->getArg(0), "").second;
-  Intrinsic *Callee = Intr.Emitter.getIntrinsic(N, Types);
-  assert(Callee && "getIntrinsic should not return us nullptr!");
+  Intrinsic &Callee = Intr.Emitter.getIntrinsic(N, Types);
 
   // Make sure the callee is known as an early def.
-  Callee->setNeededEarly();
-  Intr.Dependencies.insert(Callee);
+  Callee.setNeededEarly();
+  Intr.Dependencies.insert(&Callee);
 
   // Now create the call itself.
-  std::string S = CallPrefix.str() + Callee->getMangledName(true) + "(";
+  std::string S = CallPrefix.str() + Callee.getMangledName(true) + "(";
   for (unsigned I = 0; I < DI->getNumArgs() - 1; ++I) {
     if (I != 0)
       S += ", ";
@@ -1507,7 +1507,7 @@ std::pair<Type, std::string> Intrinsic::DagEmitter::emitDagCall(DagInit *DI) {
   }
   S += ")";
 
-  return std::make_pair(Callee->getReturnType(), S);
+  return std::make_pair(Callee.getReturnType(), S);
 }
 
 std::pair<Type, std::string> Intrinsic::DagEmitter::emitDagCast(DagInit *DI,
@@ -1855,11 +1855,11 @@ void Intrinsic::indexBody() {
 // NeonEmitter implementation
 //===----------------------------------------------------------------------===//
 
-Intrinsic *NeonEmitter::getIntrinsic(StringRef Name, ArrayRef<Type> Types) {
+Intrinsic &NeonEmitter::getIntrinsic(StringRef Name, ArrayRef<Type> Types) {
   // First, look up the name in the intrinsic map.
   assert_with_loc(IntrinsicMap.find(Name.str()) != IntrinsicMap.end(),
                   ("Intrinsic '" + Name + "' not found!").str());
-  std::vector<Intrinsic *> &V = IntrinsicMap[Name.str()];
+  auto &V = IntrinsicMap.find(Name.str())->second;
   std::vector<Intrinsic *> GoodVec;
 
   // Create a string to print if we end up failing.
@@ -1874,35 +1874,35 @@ Intrinsic *NeonEmitter::getIntrinsic(StringRef Name, ArrayRef<Type> Types) {
 
   // Now, look through each intrinsic implementation and see if the types are
   // compatible.
-  for (auto *I : V) {
-    ErrMsg += "  - " + I->getReturnType().str() + " " + I->getMangledName();
+  for (auto &I : V) {
+    ErrMsg += "  - " + I.getReturnType().str() + " " + I.getMangledName();
     ErrMsg += "(";
-    for (unsigned A = 0; A < I->getNumParams(); ++A) {
+    for (unsigned A = 0; A < I.getNumParams(); ++A) {
       if (A != 0)
         ErrMsg += ", ";
-      ErrMsg += I->getParamType(A).str();
+      ErrMsg += I.getParamType(A).str();
     }
     ErrMsg += ")\n";
 
-    if (I->getNumParams() != Types.size())
+    if (I.getNumParams() != Types.size())
       continue;
 
     bool Good = true;
     for (unsigned Arg = 0; Arg < Types.size(); ++Arg) {
-      if (I->getParamType(Arg) != Types[Arg]) {
+      if (I.getParamType(Arg) != Types[Arg]) {
         Good = false;
         break;
       }
     }
     if (Good)
-      GoodVec.push_back(I);
+      GoodVec.push_back(&I);
   }
 
   assert_with_loc(GoodVec.size() > 0,
                   "No compatible intrinsic found - " + ErrMsg);
   assert_with_loc(GoodVec.size() == 1, "Multiple overloads found - " + ErrMsg);
 
-  return GoodVec.front();
+  return *GoodVec.front();
 }
 
 void NeonEmitter::createIntrinsic(Record *R,
@@ -1947,13 +1947,12 @@ void NeonEmitter::createIntrinsic(Record *R,
   std::sort(NewTypeSpecs.begin(), NewTypeSpecs.end());
   NewTypeSpecs.erase(std::unique(NewTypeSpecs.begin(), NewTypeSpecs.end()),
 		     NewTypeSpecs.end());
+  auto &Entry = IntrinsicMap[Name];
 
   for (auto &I : NewTypeSpecs) {
-    Intrinsic *IT = new Intrinsic(R, Name, Proto, I.first, I.second, CK, Body,
-                                  *this, Guard, IsUnavailable, BigEndianSafe);
-
-    IntrinsicMap[Name].push_back(IT);
-    Out.push_back(IT);
+    Entry.emplace_back(R, Name, Proto, I.first, I.second, CK, Body, *this,
+                       Guard, IsUnavailable, BigEndianSafe);
+    Out.push_back(&Entry.back());
   }
 
   CurrentRecord = nullptr;
@@ -2023,8 +2022,8 @@ void NeonEmitter::genOverloadTypeCheckCode(raw_ostream &OS,
 
     uint64_t Mask = 0ULL;
     Type Ty = Def->getReturnType();
-    if (Def->getProto()[0] == 'v' || Def->getProto()[0] == 'f' ||
-        Def->getProto()[0] == 'F')
+    if (Def->getProto()[0] == 'v' ||
+        isFloatingPointProtoModifier(Def->getProto()[0]))
       Ty = Def->getParamType(0);
     if (Ty.isPointer())
       Ty = Def->getParamType(1);
diff --git a/utils/analyzer/CmpRuns.py b/utils/analyzer/CmpRuns.py
index ce5ddfb3add4..2d1f44f6880c 100755
--- a/utils/analyzer/CmpRuns.py
+++ b/utils/analyzer/CmpRuns.py
@@ -5,7 +5,7 @@ CmpRuns - A simple tool for comparing two static analyzer runs to determine
 which reports have been added, removed, or changed.
 
 This is designed to support automated testing using the static analyzer, from
-two perspectives: 
+two perspectives:
   1. To monitor changes in the static analyzer's reports on real code bases, for
      regression testing.
 
@@ -19,11 +19,11 @@ Usage:
     #
     resultsA = loadResultsFromSingleRun(singleRunInfoA, deleteEmpty)
     resultsB = loadResultsFromSingleRun(singleRunInfoB, deleteEmpty)
-    
-    # Generate a relation from diagnostics in run A to diagnostics in run B 
-    # to obtain a list of triples (a, b, confidence). 
+
+    # Generate a relation from diagnostics in run A to diagnostics in run B
+    # to obtain a list of triples (a, b, confidence).
     diff = compareResults(resultsA, resultsB)
-           
+
 """
 
 import os
@@ -32,7 +32,7 @@ import CmpRuns
 
 # Information about analysis run:
 # path - the analysis output directory
-# root - the name of the root directory, which will be disregarded when 
+# root - the name of the root directory, which will be disregarded when
 # determining the source file name
 class SingleRunInfo:
     def __init__(self, path, root="", verboseLog=None):
@@ -56,7 +56,7 @@ class AnalysisDiagnostic:
 
     def getLine(self):
         return self._loc['line']
-        
+
     def getColumn(self):
         return self._loc['col']
 
@@ -70,8 +70,8 @@ class AnalysisDiagnostic:
         id = self.getFileName() + "+"
         if 'issue_context' in self._data :
           id += self._data['issue_context'] + "+"
-        if 'issue_hash' in self._data :
-          id += str(self._data['issue_hash'])
+        if 'issue_hash_content_of_line_in_context' in self._data :
+          id += str(self._data['issue_hash_content_of_line_in_context'])
         return id
 
     def getReport(self):
@@ -80,12 +80,12 @@ class AnalysisDiagnostic:
         return os.path.join(self._report.run.path, self._htmlReport)
 
     def getReadableName(self):
-        return '%s:%d:%d, %s: %s' % (self.getFileName(), self.getLine(), 
-                                     self.getColumn(), self.getCategory(), 
+        return '%s:%d:%d, %s: %s' % (self.getFileName(), self.getLine(),
+                                     self.getColumn(), self.getCategory(),
                                      self.getDescription())
-        
-    # Note, the data format is not an API and may change from one analyzer 
-    # version to another.        
+
+    # Note, the data format is not an API and may change from one analyzer
+    # version to another.
     def getRawData(self):
         return self._data
 
@@ -94,7 +94,7 @@ class multidict:
         self.data = {}
         for key,value in elts:
             self[key] = value
-    
+
     def __getitem__(self, item):
         return self.data[item]
     def __setitem__(self, key, value):
@@ -134,15 +134,15 @@ class AnalysisRun:
         # Cumulative list of all diagnostics from all the reports.
         self.diagnostics = []
         self.clang_version = None
-    
+
     def getClangVersion(self):
         return self.clang_version
 
     def readSingleFile(self, p, deleteEmpty):
         data = plistlib.readPlist(p)
 
-        # We want to retrieve the clang version even if there are no 
-        # reports. Assume that all reports were created using the same 
+        # We want to retrieve the clang version even if there are no
+        # reports. Assume that all reports were created using the same
         # clang version (this is always true and is more efficient).
         if 'clang_version' in data:
             if self.clang_version == None:
@@ -166,9 +166,9 @@ class AnalysisRun:
                 htmlFiles.append(d.pop('HTMLDiagnostics_files')[0])
         else:
             htmlFiles = [None] * len(data['diagnostics'])
-            
+
         report = AnalysisReport(self, data.pop('files'))
-        diagnostics = [AnalysisDiagnostic(d, report, h) 
+        diagnostics = [AnalysisDiagnostic(d, report, h)
                        for d,h in zip(data.pop('diagnostics'),
                                       htmlFiles)]
 
@@ -179,7 +179,7 @@ class AnalysisRun:
         self.diagnostics.extend(diagnostics)
 
 
-# Backward compatibility API. 
+# Backward compatibility API.
 def loadResults(path, opts, root = "", deleteEmpty=True):
     return loadResultsFromSingleRun(SingleRunInfo(path, root, opts.verboseLog),
                                     deleteEmpty)
@@ -257,7 +257,7 @@ def dumpScanBuildResultsDiff(dirA, dirB, opts, deleteEmpty=True):
     # Load the run results.
     resultsA = loadResults(dirA, opts, opts.rootA, deleteEmpty)
     resultsB = loadResults(dirB, opts, opts.rootB, deleteEmpty)
-    
+
     # Open the verbose log, if given.
     if opts.verboseLog:
         auxLog = open(opts.verboseLog, "wb")
@@ -285,7 +285,7 @@ def dumpScanBuildResultsDiff(dirA, dirB, opts, deleteEmpty=True):
                                          b.getReadableName())
             foundDiffs += 1
             if auxLog:
-                print >>auxLog, ("('CHANGED', %r, %r, %r, %r)" 
+                print >>auxLog, ("('CHANGED', %r, %r, %r, %r)"
                                  % (a.getReadableName(),
                                     b.getReadableName(),
                                     a.getReport(),
@@ -299,7 +299,7 @@ def dumpScanBuildResultsDiff(dirA, dirB, opts, deleteEmpty=True):
     if auxLog:
         print >>auxLog, "('TOTAL NEW REPORTS', %r)" % TotalReports
         print >>auxLog, "('TOTAL DIFFERENCES', %r)" % foundDiffs
-        
+
     return foundDiffs, len(resultsA.diagnostics), len(resultsB.diagnostics)
 
 def main():
@@ -322,7 +322,7 @@ def main():
 
     dirA,dirB = args
 
-    dumpScanBuildResultsDiff(dirA, dirB, opts)    
+    dumpScanBuildResultsDiff(dirA, dirB, opts)
 
 if __name__ == '__main__':
     main()
diff --git a/utils/analyzer/SATestAdd.py b/utils/analyzer/SATestAdd.py
index 64ff4ff65683..4b94a109ce64 100644
--- a/utils/analyzer/SATestAdd.py
+++ b/utils/analyzer/SATestAdd.py
@@ -1,20 +1,44 @@
 #!/usr/bin/env python
 
 """
-Static Analyzer qualification infrastructure: adding a new project to 
+Static Analyzer qualification infrastructure: adding a new project to
 the Repository Directory.
 
  Add a new project for testing: build it and add to the Project Map file.
    Assumes it's being run from the Repository Directory.
-   The project directory should be added inside the Repository Directory and 
+   The project directory should be added inside the Repository Directory and
    have the same name as the project ID
-   
+
  The project should use the following files for set up:
-      - pre_run_static_analyzer.sh - prepare the build environment.
+      - cleanup_run_static_analyzer.sh - prepare the build environment.
                                      Ex: make clean can be a part of it.
       - run_static_analyzer.cmd - a list of commands to run through scan-build.
                                      Each command should be on a separate line.
-                                     Choose from: configure, make, xcodebuild 
+                                     Choose from: configure, make, xcodebuild
+      - download_project.sh - download the project into the CachedSource/
+                                     directory. For example, download a zip of
+                                     the project source from GitHub, unzip it,
+                                     and rename the unzipped directory to
+                                     'CachedSource'. This script is not called
+                                     when 'CachedSource' is already present,
+                                     so an alternative is to check the
+                                     'CachedSource' directory into the
+                                     repository directly.
+      - CachedSource/ - An optional directory containing the source of the
+                                     project being analyzed. If present,
+                                     download_project.sh will not be called.
+      - changes_for_analyzer.patch - An optional patch file for any local changes
+                                     (e.g., to adapt to newer version of clang)
+                                     that should be applied to CachedSource
+                                     before analysis. To construct this patch,
+                                     run the the download script to download
+                                     the project to CachedSource, copy the
+                                     CachedSource to another directory (for
+                                     example, PatchedSource) and make any needed
+                                     modifications to the the copied source.
+                                     Then run:
+                                          diff -ur CachedSource PatchedSource \
+                                              > changes_for_analyzer.patch
 """
 import SATestBuild
 
@@ -27,7 +51,7 @@ def isExistingProject(PMapFile, projectID) :
     for I in PMapReader:
         if projectID == I[0]:
             return True
-    return False    
+    return False
 
 # Add a new project for testing: build it and add to the Project Map file.
 # Params:
@@ -39,7 +63,7 @@ def addNewProject(ID, BuildMode) :
     if not os.path.exists(Dir):
         print "Error: Project directory is missing: %s" % Dir
         sys.exit(-1)
-        
+
     # Build the project.
     SATestBuild.testProject(ID, BuildMode, IsReferenceBuild=True, Dir=Dir)
 
@@ -51,19 +75,19 @@ def addNewProject(ID, BuildMode) :
         print "Warning: Creating the Project Map file!!"
         PMapFile = open(ProjectMapPath, "w+b")
     try:
-        if (isExistingProject(PMapFile, ID)) :        
+        if (isExistingProject(PMapFile, ID)) :
             print >> sys.stdout, 'Warning: Project with ID \'', ID, \
                                  '\' already exists.'
             print >> sys.stdout, "Reference output has been regenerated."
-        else:                     
+        else:
             PMapWriter = csv.writer(PMapFile)
             PMapWriter.writerow( (ID, int(BuildMode)) );
             print "The project map is updated: ", ProjectMapPath
     finally:
         PMapFile.close()
-            
 
-# TODO: Add an option not to build. 
+
+# TODO: Add an option not to build.
 # TODO: Set the path to the Repository directory.
 if __name__ == '__main__':
     if len(sys.argv) < 2:
@@ -73,10 +97,10 @@ if __name__ == '__main__':
                              '1 for scan_build; ' \
                              '2 for single file c++11 project'
         sys.exit(-1)
-    
-    BuildMode = 1    
+
+    BuildMode = 1
     if (len(sys.argv) >= 3):
-        BuildMode = int(sys.argv[2])  
+        BuildMode = int(sys.argv[2])
     assert((BuildMode == 0) | (BuildMode == 1) | (BuildMode == 2))
-        
+
     addNewProject(sys.argv[1], BuildMode)
diff --git a/utils/analyzer/SATestBuild.py b/utils/analyzer/SATestBuild.py
index 2d91ba41e24c..d0503c6389c9 100644
--- a/utils/analyzer/SATestBuild.py
+++ b/utils/analyzer/SATestBuild.py
@@ -6,8 +6,8 @@ Static Analyzer qualification infrastructure.
 The goal is to test the analyzer against different projects, check for failures,
 compare results, and measure performance.
 
-Repository Directory will contain sources of the projects as well as the 
-information on how to build them and the expected output. 
+Repository Directory will contain sources of the projects as well as the
+information on how to build them and the expected output.
 Repository Directory structure:
    - ProjectMap file
    - Historical Performance Data
@@ -19,16 +19,16 @@ Repository Directory structure:
 Note that the build tree must be inside the project dir.
 
 To test the build of the analyzer one would:
-   - Copy over a copy of the Repository Directory. (TODO: Prefer to ensure that 
+   - Copy over a copy of the Repository Directory. (TODO: Prefer to ensure that
      the build directory does not pollute the repository to min network traffic).
    - Build all projects, until error. Produce logs to report errors.
-   - Compare results.  
+   - Compare results.
 
-The files which should be kept around for failure investigations: 
+The files which should be kept around for failure investigations:
    RepositoryCopy/Project DirI/ScanBuildResults
-   RepositoryCopy/Project DirI/run_static_analyzer.log      
-   
-Assumptions (TODO: shouldn't need to assume these.):   
+   RepositoryCopy/Project DirI/run_static_analyzer.log
+
+Assumptions (TODO: shouldn't need to assume these.):
    The script is being run from the Repository Directory.
    The compiler for scan-build and scan-build are in the PATH.
    export PATH=/Users/zaks/workspace/c2llvm/build/Release+Asserts/bin:$PATH
@@ -36,6 +36,10 @@ Assumptions (TODO: shouldn't need to assume these.):
 For more logging, set the  env variables:
    zaks:TI zaks$ export CCC_ANALYZER_LOG=1
    zaks:TI zaks$ export CCC_ANALYZER_VERBOSE=1
+
+The list of checkers tested are hardcoded in the Checkers variable.
+For testing additional checkers, use the SA_ADDITIONAL_CHECKERS environment
+variable. It should contain a comma separated list.
 """
 import CmpRuns
 
@@ -48,7 +52,7 @@ import shutil
 import time
 import plistlib
 import argparse
-from subprocess import check_call, CalledProcessError
+from subprocess import check_call, check_output, CalledProcessError
 
 #------------------------------------------------------------------------------
 # Helper functions.
@@ -116,16 +120,16 @@ class flushfile(object):
 sys.stdout = flushfile(sys.stdout)
 
 def getProjectMapPath():
-    ProjectMapPath = os.path.join(os.path.abspath(os.curdir), 
+    ProjectMapPath = os.path.join(os.path.abspath(os.curdir),
                                   ProjectMapFile)
     if not os.path.exists(ProjectMapPath):
         print "Error: Cannot find the Project Map file " + ProjectMapPath +\
                 "\nRunning script for the wrong directory?"
-        sys.exit(-1)  
-    return ProjectMapPath         
+        sys.exit(-1)
+    return ProjectMapPath
 
 def getProjectDir(ID):
-    return os.path.join(os.path.abspath(os.curdir), ID)        
+    return os.path.join(os.path.abspath(os.curdir), ID)
 
 def getSBOutputDirName(IsReferenceBuild) :
     if IsReferenceBuild == True :
@@ -150,15 +154,17 @@ Jobs = int(math.ceil(detectCPUs() * 0.75))
 ProjectMapFile = "projectMap.csv"
 
 # Names of the project specific scripts.
+# The script that downloads the project.
+DownloadScript = "download_project.sh"
 # The script that needs to be executed before the build can start.
 CleanupScript = "cleanup_run_static_analyzer.sh"
-# This is a file containing commands for scan-build.  
+# This is a file containing commands for scan-build.
 BuildScript = "run_static_analyzer.cmd"
 
 # The log file name.
 LogFolderName = "Logs"
 BuildLogName = "run_static_analyzer.log"
-# Summary file - contains the summary of the failures. Ex: This info can be be  
+# Summary file - contains the summary of the failures. Ex: This info can be be
 # displayed when buildbot detects a build failure.
 NumOfFailuresInSummary = 10
 FailuresSummaryFileName = "failures.txt"
@@ -169,6 +175,21 @@ DiffsSummaryFileName = "diffs.txt"
 SBOutputDirName = "ScanBuildResults"
 SBOutputDirReferencePrefix = "Ref"
 
+# The name of the directory storing the cached project source. If this directory
+# does not exist, the download script will be executed. That script should
+# create the "CachedSource" directory and download the project source into it.
+CachedSourceDirName = "CachedSource"
+
+# The name of the directory containing the source code that will be analyzed.
+# Each time a project is analyzed, a fresh copy of its CachedSource directory
+# will be copied to the PatchedSource directory and then the local patches
+# in PatchfileName will be applied (if PatchfileName exists).
+PatchedSourceDirName = "PatchedSource"
+
+# The name of the patchfile specifying any changes that should be applied
+# to the CachedSource before analyzing.
+PatchfileName = "changes_for_analyzer.patch"
+
 # The list of checkers used during analyzes.
 # Currently, consists of all the non-experimental checkers, plus a few alpha
 # checkers we don't want to regress on.
@@ -182,35 +203,93 @@ Verbose = 1
 
 # Run pre-processing script if any.
 def runCleanupScript(Dir, PBuildLogFile):
+    Cwd = os.path.join(Dir, PatchedSourceDirName)
     ScriptPath = os.path.join(Dir, CleanupScript)
+    runScript(ScriptPath, PBuildLogFile, Cwd)
+
+# Run the script to download the project, if it exists.
+def runDownloadScript(Dir, PBuildLogFile):
+    ScriptPath = os.path.join(Dir, DownloadScript)
+    runScript(ScriptPath, PBuildLogFile, Dir)
+
+# Run the provided script if it exists.
+def runScript(ScriptPath, PBuildLogFile, Cwd):
     if os.path.exists(ScriptPath):
         try:
-            if Verbose == 1:        
+            if Verbose == 1:
                 print "  Executing: %s" % (ScriptPath,)
-            check_call("chmod +x %s" % ScriptPath, cwd = Dir, 
+            check_call("chmod +x %s" % ScriptPath, cwd = Cwd,
                                               stderr=PBuildLogFile,
-                                              stdout=PBuildLogFile, 
-                                              shell=True)    
-            check_call(ScriptPath, cwd = Dir, stderr=PBuildLogFile,
-                                              stdout=PBuildLogFile, 
+                                              stdout=PBuildLogFile,
+                                              shell=True)
+            check_call(ScriptPath, cwd = Cwd, stderr=PBuildLogFile,
+                                              stdout=PBuildLogFile,
                                               shell=True)
         except:
-            print "Error: The pre-processing step failed. See ", \
-                  PBuildLogFile.name, " for details."
+            print "Error: Running %s failed. See %s for details." % (ScriptPath,
+                PBuildLogFile.name)
             sys.exit(-1)
 
-# Build the project with scan-build by reading in the commands and 
+# Download the project and apply the local patchfile if it exists.
+def downloadAndPatch(Dir, PBuildLogFile):
+    CachedSourceDirPath = os.path.join(Dir, CachedSourceDirName)
+
+    # If the we don't already have the cached source, run the project's
+    # download script to download it.
+    if not os.path.exists(CachedSourceDirPath):
+      runDownloadScript(Dir, PBuildLogFile)
+      if not os.path.exists(CachedSourceDirPath):
+        print "Error: '%s' not found after download." % (CachedSourceDirPath)
+        exit(-1)
+
+    PatchedSourceDirPath = os.path.join(Dir, PatchedSourceDirName)
+
+    # Remove potentially stale patched source.
+    if os.path.exists(PatchedSourceDirPath):
+        shutil.rmtree(PatchedSourceDirPath)
+
+    # Copy the cached source and apply any patches to the copy.
+    shutil.copytree(CachedSourceDirPath, PatchedSourceDirPath, symlinks=True)
+    applyPatch(Dir, PBuildLogFile)
+
+def applyPatch(Dir, PBuildLogFile):
+    PatchfilePath = os.path.join(Dir, PatchfileName)
+    PatchedSourceDirPath = os.path.join(Dir, PatchedSourceDirName)
+    if not os.path.exists(PatchfilePath):
+        print "  No local patches."
+        return
+
+    print "  Applying patch."
+    try:
+        check_call("patch -p1 < %s" % (PatchfilePath),
+                    cwd = PatchedSourceDirPath,
+                    stderr=PBuildLogFile,
+                    stdout=PBuildLogFile,
+                    shell=True)
+    except:
+        print "Error: Patch failed. See %s for details." % (PBuildLogFile.name)
+        sys.exit(-1)
+
+# Build the project with scan-build by reading in the commands and
 # prefixing them with the scan-build options.
 def runScanBuild(Dir, SBOutputDir, PBuildLogFile):
     BuildScriptPath = os.path.join(Dir, BuildScript)
     if not os.path.exists(BuildScriptPath):
         print "Error: build script is not defined: %s" % BuildScriptPath
         sys.exit(-1)
+
+    AllCheckers = Checkers
+    if os.environ.has_key('SA_ADDITIONAL_CHECKERS'):
+        AllCheckers = AllCheckers + ',' + os.environ['SA_ADDITIONAL_CHECKERS']
+
+    # Run scan-build from within the patched source directory.
+    SBCwd = os.path.join(Dir, PatchedSourceDirName)
+
     SBOptions = "--use-analyzer " + Clang + " "
     SBOptions += "-plist-html -o " + SBOutputDir + " "
-    SBOptions += "-enable-checker " + Checkers + " "  
+    SBOptions += "-enable-checker " + AllCheckers + " "
     SBOptions += "--keep-empty "
-    # Always use ccc-analyze to ensure that we can locate the failures 
+    # Always use ccc-analyze to ensure that we can locate the failures
     # directory.
     SBOptions += "--override-compiler "
     try:
@@ -227,11 +306,11 @@ def runScanBuild(Dir, SBOutputDir, PBuildLogFile):
                 "-j" not in Command:
                 Command += " -j%d" % Jobs
             SBCommand = SBPrefix + Command
-            if Verbose == 1:        
+            if Verbose == 1:
                 print "  Executing: %s" % (SBCommand,)
-            check_call(SBCommand, cwd = Dir, stderr=PBuildLogFile,
-                                             stdout=PBuildLogFile, 
-                                             shell=True)
+            check_call(SBCommand, cwd = SBCwd, stderr=PBuildLogFile,
+                                               stdout=PBuildLogFile,
+                                               shell=True)
     except:
         print "Error: scan-build failed. See ",PBuildLogFile.name,\
               " for details."
@@ -245,97 +324,114 @@ def hasNoExtension(FileName):
 
 def isValidSingleInputFile(FileName):
     (Root, Ext) = os.path.splitext(FileName)
-    if ((Ext == ".i") | (Ext == ".ii") | 
-        (Ext == ".c") | (Ext == ".cpp") | 
+    if ((Ext == ".i") | (Ext == ".ii") |
+        (Ext == ".c") | (Ext == ".cpp") |
         (Ext == ".m") | (Ext == "")) :
         return True
     return False
-   
+
+# Get the path to the SDK for the given SDK name. Returns None if
+# the path cannot be determined.
+def getSDKPath(SDKName):
+    if which("xcrun") is None:
+        return None
+
+    Cmd = "xcrun --sdk " + SDKName + " --show-sdk-path"
+    return check_output(Cmd, shell=True).rstrip()
+
 # Run analysis on a set of preprocessed files.
 def runAnalyzePreprocessed(Dir, SBOutputDir, Mode):
     if os.path.exists(os.path.join(Dir, BuildScript)):
         print "Error: The preprocessed files project should not contain %s" % \
                BuildScript
-        raise Exception()       
+        raise Exception()
+
+    CmdPrefix = Clang + " -cc1 "
+
+    # For now, we assume the preprocessed files should be analyzed
+    # with the OS X SDK.
+    SDKPath = getSDKPath("macosx")
+    if SDKPath is not None:
+      CmdPrefix += "-isysroot " + SDKPath + " "
+
+    CmdPrefix += "-analyze -analyzer-output=plist -w "
+    CmdPrefix += "-analyzer-checker=" + Checkers +" -fcxx-exceptions -fblocks "
 
-    CmdPrefix = Clang + " -cc1 -analyze -analyzer-output=plist -w "
-    CmdPrefix += "-analyzer-checker=" + Checkers +" -fcxx-exceptions -fblocks "   
-    
     if (Mode == 2) :
-        CmdPrefix += "-std=c++11 " 
-    
+        CmdPrefix += "-std=c++11 "
+
     PlistPath = os.path.join(Dir, SBOutputDir, "date")
     FailPath = os.path.join(PlistPath, "failures");
     os.makedirs(FailPath);
- 
+
     for FullFileName in glob.glob(Dir + "/*"):
         FileName = os.path.basename(FullFileName)
         Failed = False
-        
+
         # Only run the analyzes on supported files.
         if (hasNoExtension(FileName)):
             continue
         if (isValidSingleInputFile(FileName) == False):
             print "Error: Invalid single input file %s." % (FullFileName,)
             raise Exception()
-        
+
         # Build and call the analyzer command.
         OutputOption = "-o " + os.path.join(PlistPath, FileName) + ".plist "
         Command = CmdPrefix + OutputOption + FileName
         LogFile = open(os.path.join(FailPath, FileName + ".stderr.txt"), "w+b")
         try:
-            if Verbose == 1:        
+            if Verbose == 1:
                 print "  Executing: %s" % (Command,)
             check_call(Command, cwd = Dir, stderr=LogFile,
-                                           stdout=LogFile, 
+                                           stdout=LogFile,
                                            shell=True)
         except CalledProcessError, e:
             print "Error: Analyzes of %s failed. See %s for details." \
                   "Error code %d." % \
                    (FullFileName, LogFile.name, e.returncode)
-            Failed = True       
+            Failed = True
         finally:
-            LogFile.close()            
-        
+            LogFile.close()
+
         # If command did not fail, erase the log file.
         if Failed == False:
             os.remove(LogFile.name);
 
 def buildProject(Dir, SBOutputDir, ProjectBuildMode, IsReferenceBuild):
-    TBegin = time.time() 
+    TBegin = time.time()
 
     BuildLogPath = os.path.join(SBOutputDir, LogFolderName, BuildLogName)
-    print "Log file: %s" % (BuildLogPath,) 
+    print "Log file: %s" % (BuildLogPath,)
     print "Output directory: %s" %(SBOutputDir, )
-    
+
     # Clean up the log file.
     if (os.path.exists(BuildLogPath)) :
         RmCommand = "rm " + BuildLogPath
         if Verbose == 1:
             print "  Executing: %s" % (RmCommand,)
         check_call(RmCommand, shell=True)
-    
+
     # Clean up scan build results.
     if (os.path.exists(SBOutputDir)) :
         RmCommand = "rm -r " + SBOutputDir
-        if Verbose == 1: 
+        if Verbose == 1:
             print "  Executing: %s" % (RmCommand,)
             check_call(RmCommand, shell=True)
     assert(not os.path.exists(SBOutputDir))
     os.makedirs(os.path.join(SBOutputDir, LogFolderName))
-        
+
     # Open the log file.
     PBuildLogFile = open(BuildLogPath, "wb+")
-    
+
     # Build and analyze the project.
     try:
-        runCleanupScript(Dir, PBuildLogFile)
-        
         if (ProjectBuildMode == 1):
+            downloadAndPatch(Dir, PBuildLogFile)
+            runCleanupScript(Dir, PBuildLogFile)
             runScanBuild(Dir, SBOutputDir, PBuildLogFile)
         else:
             runAnalyzePreprocessed(Dir, SBOutputDir, ProjectBuildMode)
-        
+
         if IsReferenceBuild :
             runCleanupScript(Dir, PBuildLogFile)
 
@@ -346,32 +442,36 @@ def buildProject(Dir, SBOutputDir, ProjectBuildMode, IsReferenceBuild):
                         continue
                     Plist = os.path.join(DirPath, F)
                     Data = plistlib.readPlist(Plist)
-                    Paths = [SourceFile[len(Dir)+1:] if SourceFile.startswith(Dir)\
-                            else SourceFile for SourceFile in Data['files']]
+                    PathPrefix = Dir
+                    if (ProjectBuildMode == 1):
+                        PathPrefix = os.path.join(Dir, PatchedSourceDirName)
+                    Paths = [SourceFile[len(PathPrefix)+1:]\
+                              if SourceFile.startswith(PathPrefix)\
+                              else SourceFile for SourceFile in Data['files']]
                     Data['files'] = Paths
                     plistlib.writePlist(Data, Plist)
-           
+
     finally:
         PBuildLogFile.close()
-        
+
     print "Build complete (time: %.2f). See the log for more details: %s" % \
-           ((time.time()-TBegin), BuildLogPath) 
-       
+           ((time.time()-TBegin), BuildLogPath)
+
 # A plist file is created for each call to the analyzer(each source file).
-# We are only interested on the once that have bug reports, so delete the rest.        
+# We are only interested on the once that have bug reports, so delete the rest.
 def CleanUpEmptyPlists(SBOutputDir):
     for F in glob.glob(SBOutputDir + "/*/*.plist"):
         P = os.path.join(SBOutputDir, F)
-        
+
         Data = plistlib.readPlist(P)
         # Delete empty reports.
         if not Data['files']:
             os.remove(P)
             continue
 
-# Given the scan-build output directory, checks if the build failed 
-# (by searching for the failures directories). If there are failures, it 
-# creates a summary file in the output directory.         
+# Given the scan-build output directory, checks if the build failed
+# (by searching for the failures directories). If there are failures, it
+# creates a summary file in the output directory.
 def checkBuild(SBOutputDir):
     # Check if there are failures.
     Failures = glob.glob(SBOutputDir + "/*/failures/*.stderr.txt")
@@ -382,37 +482,37 @@ def checkBuild(SBOutputDir):
         print "Number of bug reports (non-empty plist files) produced: %d" %\
            len(Plists)
         return;
-    
+
     # Create summary file to display when the build fails.
     SummaryPath = os.path.join(SBOutputDir, LogFolderName, FailuresSummaryFileName)
     if (Verbose > 0):
         print "  Creating the failures summary file %s" % (SummaryPath,)
-    
+
     SummaryLog = open(SummaryPath, "w+")
     try:
         SummaryLog.write("Total of %d failures discovered.\n" % (TotalFailed,))
         if TotalFailed > NumOfFailuresInSummary:
-            SummaryLog.write("See the first %d below.\n" 
+            SummaryLog.write("See the first %d below.\n"
                                                    % (NumOfFailuresInSummary,))
         # TODO: Add a line "See the results folder for more."
-    
+
         FailuresCopied = NumOfFailuresInSummary
         Idx = 0
         for FailLogPathI in Failures:
             if Idx >= NumOfFailuresInSummary:
                 break;
-            Idx += 1 
+            Idx += 1
             SummaryLog.write("\n-- Error #%d -----------\n" % (Idx,));
             FailLogI = open(FailLogPathI, "r");
-            try: 
+            try:
                 shutil.copyfileobj(FailLogI, SummaryLog);
             finally:
                 FailLogI.close()
     finally:
         SummaryLog.close()
-    
+
     print "Error: analysis failed. See ", SummaryPath
-    sys.exit(-1)       
+    sys.exit(-1)
 
 # Auxiliary object to discard stdout.
 class Discarder(object):
@@ -424,46 +524,47 @@ class Discarder(object):
 #   0 - success if there are no crashes or analyzer failure.
 #   1 - success if there are no difference in the number of reported bugs.
 #   2 - success if all the bug reports are identical.
-def runCmpResults(Dir, Strictness = 0):   
-    TBegin = time.time() 
+def runCmpResults(Dir, Strictness = 0):
+    TBegin = time.time()
 
     RefDir = os.path.join(Dir, SBOutputDirReferencePrefix + SBOutputDirName)
     NewDir = os.path.join(Dir, SBOutputDirName)
-    
+
     # We have to go one level down the directory tree.
-    RefList = glob.glob(RefDir + "/*") 
+    RefList = glob.glob(RefDir + "/*")
     NewList = glob.glob(NewDir + "/*")
-    
+
     # Log folders are also located in the results dir, so ignore them.
     RefLogDir = os.path.join(RefDir, LogFolderName)
     if RefLogDir in RefList:
         RefList.remove(RefLogDir)
     NewList.remove(os.path.join(NewDir, LogFolderName))
-    
+
     if len(RefList) == 0 or len(NewList) == 0:
         return False
     assert(len(RefList) == len(NewList))
 
-    # There might be more then one folder underneath - one per each scan-build 
+    # There might be more then one folder underneath - one per each scan-build
     # command (Ex: one for configure and one for make).
     if (len(RefList) > 1):
         # Assume that the corresponding folders have the same names.
         RefList.sort()
         NewList.sort()
-    
+
     # Iterate and find the differences.
     NumDiffs = 0
-    PairList = zip(RefList, NewList)    
-    for P in PairList:    
-        RefDir = P[0] 
+    PairList = zip(RefList, NewList)
+    for P in PairList:
+        RefDir = P[0]
         NewDir = P[1]
-    
-        assert(RefDir != NewDir) 
-        if Verbose == 1:        
+
+        assert(RefDir != NewDir)
+        if Verbose == 1:
             print "  Comparing Results: %s %s" % (RefDir, NewDir)
-    
+
         DiffsPath = os.path.join(NewDir, DiffsSummaryFileName)
-        Opts = CmpRuns.CmpOptions(DiffsPath, "", Dir)
+        PatchedSourceDirPath = os.path.join(Dir, PatchedSourceDirName)
+        Opts = CmpRuns.CmpOptions(DiffsPath, "", PatchedSourceDirPath)
         # Discard everything coming out of stdout (CmpRun produces a lot of them).
         OLD_STDOUT = sys.stdout
         sys.stdout = Discarder()
@@ -476,70 +577,70 @@ def runCmpResults(Dir, Strictness = 0):
                   (NumDiffs, DiffsPath,)
         if Strictness >= 2 and NumDiffs > 0:
             print "Error: Diffs found in strict mode (2)."
-            sys.exit(-1)       
+            sys.exit(-1)
         elif Strictness >= 1 and ReportsInRef != ReportsInNew:
             print "Error: The number of results are different in strict mode (1)."
-            sys.exit(-1)       
-                    
-    print "Diagnostic comparison complete (time: %.2f)." % (time.time()-TBegin) 
+            sys.exit(-1)
+
+    print "Diagnostic comparison complete (time: %.2f)." % (time.time()-TBegin)
     return (NumDiffs > 0)
-    
+
 def updateSVN(Mode, ProjectsMap):
     try:
-        ProjectsMap.seek(0)    
+        ProjectsMap.seek(0)
         for I in csv.reader(ProjectsMap):
-            ProjName = I[0] 
+            ProjName = I[0]
             Path = os.path.join(ProjName, getSBOutputDirName(True))
-    
+
             if Mode == "delete":
                 Command = "svn delete %s" % (Path,)
             else:
                 Command = "svn add %s" % (Path,)
 
-            if Verbose == 1:        
+            if Verbose == 1:
                 print "  Executing: %s" % (Command,)
-            check_call(Command, shell=True)    
-    
+            check_call(Command, shell=True)
+
         if Mode == "delete":
             CommitCommand = "svn commit -m \"[analyzer tests] Remove " \
-                            "reference results.\""     
+                            "reference results.\""
         else:
             CommitCommand = "svn commit -m \"[analyzer tests] Add new " \
                             "reference results.\""
-        if Verbose == 1:        
+        if Verbose == 1:
             print "  Executing: %s" % (CommitCommand,)
-        check_call(CommitCommand, shell=True)    
+        check_call(CommitCommand, shell=True)
     except:
         print "Error: SVN update failed."
         sys.exit(-1)
-        
+
 def testProject(ID, ProjectBuildMode, IsReferenceBuild=False, Dir=None, Strictness = 0):
     print " \n\n--- Building project %s" % (ID,)
 
-    TBegin = time.time() 
+    TBegin = time.time()
 
     if Dir is None :
-        Dir = getProjectDir(ID)        
-    if Verbose == 1:        
+        Dir = getProjectDir(ID)
+    if Verbose == 1:
         print "  Build directory: %s." % (Dir,)
-    
+
     # Set the build results directory.
     RelOutputDir = getSBOutputDirName(IsReferenceBuild)
     SBOutputDir = os.path.join(Dir, RelOutputDir)
-                
+
     buildProject(Dir, SBOutputDir, ProjectBuildMode, IsReferenceBuild)
 
     checkBuild(SBOutputDir)
-    
+
     if IsReferenceBuild == False:
         runCmpResults(Dir, Strictness)
-        
+
     print "Completed tests for project %s (time: %.2f)." % \
           (ID, (time.time()-TBegin))
-    
+
 def testAll(IsReferenceBuild = False, UpdateSVN = False, Strictness = 0):
     PMapFile = open(getProjectMapPath(), "rb")
-    try:        
+    try:
         # Validate the input.
         for I in csv.reader(PMapFile):
             if (len(I) != 2) :
@@ -548,16 +649,16 @@ def testAll(IsReferenceBuild = False, UpdateSVN = False, Strictness = 0):
             if (not ((I[1] == "0") | (I[1] == "1") | (I[1] == "2"))):
                 print "Error: Second entry in the ProjectMapFile should be 0" \
                       " (single file), 1 (project), or 2(single file c++11)."
-                raise Exception()              
+                raise Exception()
 
-        # When we are regenerating the reference results, we might need to 
+        # When we are regenerating the reference results, we might need to
         # update svn. Remove reference results from SVN.
         if UpdateSVN == True:
             assert(IsReferenceBuild == True);
             updateSVN("delete",  PMapFile);
-            
+
         # Test the projects.
-        PMapFile.seek(0)    
+        PMapFile.seek(0)
         for I in csv.reader(PMapFile):
             testProject(I[0], int(I[1]), IsReferenceBuild, None, Strictness)
 
@@ -567,10 +668,10 @@ def testAll(IsReferenceBuild = False, UpdateSVN = False, Strictness = 0):
 
     except:
         print "Error occurred. Premature termination."
-        raise                            
+        raise
     finally:
-        PMapFile.close()    
-            
+        PMapFile.close()
+
 if __name__ == '__main__':
     # Parse command line arguments.
     Parser = argparse.ArgumentParser(description='Test the Clang Static Analyzer.')
diff --git a/utils/analyzer/SumTimerInfo.py b/utils/analyzer/SumTimerInfo.py
index 4ef1ceb4cec5..0c3585bbc279 100644
--- a/utils/analyzer/SumTimerInfo.py
+++ b/utils/analyzer/SumTimerInfo.py
@@ -3,7 +3,7 @@
 """
 Script to Summarize statistics in the scan-build output.
 
-Statistics are enabled by passing '-internal-stats' option to scan-build 
+Statistics are enabled by passing '-internal-stats' option to scan-build
 (or '-analyzer-stats' to the analyzer).
 
 """
@@ -69,7 +69,7 @@ if __name__ == '__main__':
         if ((")  Total" in line) and (Mode == 1)) :
           s = line.split()
           TotalTime = TotalTime + float(s[6])
-          
+
     print "TU Count %d" % (Count)
     print "Time %f" % (Time)
     print "Warnings %d" % (Warnings)
@@ -81,4 +81,3 @@ if __name__ == '__main__':
     print "MaxTime %f" % (MaxTime)
     print "TotalTime %f" % (TotalTime)
     print "Max CFG Size %d" % (MaxCFGSize)
-    
\ No newline at end of file
diff --git a/utils/analyzer/ubiviz b/utils/analyzer/ubiviz
index 1582797c63f9..9d821c3c10a0 100755
--- a/utils/analyzer/ubiviz
+++ b/utils/analyzer/ubiviz
@@ -18,7 +18,7 @@ import sys
 def Error(message):
     print >> sys.stderr, 'ubiviz: ' + message
     sys.exit(1)
-    
+
 def StreamData(filename):
   file = open(filename)
   for ln in file:
@@ -55,20 +55,19 @@ def Display(G, data):
 
 def main(args):
   if len(args) == 0:
-    Error('no input files')    
+    Error('no input files')
 
   server = xmlrpclib.Server('http://127.0.0.1:20738/RPC2')
   G = server.ubigraph
-            
+
   for arg in args:
     G.clear()
     for x in StreamData(arg):
       Display(G,x)
-  
+
   sys.exit(0)
-  
+
 
 if __name__ == '__main__':
     main(sys.argv[1:])
-    
-    
\ No newline at end of file
+
diff --git a/utils/clang.natvis b/utils/clang.natvis
index 107af502f7d0..a0004e9b0163 100644
--- a/utils/clang.natvis
+++ b/utils/clang.natvis
@@ -6,6 +6,14 @@ Put this file into "%USERPROFILE%\Documents\Visual Studio 2012\Visualizers"
 or create a symbolic link so it updates automatically.

 -->

 <AutoVisualizer xmlns="http://schemas.microsoft.com/vstudio/debugger/natvis/2010">

+  <Type Name="clang::Type">

+    <DisplayString Condition="(clang::Type::TypeClass)TypeBits.TC == clang::Type::Builtin">Builtin Type={(clang::BuiltinType::Kind)BuiltinTypeBits.Kind}</DisplayString>

+    <DisplayString Condition="(clang::Type::TypeClass)TypeBits.TC == clang::Type::Attributed">Modified Type={((clang::AttributedType*)this)->ModifiedType} Attribute={(clang::AttributedType::Kind)AttributedTypeBits.AttrKind}</DisplayString>

+    <DisplayString>Type Class={(clang::Type::TypeClass)TypeBits.TC}</DisplayString>

+  </Type>

+  <Type Name="clang::QualType">

+    <DisplayString>{((clang::ExtQualsTypeCommonBase *)(((uintptr_t)Value.Value) &amp; ~(uintptr_t)((1 &lt;&lt; clang::TypeAlignmentInBits) - 1)))-&gt;BaseType}</DisplayString>

+  </Type>

   <Type Name="clang::IdentifierInfo">

     <DisplayString Condition="Entry != 0">({((llvm::StringMapEntry&lt;clang::IdentifierInfo *&gt;*)Entry)+1,s})</DisplayString>

     <Expand>

@@ -24,13 +32,14 @@ or create a symbolic link so it updates automatically.
       <Item Condition="(Ptr &amp; PtrMask) == StoredObjCZeroArgSelector" Name="[ObjC Zero Arg Selector]">*(clang::IdentifierInfo *)(Ptr &amp; ~PtrMask)</Item>

       <Item Condition="(Ptr &amp; PtrMask) == StoredObjCOneArgSelector" Name="[ObjC One Arg Selector]">*(clang::IdentifierInfo *)(Ptr &amp; ~PtrMask)</Item>

       <Item Condition="(Ptr &amp; PtrMask) == StoredDeclarationNameExtra" Name="[Extra]">(clang::DeclarationNameExtra::ExtraKind)((clang::DeclarationNameExtra *)(Ptr &amp; ~PtrMask))-&gt;ExtraKindOrNumArgs</Item>

-    </Expand>

-  </Type>

-  <Type Name="clang::Token">

-    <DisplayString>{(clang::tok::TokenKind)Kind}</DisplayString>

-  </Type>

-  <Type Name="clang::DeclSpec">

-    <DisplayString>[{(clang::DeclSpec::SCS)StorageClassSpec}], [{(clang::TypeSpecifierType)TypeSpecType}]</DisplayString>

+    </Expand>
+  </Type>
+  <Type Name="clang::Token">
+    <DisplayString Condition="Kind != clang::tok::identifier">{(clang::tok::TokenKind)Kind}</DisplayString>
+    <DisplayString Condition="Kind == clang::tok::identifier">{{Identifier ({*(clang::IdentifierInfo *)(PtrData)})}}</DisplayString>
+  </Type>
+  <Type Name="clang::DeclSpec">
+    <DisplayString>[{(clang::DeclSpec::SCS)StorageClassSpec}], [{(clang::TypeSpecifierType)TypeSpecType}]</DisplayString>
   </Type>

   <Type Name="clang::PragmaHandler">

     <DisplayString>{Name,s}</DisplayString>

diff --git a/utils/perf-training/CMakeLists.txt b/utils/perf-training/CMakeLists.txt
new file mode 100644
index 000000000000..ccedcf5d5163
--- /dev/null
+++ b/utils/perf-training/CMakeLists.txt
@@ -0,0 +1,36 @@
+if(LLVM_BUILD_INSTRUMENTED)
+  if (CMAKE_CFG_INTDIR STREQUAL ".")
+    set(LLVM_BUILD_MODE ".")
+  else ()
+    set(LLVM_BUILD_MODE "%(build_mode)s")
+  endif ()
+
+  string(REPLACE ${CMAKE_CFG_INTDIR} ${LLVM_BUILD_MODE} CLANG_TOOLS_DIR ${LLVM_RUNTIME_OUTPUT_INTDIR})
+
+  configure_lit_site_cfg(
+    ${CMAKE_CURRENT_SOURCE_DIR}/lit.site.cfg.in
+    ${CMAKE_CURRENT_BINARY_DIR}/lit.site.cfg
+    )
+
+  add_lit_testsuite(generate-profraw "Generating clang PGO data"
+    ${CMAKE_CURRENT_BINARY_DIR}
+    DEPENDS clang clear-profraw
+    )
+
+  add_custom_target(clear-profraw
+    COMMAND ${PYTHON_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/perf-helper.py clean ${CMAKE_CURRENT_BINARY_DIR}
+    COMMENT "Clearing old profraw data")
+
+  if(NOT LLVM_PROFDATA)
+    find_program(LLVM_PROFDATA llvm-profdata)
+  endif()
+
+  if(NOT LLVM_PROFDATA)
+    message(FATAL_ERROR "Must set LLVM_PROFDATA to point to llvm-profdata to use for merging PGO data")
+  endif()
+
+  add_custom_target(generate-profdata
+    COMMAND ${PYTHON_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/perf-helper.py merge ${LLVM_PROFDATA} ${CMAKE_CURRENT_BINARY_DIR}/clang.profdata ${CMAKE_CURRENT_BINARY_DIR}
+    COMMENT "Merging profdata"
+    DEPENDS generate-profraw)
+endif()
diff --git a/utils/perf-training/README.txt b/utils/perf-training/README.txt
new file mode 100644
index 000000000000..cfbce40ca3ef
--- /dev/null
+++ b/utils/perf-training/README.txt
@@ -0,0 +1,6 @@
+==========================
+ Performance Training Data
+==========================
+
+This directory contains simple source files for use as training data for
+generating PGO data and linker order files for clang.
diff --git a/utils/perf-training/cxx/hello_world.cpp b/utils/perf-training/cxx/hello_world.cpp
new file mode 100644
index 000000000000..66e00d00d261
--- /dev/null
+++ b/utils/perf-training/cxx/hello_world.cpp
@@ -0,0 +1,7 @@
+// RUN: %clang_cpp -c %s
+#include <iostream>
+
+int main(int, char**) {
+  std::cout << "Hello, World!";
+  return 0;
+}
diff --git a/utils/perf-training/lit.cfg b/utils/perf-training/lit.cfg
new file mode 100644
index 000000000000..af4b43b78b09
--- /dev/null
+++ b/utils/perf-training/lit.cfg
@@ -0,0 +1,37 @@
+# -*- Python -*-
+
+from lit import Test
+import lit.formats
+import lit.util
+
+def getSysrootFlagsOnDarwin(config, lit_config):
+    # On Darwin, support relocatable SDKs by providing Clang with a
+    # default system root path.
+    if 'darwin' in config.target_triple:
+        try:
+            out = lit.util.capture(['xcrun', '--show-sdk-path']).strip()
+            res = 0
+        except OSError:
+            res = -1
+        if res == 0 and out:
+            sdk_path = out
+            lit_config.note('using SDKROOT: %r' % sdk_path)
+            return '-isysroot %s' % sdk_path
+    return ''
+
+sysroot_flags = getSysrootFlagsOnDarwin(config, lit_config)
+
+config.clang = lit.util.which('clang', config.clang_tools_dir).replace('\\', '/')
+
+config.name = 'Clang Perf Training'
+config.suffixes = ['.c', '.cpp', '.m', '.mm', '.cu', '.ll', '.cl', '.s', '.S', '.modulemap']
+
+use_lit_shell = os.environ.get("LIT_USE_INTERNAL_SHELL")
+config.test_format = lit.formats.ShTest(use_lit_shell == "0")
+config.substitutions.append( ('%clang_cpp', ' %s --driver-mode=cpp %s ' % (config.clang, sysroot_flags)))
+config.substitutions.append( ('%clang_cc1', ' %s -cc1 %s ' % (config.clang, sysroot_flags)))
+config.substitutions.append( ('%clang', ' %s %s ' % (config.clang, sysroot_flags) ) )
+config.substitutions.append( ('%test_root', config.test_exec_root ) )
+
+config.environment['LLVM_PROFILE_FILE'] = 'perf-training-%p.profraw'
+
diff --git a/utils/perf-training/lit.site.cfg.in b/utils/perf-training/lit.site.cfg.in
new file mode 100644
index 000000000000..9dc380242e52
--- /dev/null
+++ b/utils/perf-training/lit.site.cfg.in
@@ -0,0 +1,20 @@
+import sys
+
+## Autogenerated by LLVM/Clang configuration.
+# Do not edit!
+config.clang_tools_dir = "@CLANG_TOOLS_DIR@"
+config.test_exec_root = "@CMAKE_CURRENT_BINARY_DIR@"
+config.test_source_root = "@CMAKE_CURRENT_SOURCE_DIR@"
+config.target_triple = "@TARGET_TRIPLE@"
+
+# Support substitution of the tools and libs dirs with user parameters. This is
+# used when we can't determine the tool dir at configuration time.
+try:
+    config.clang_tools_dir = config.clang_tools_dir % lit_config.params
+except KeyError:
+    e = sys.exc_info()[1]
+    key, = e.args
+    lit_config.fatal("unable to find %r parameter, use '--param=%s=VALUE'" % (key,key))
+
+# Let the main config do the real work.
+lit_config.load_config(config, "@CLANG_SOURCE_DIR@/utils/perf-training/lit.cfg")
diff --git a/utils/perf-training/perf-helper.py b/utils/perf-training/perf-helper.py
new file mode 100644
index 000000000000..448801133e0a
--- /dev/null
+++ b/utils/perf-training/perf-helper.py
@@ -0,0 +1,46 @@
+#===- perf-helper.py - Clang Python Bindings -----------------*- python -*--===#
+#
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+import sys
+import os
+import subprocess
+
+def findProfrawFiles(path):
+  profraw_files = []
+  for root, dirs, files in os.walk(path): 
+    for filename in files:
+      if filename.endswith(".profraw"):
+        profraw_files.append(os.path.join(root, filename))
+  return profraw_files
+
+def clean(args):
+  if len(args) != 1:
+    print 'Usage: %s clean <path>\n\tRemoves all *.profraw files from <path>.' % __file__
+    return 1
+  for profraw in findProfrawFiles(args[0]):
+    os.remove(profraw)
+  return 0
+
+def merge(args):
+  if len(args) != 3:
+    print 'Usage: %s clean <llvm-profdata> <output> <path>\n\tMerges all profraw files from path into output.' % __file__
+    return 1
+  cmd = [args[0], 'merge', '-o', args[1]]
+  cmd.extend(findProfrawFiles(args[2]))
+  subprocess.check_call(cmd)
+  return 0
+
+commands = {'clean' : clean, 'merge' : merge}
+
+def main():
+  f = commands[sys.argv[1]]
+  sys.exit(f(sys.argv[2:]))
+
+if __name__ == '__main__':
+  main()