../deps/v8/src/torque/implementation-visitor.cc

Bug Summary

File:	out/../deps/v8/src/torque/implementation-visitor.cc
Warning:	line 387, column 5 1st function call argument is an uninitialized value
Annotated Source Code

Press '?' to see keyboard shortcuts
Show analyzer invocation
clang -cc1 -cc1 -triple x86_64-unknown-linux-gnu -analyze -disable-free -clear-ast-before-backend -disable-llvm-verifier -discard-value-names -main-file-name implementation-visitor.cc -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=cplusplus -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=all -relaxed-aliasing -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -ffunction-sections -fdata-sections -fcoverage-compilation-dir=/home/maurizio/node-v18.6.0/out -resource-dir /usr/local/lib/clang/16.0.0 -D _GLIBCXX_USE_CXX11_ABI=1 -D NODE_OPENSSL_CONF_NAME=nodejs_conf -D NODE_OPENSSL_HAS_QUIC -D V8_GYP_BUILD -D V8_TYPED_ARRAY_MAX_SIZE_IN_HEAP=64 -D __STDC_FORMAT_MACROS -D OPENSSL_NO_PINSHARED -D OPENSSL_THREADS -D V8_TARGET_ARCH_X64 -D V8_HAVE_TARGET_OS -D V8_TARGET_OS_LINUX -D V8_EMBEDDER_STRING="-node.8" -D ENABLE_DISASSEMBLER -D V8_PROMISE_INTERNAL_FIELD_COUNT=1 -D V8_SHORT_BUILTIN_CALLS -D OBJECT_PRINT -D V8_INTL_SUPPORT -D V8_ATOMIC_OBJECT_FIELD_WRITES -D V8_ENABLE_LAZY_SOURCE_POSITIONS -D V8_USE_SIPHASH -D V8_SHARED_RO_HEAP -D V8_WIN64_UNWINDING_INFO -D V8_ENABLE_REGEXP_INTERPRETER_THREADED_DISPATCH -D V8_SNAPSHOT_COMPRESSION -D V8_ENABLE_WEBASSEMBLY -D V8_ENABLE_JAVASCRIPT_PROMISE_HOOKS -D V8_ALLOCATION_FOLDING -D V8_ALLOCATION_SITE_TRACKING -D V8_SCRIPTORMODULE_LEGACY_LIFETIME -D V8_ADVANCED_BIGINT_ALGORITHMS -I ../deps/v8 -I ../deps/v8/include -internal-isystem /usr/lib/gcc/x86_64-redhat-linux/8/../../../../include/c++/8 -internal-isystem /usr/lib/gcc/x86_64-redhat-linux/8/../../../../include/c++/8/x86_64-redhat-linux -internal-isystem /usr/lib/gcc/x86_64-redhat-linux/8/../../../../include/c++/8/backward -internal-isystem /usr/local/lib/clang/16.0.0/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-redhat-linux/8/../../../../x86_64-redhat-linux/include -internal-externc-isystem /include -internal-externc-isystem /usr/include -O3 -Wno-unused-parameter -Wno-return-type -std=gnu++17 -fdeprecated-macro -fdebug-compilation-dir=/home/maurizio/node-v18.6.0/out -ferror-limit 19 -fno-rtti -fgnuc-version=4.2.1 -fcxx-exceptions -fexceptions -vectorize-loops -vectorize-slp -analyzer-output=html -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/scan-build-2022-08-22-142216-507842-1 -x c++ ../deps/v8/src/torque/implementation-visitor.cc
1// Copyright 2017 the V8 project authors. All rights reserved.
2// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.

5#include "src/torque/implementation-visitor.h"

7#include <algorithm>
8#include <iomanip>
9#include <string>

11#include "src/base/optional.h"
12#include "src/common/globals.h"
13#include "src/numbers/integer-literal-inl.h"
14#include "src/torque/cc-generator.h"
15#include "src/torque/cfg.h"
16#include "src/torque/constants.h"
17#include "src/torque/cpp-builder.h"
18#include "src/torque/csa-generator.h"
19#include "src/torque/declaration-visitor.h"
20#include "src/torque/global-context.h"
21#include "src/torque/kythe-data.h"
22#include "src/torque/parameter-difference.h"
23#include "src/torque/server-data.h"
24#include "src/torque/source-positions.h"
25#include "src/torque/type-inference.h"
26#include "src/torque/type-visitor.h"
27#include "src/torque/types.h"
28#include "src/torque/utils.h"

30namespace v8 {
31namespace internal {
32namespace torque {

34uint64_t next_unique_binding_index = 0;

36// Sadly, 'using std::string_literals::operator""s;' is bugged in MSVC (see
37// https://developercommunity.visualstudio.com/t/Incorrect-warning-when-using-standard-st/673948).
38// TODO(nicohartmann@): Change to 'using std::string_literals::operator""s;'
39// once this is fixed.
40using namespace std::string_literals;  // NOLINT(build/namespaces)

42namespace {
43const char* BuiltinIncludesMarker = "// __BUILTIN_INCLUDES_MARKER__\n";
44}  // namespace

46VisitResult ImplementationVisitor::Visit(Expression* expr) {
CurrentSourcePosition::Scope scope(expr->pos);
switch (expr->kind) {
49#define ENUM_ITEM(name)        \
case AstNode::Kind::k##name: \
  return Visit(name::cast(expr));
  AST_EXPRESSION_NODE_KIND_LIST(ENUM_ITEM)ENUM_ITEM(CallExpression) ENUM_ITEM(CallMethodExpression) ENUM_ITEM
(IntrinsicCallExpression) ENUM_ITEM(StructExpression) ENUM_ITEM
(LogicalOrExpression) ENUM_ITEM(LogicalAndExpression) ENUM_ITEM
(SpreadExpression) ENUM_ITEM(ConditionalExpression) ENUM_ITEM
(IdentifierExpression) ENUM_ITEM(StringLiteralExpression) ENUM_ITEM
(IntegerLiteralExpression) ENUM_ITEM(FloatingPointLiteralExpression
) ENUM_ITEM(FieldAccessExpression) ENUM_ITEM(ElementAccessExpression
) ENUM_ITEM(DereferenceExpression) ENUM_ITEM(AssignmentExpression
) ENUM_ITEM(IncrementDecrementExpression) ENUM_ITEM(NewExpression
) ENUM_ITEM(AssumeTypeImpossibleExpression) ENUM_ITEM(StatementExpression
) ENUM_ITEM(TryLabelExpression)
53#undef ENUM_ITEM
  default:
    UNREACHABLE()V8_Fatal("unreachable code");
}
57}

59const Type* ImplementationVisitor::Visit(Statement* stmt) {
CurrentSourcePosition::Scope scope(stmt->pos);
StackScope stack_scope(this);
const Type* result;
switch (stmt->kind) {
64#define ENUM_ITEM(name)               \
case AstNode::Kind::k##name:        \
  result = Visit(name::cast(stmt)); \
  break;
  AST_STATEMENT_NODE_KIND_LIST(ENUM_ITEM)ENUM_ITEM(BlockStatement) ENUM_ITEM(ExpressionStatement) ENUM_ITEM
(IfStatement) ENUM_ITEM(WhileStatement) ENUM_ITEM(ForLoopStatement
) ENUM_ITEM(BreakStatement) ENUM_ITEM(ContinueStatement) ENUM_ITEM
(ReturnStatement) ENUM_ITEM(DebugStatement) ENUM_ITEM(AssertStatement
) ENUM_ITEM(TailCallStatement) ENUM_ITEM(VarDeclarationStatement
) ENUM_ITEM(GotoStatement)
69#undef ENUM_ITEM
  default:
    UNREACHABLE()V8_Fatal("unreachable code");
}
DCHECK_EQ(result == TypeOracle::GetNeverType(),((void) 0)
          assembler().CurrentBlockIsComplete())((void) 0);
return result;
76}

78void ImplementationVisitor::BeginGeneratedFiles() {
std::set<SourceId> contains_class_definitions;
for (const ClassType* type : TypeOracle::GetClasses()) {
  if (type->ShouldGenerateCppClassDefinitions()) {
    contains_class_definitions.insert(type->AttributedToFile());
  }
}

for (SourceId source : SourceFileMap::AllSources()) {
  auto& streams = GlobalContext::GeneratedPerFile(source);
  // Output beginning of CSA .cc file.
  {
    cpp::File& file = streams.csa_cc;

    for (const std::string& include_path : GlobalContext::CppIncludes()) {
      file << "#include " << StringLiteralQuote(include_path) << "\n";
    }

    file << "// Required Builtins:\n";
    file << "#include \"torque-generated/" +
                SourceFileMap::PathFromV8RootWithoutExtension(source) +
                "-tq-csa.h\"\n";
    // Now that required include files are collected while generting the file,
    // we only know the full set at the end. Insert a marker here that is
    // replaced with the list of includes at the very end.
    // TODO(nicohartmann@): This is not the most beautiful way to do this,
    // replace once the cpp file builder is available, where this can be
    // handled easily.
    file << BuiltinIncludesMarker;
    file << "\n";

    streams.csa_cc.BeginNamespace("v8", "internal");
    streams.csa_ccfile << "\n";
  }
  // Output beginning of CSA .h file.
  {
    cpp::File& file = streams.csa_header;
    std::string header_define =
        "V8_GEN_TORQUE_GENERATED_" +
        UnderlinifyPath(SourceFileMap::PathFromV8Root(source)) + "_CSA_H_";
    streams.csa_header.BeginIncludeGuard(header_define);
    file << "#include \"src/builtins/torque-csa-header-includes.h\"\n";
    file << "\n";

    streams.csa_header.BeginNamespace("v8", "internal");
    streams.csa_headerfile << "\n";
  }
  // Output beginning of class definition .cc file.
  {
    cpp::File& file = streams.class_definition_cc;
    if (contains_class_definitions.count(source) != 0) {
      file << "#include \""
           << SourceFileMap::PathFromV8RootWithoutExtension(source)
           << "-inl.h\"\n\n";
      file << "#include \"torque-generated/class-verifiers.h\"\n";
      file << "#include \"src/objects/instance-type-inl.h\"\n\n";
    }

    streams.class_definition_cc.BeginNamespace("v8", "internal");
    streams.class_definition_ccfile << "\n";
  }
}
140}

142void ImplementationVisitor::EndGeneratedFiles() {
for (SourceId file : SourceFileMap::AllSources()) {
  auto& streams = GlobalContext::GeneratedPerFile(file);

  // Output ending of CSA .cc file.
  streams.csa_cc.EndNamespace("v8", "internal");

  // Output ending of CSA .h file.
  {
    std::string header_define =
        "V8_GEN_TORQUE_GENERATED_" +
        UnderlinifyPath(SourceFileMap::PathFromV8Root(file)) + "_CSA_H_";

    streams.csa_header.EndNamespace("v8", "internal");
    streams.csa_headerfile << "\n";
    streams.csa_header.EndIncludeGuard(header_define);
  }

  // Output ending of class definition .cc file.
  streams.class_definition_cc.EndNamespace("v8", "internal");
}
163}

165void ImplementationVisitor::BeginDebugMacrosFile() {
// TODO(torque-builer): Can use builder for debug_macros_*_
std::ostream& source = debug_macros_cc_;
std::ostream& header = debug_macros_h_;

source << "#include \"torque-generated/debug-macros.h\"\n\n";
source << "#include \"src/objects/swiss-name-dictionary.h\"\n";
source << "#include \"src/objects/ordered-hash-table.h\"\n";
source << "#include \"tools/debug_helper/debug-macro-shims.h\"\n";
source << "#include \"include/v8-internal.h\"\n";
source << "\n";

source << "namespace v8 {\n"
       << "namespace internal {\n"
       << "namespace debug_helper_internal {\n"
       << "\n";

const char* kHeaderDefine = "V8_GEN_TORQUE_GENERATED_DEBUG_MACROS_H_";
header << "#ifndef " << kHeaderDefine << "\n";
header << "#define " << kHeaderDefine << "\n\n";
header << "#include \"tools/debug_helper/debug-helper-internal.h\"\n";
header << "#include \"src/numbers/integer-literal.h\"\n";
header << "\n";

header << "namespace v8 {\n"
       << "namespace internal {\n"
       << "namespace debug_helper_internal {\n"
       << "\n";
193}

195void ImplementationVisitor::EndDebugMacrosFile() {
// TODO(torque-builder): Can use builder for debug_macros_*_
std::ostream& source = debug_macros_cc_;
std::ostream& header = debug_macros_h_;

source << "}  // namespace internal\n"
       << "}  // namespace v8\n"
       << "}  // namespace debug_helper_internal\n"
       << "\n";

header << "\n}  // namespace internal\n"
       << "}  // namespace v8\n"
       << "}  // namespace debug_helper_internal\n"
       << "\n";
header << "#endif  // V8_GEN_TORQUE_GENERATED_DEBUG_MACROS_H_\n";
210}

212void ImplementationVisitor::Visit(NamespaceConstant* decl) {
Signature signature{{}, base::nullopt, {{}, false}, 0, decl->type(),
                    {}, false};

BindingsManagersScope bindings_managers_scope;

cpp::Function f =
    GenerateFunction(nullptr, decl->external_name(), signature, {});

f.PrintDeclaration(csa_headerfile());

f.PrintDefinition(csa_ccfile(), [&](std::ostream& stream) {
  stream << "  compiler::CodeAssembler ca_(state_);\n";

  DCHECK(!signature.return_type->IsVoidOrNever())((void) 0);

  assembler_ = CfgAssembler(Stack<const Type*>{});

  VisitResult expression_result = Visit(decl->body());
  VisitResult return_result =
      GenerateImplicitConvert(signature.return_type, expression_result);

  CSAGenerator csa_generator{assembler().Result(), stream};
  Stack<std::string> values = *csa_generator.EmitGraph(Stack<std::string>{});

  assembler_ = base::nullopt;

  stream << "  return ";
  CSAGenerator::EmitCSAValue(return_result, values, stream);
  stream << ";";
});
243}

245void ImplementationVisitor::Visit(TypeAlias* alias) {
if (alias->IsRedeclaration()) return;
if (const ClassType* class_type = ClassType::DynamicCast(alias->type())) {
  if (class_type->IsExtern() && !class_type->nspace()->IsDefaultNamespace()) {
    Error(
        "extern classes are currently only supported in the default "
        "namespace");
  }
}
254}

256class ImplementationVisitor::MacroInliningScope {
public:
MacroInliningScope(ImplementationVisitor* visitor, const Macro* macro)
    : visitor_(visitor), macro_(macro) {
  if (!visitor_->inlining_macros_.insert(macro).second) {
    // Recursive macro expansion would just keep going until stack overflow.
    // To avoid crashes, throw an error immediately.
    ReportError("Recursive macro call to ", *macro);
  }
}
~MacroInliningScope() { visitor_->inlining_macros_.erase(macro_); }

private:
ImplementationVisitor* visitor_;
const Macro* macro_;
271};

273VisitResult ImplementationVisitor::InlineMacro(
  Macro* macro, base::Optional<LocationReference> this_reference,
  const std::vector<VisitResult>& arguments,
  const std::vector<Block*> label_blocks) {
MacroInliningScope macro_inlining_scope(this, macro);
CurrentScope::Scope current_scope(macro);
BindingsManagersScope bindings_managers_scope;
CurrentCallable::Scope current_callable(macro);
CurrentReturnValue::Scope current_return_value;
const Signature& signature = macro->signature();
const Type* return_type = macro->signature().return_type;
bool can_return = return_type != TypeOracle::GetNeverType();
1
Assuming the condition is false→

BlockBindings<LocalValue> parameter_bindings(&ValueBindingsManager::Get());
BlockBindings<LocalLabel> label_bindings(&LabelBindingsManager::Get());
DCHECK_EQ(macro->signature().parameter_names.size(),((void) 0)
          arguments.size() + (this_reference ? 1 : 0))((void) 0);
DCHECK_EQ(this_reference.has_value(), macro->IsMethod())((void) 0);

// Bind the this for methods. Methods that modify a struct-type "this" must
// only be called if the this is in a variable, in which case the
// LocalValue is non-const. Otherwise, the LocalValue used for the parameter
// binding is const, and thus read-only, which will cause errors if
// modified, e.g. when called by a struct method that sets the structs
// fields. This prevents using temporary struct values for anything other
// than read operations.
if (this_reference) {
2
←
Assuming the condition is false→
3
←
Taking false branch→
  DCHECK(macro->IsMethod())((void) 0);
  parameter_bindings.Add(kThisParameterName, LocalValue{*this_reference},
                         true);
  // TODO(v8:12261): Tracking 'this'-binding for kythe led to a few weird
  // issues. Review to fully support 'this' in methods.
}

size_t count = 0;
for (auto arg : arguments) {
  if (this_reference && count == signature.implicit_count) count++;
  const bool mark_as_used = signature.implicit_count > count;
  const Identifier* name = macro->parameter_names()[count++];
  Binding<LocalValue>* binding =
      parameter_bindings.Add(name,
                             LocalValue{LocationReference::Temporary(
                                 arg, "parameter " + name->value)},
                             mark_as_used);
  if (GlobalContext::collect_kythe_data()) {
    KytheData::AddBindingDefinition(binding);
  }
}

DCHECK_EQ(label_blocks.size(), signature.labels.size())((void) 0);
for (size_t i = 0; i < signature.labels.size(); ++i) {
4
←
Assuming the condition is false→
5
←
Loop condition is false. Execution continues on line 332→
  const LabelDeclaration& label_info = signature.labels[i];
  Binding<LocalLabel>* binding = label_bindings.Add(
      label_info.name, LocalLabel{label_blocks[i], label_info.types});
  if (GlobalContext::collect_kythe_data()) {
    KytheData::AddBindingDefinition(binding);
  }
}

Block* macro_end;
6
←
'macro_end' declared without an initial value→
base::Optional<Binding<LocalLabel>> macro_end_binding;
if (can_return6.1
'can_return' is false
) {
7
←
Taking false branch→
  Stack<const Type*> stack = assembler().CurrentStack();
  std::vector<const Type*> lowered_return_types = LowerType(return_type);
  stack.PushMany(lowered_return_types);
  if (!return_type->IsConstexpr()) {
    SetReturnValue(VisitResult(return_type,
                               stack.TopRange(lowered_return_types.size())));
  }
  // The stack copy used to initialize the _macro_end block is only used
  // as a template for the actual gotos generated by return statements. It
  // doesn't correspond to any real return values, and thus shouldn't contain
  // top types, because these would pollute actual return value types that get
  // unioned with them for return statements, erroneously forcing them to top.
  for (auto i = stack.begin(); i != stack.end(); ++i) {
    if ((*i)->IsTopType()) {
      *i = TopType::cast(*i)->source_type();
    }
  }
  macro_end = assembler().NewBlock(std::move(stack));
  macro_end_binding.emplace(&LabelBindingsManager::Get(), kMacroEndLabelName,
                            LocalLabel{macro_end, {return_type}});
} else {
  SetReturnValue(VisitResult::NeverResult());
}

const Type* result = Visit(*macro->body());

if (result->IsNever()) {
8
←
Assuming the condition is false→
9
←
Taking false branch→
  if (!return_type->IsNever() && !macro->HasReturns()) {
    std::stringstream s;
    s << "macro " << macro->ReadableName()
      << " that never returns must have return type never";
    ReportError(s.str());
  }
} else {
  if (return_type->IsNever()) {
10
←
Assuming the condition is false→
11
←
Taking false branch→
    std::stringstream s;
    s << "macro " << macro->ReadableName()
      << " has implicit return at end of its declartion but return type "
         "never";
    ReportError(s.str());
  } else if (!macro->signature().return_type->IsVoid()) {
12
←
Assuming the condition is false→
13
←
Taking false branch→
    std::stringstream s;
    s << "macro " << macro->ReadableName()
      << " expects to return a value but doesn't on all paths";
    ReportError(s.str());
  }
}
if (!result->IsNever()) {
14
←
Assuming the condition is false→
  assembler().Goto(macro_end);
}

if (macro->HasReturns() || !result->IsNever()) {
15
←
Assuming the condition is false→
16
←
Assuming the condition is true→
17
←
Taking true branch→
  assembler().Bind(macro_end);
18
←
1st function call argument is an uninitialized value
}

return GetAndClearReturnValue();
391}

393void ImplementationVisitor::VisitMacroCommon(Macro* macro) {
CurrentCallable::Scope current_callable(macro);
const Signature& signature = macro->signature();
const Type* return_type = macro->signature().return_type;
bool can_return = return_type != TypeOracle::GetNeverType();
bool has_return_value =
    can_return && return_type != TypeOracle::GetVoidType();

cpp::Function f = GenerateMacroFunctionDeclaration(macro);
f.PrintDeclaration(csa_headerfile());
csa_headerfile() << "\n";

cpp::File csa_cc(csa_ccfile());

// Avoid multiple-definition errors since it is possible for multiple
// generated -inl.inc files to all contain function definitions for the same
// Torque macro.
base::Optional<cpp::IncludeGuardScope> include_guard;
if (output_type_ == OutputType::kCC) {
  include_guard.emplace(&csa_cc, "V8_INTERNAL_DEFINED_"s + macro->CCName());
} else if (output_type_ == OutputType::kCCDebug) {
  include_guard.emplace(&csa_cc,
                        "V8_INTERNAL_DEFINED_"s + macro->CCDebugName());
}

f.PrintBeginDefinition(csa_ccfile());

if (output_type_ == OutputType::kCC) {
  // For now, generated C++ is only for field offset computations. If we ever
  // generate C++ code that can allocate, then it should be handlified.
  csa_ccfile() << "  DisallowGarbageCollection no_gc;\n";
} else if (output_type_ == OutputType::kCSA) {
  csa_ccfile() << "  compiler::CodeAssembler ca_(state_);\n";
  csa_ccfile()
      << "  compiler::CodeAssembler::SourcePositionScope pos_scope(&ca_);\n";
}

Stack<std::string> lowered_parameters;
Stack<const Type*> lowered_parameter_types;

std::vector<VisitResult> arguments;

base::Optional<LocationReference> this_reference;
if (Method* method = Method::DynamicCast(macro)) {
  const Type* this_type = method->aggregate_type();
  LowerParameter(this_type, ExternalParameterName(kThisParameterName),
                 &lowered_parameters);
  StackRange range = lowered_parameter_types.PushMany(LowerType(this_type));
  VisitResult this_result = VisitResult(this_type, range);
  // For classes, mark 'this' as a temporary to prevent assignment to it.
  // Note that using a VariableAccess for non-class types is technically
  // incorrect because changes to the 'this' variable do not get reflected
  // to the caller. Therefore struct methods should always be inlined and a
  // C++ version should never be generated, since it would be incorrect.
  // However, in order to be able to type- and semantics-check even unused
  // struct methods, set the this_reference to be the local variable copy of
  // the passed-in this, which allows the visitor to at least find and report
  // errors.
  this_reference =
      (this_type->IsClassType())
          ? LocationReference::Temporary(this_result, "this parameter")
          : LocationReference::VariableAccess(this_result);
}

for (size_t i = 0; i < macro->signature().parameter_names.size(); ++i) {
  if (this_reference && i == macro->signature().implicit_count) continue;
  const std::string& name = macro->parameter_names()[i]->value;
  std::string external_name = ExternalParameterName(name);
  const Type* type = macro->signature().types()[i];

  if (type->IsConstexpr()) {
    arguments.push_back(VisitResult(type, external_name));
  } else {
    LowerParameter(type, external_name, &lowered_parameters);
    StackRange range = lowered_parameter_types.PushMany(LowerType(type));
    arguments.push_back(VisitResult(type, range));
  }
}

DCHECK_EQ(lowered_parameters.Size(), lowered_parameter_types.Size())((void) 0);
assembler_ = CfgAssembler(lowered_parameter_types);

std::vector<Block*> label_blocks;
for (const LabelDeclaration& label_info : signature.labels) {
  Stack<const Type*> label_input_stack;
  for (const Type* type : label_info.types) {
    label_input_stack.PushMany(LowerType(type));
  }
  Block* block = assembler().NewBlock(std::move(label_input_stack));
  label_blocks.push_back(block);
}

VisitResult return_value =
    InlineMacro(macro, this_reference, arguments, label_blocks);
Block* end = assembler().NewBlock();
if (return_type != TypeOracle::GetNeverType()) {
  assembler().Goto(end);
}

for (size_t i = 0; i < label_blocks.size(); ++i) {
  Block* label_block = label_blocks[i];
  const LabelDeclaration& label_info = signature.labels[i];
  assembler().Bind(label_block);
  std::vector<std::string> label_parameter_variables;
  for (size_t j = 0; j < label_info.types.size(); ++j) {
    LowerLabelParameter(label_info.types[j],
                        ExternalLabelParameterName(label_info.name->value, j),
                        &label_parameter_variables);
  }
  assembler().Emit(GotoExternalInstruction{
      ExternalLabelName(label_info.name->value), label_parameter_variables});
}

if (return_type != TypeOracle::GetNeverType()) {
  assembler().Bind(end);
}

base::Optional<Stack<std::string>> values;
if (output_type_ == OutputType::kCC) {
  CCGenerator cc_generator{assembler().Result(), csa_ccfile()};
  values = cc_generator.EmitGraph(lowered_parameters);
} else if (output_type_ == OutputType::kCCDebug) {
  CCGenerator cc_generator{assembler().Result(), csa_ccfile(), true};
  values = cc_generator.EmitGraph(lowered_parameters);
} else {
  CSAGenerator csa_generator{assembler().Result(), csa_ccfile()};
  values = csa_generator.EmitGraph(lowered_parameters);
}

assembler_ = base::nullopt;

if (has_return_value) {
  csa_ccfile() << "  return ";
  if (output_type_ == OutputType::kCCDebug) {
    csa_ccfile() << "{d::MemoryAccessResult::kOk, ";
    CCGenerator::EmitCCValue(return_value, *values, csa_ccfile());
    csa_ccfile() << "}";
  } else if (output_type_ == OutputType::kCC) {
    CCGenerator::EmitCCValue(return_value, *values, csa_ccfile());
  } else {
    CSAGenerator::EmitCSAValue(return_value, *values, csa_ccfile());
  }
  csa_ccfile() << ";\n";
}
f.PrintEndDefinition(csa_ccfile());

include_guard.reset();
540}

542void ImplementationVisitor::Visit(TorqueMacro* macro) {
VisitMacroCommon(macro);
544}

546void ImplementationVisitor::Visit(Method* method) {
DCHECK(!method->IsExternal())((void) 0);
VisitMacroCommon(method);
549}

551namespace {

553std::string AddParameter(size_t i, Builtin* builtin,
                       Stack<std::string>* parameters,
                       Stack<const Type*>* parameter_types,
                       BlockBindings<LocalValue>* parameter_bindings,
                       bool mark_as_used) {
const Identifier* name = builtin->signature().parameter_names[i];
const Type* type = builtin->signature().types()[i];
std::string external_name = "parameter" + std::to_string(i);
parameters->Push(external_name);
StackRange range = parameter_types->PushMany(LowerType(type));
Binding<LocalValue>* binding = parameter_bindings->Add(
    name,
    LocalValue{LocationReference::Temporary(VisitResult(type, range),
                                            "parameter " + name->value)},
    mark_as_used);
if (GlobalContext::collect_kythe_data()) {
  KytheData::AddBindingDefinition(binding);
}
return external_name;
572}

574}  // namespace

576void ImplementationVisitor::Visit(Builtin* builtin) {
if (builtin->IsExternal()) return;
CurrentScope::Scope current_scope(builtin);
CurrentCallable::Scope current_callable(builtin);
CurrentReturnValue::Scope current_return_value;

const std::string& name = builtin->ExternalName();
const Signature& signature = builtin->signature();
csa_ccfile() << "TF_BUILTIN(" << name << ", CodeStubAssembler) {\n"
             << "  compiler::CodeAssemblerState* state_ = state();"
             << "  compiler::CodeAssembler ca_(state());\n";

Stack<const Type*> parameter_types;
Stack<std::string> parameters;

BindingsManagersScope bindings_managers_scope;

BlockBindings<LocalValue> parameter_bindings(&ValueBindingsManager::Get());

if (builtin->IsVarArgsJavaScript() || builtin->IsFixedArgsJavaScript()) {
  if (builtin->IsVarArgsJavaScript()) {
    DCHECK(signature.parameter_types.var_args)((void) 0);
    if (signature.ExplicitCount() > 0) {
      Error("Cannot mix explicit parameters with varargs.")
          .Position(signature.parameter_names[signature.implicit_count]->pos);
    }

    csa_ccfile() << "  TNode<Word32T> argc = UncheckedParameter<Word32T>("
                 << "Descriptor::kJSActualArgumentsCount);\n";
    csa_ccfile() << "  TNode<IntPtrT> "
                    "arguments_length(ChangeInt32ToIntPtr(UncheckedCast<"
                    "Int32T>(argc)));\n";
    csa_ccfile() << "  TNode<RawPtrT> arguments_frame = "
                    "UncheckedCast<RawPtrT>(LoadFramePointer());\n";
    csa_ccfile()
        << "  TorqueStructArguments "
           "torque_arguments(GetFrameArguments(arguments_frame, "
           "arguments_length, FrameArgumentsArgcType::kCountIncludesReceiver"
        << "));\n";
    csa_ccfile()
        << "  CodeStubArguments arguments(this, torque_arguments);\n";

    parameters.Push("torque_arguments.frame");
    parameters.Push("torque_arguments.base");
    parameters.Push("torque_arguments.length");
    parameters.Push("torque_arguments.actual_count");
    const Type* arguments_type = TypeOracle::GetArgumentsType();
    StackRange range = parameter_types.PushMany(LowerType(arguments_type));
    parameter_bindings.Add(*signature.arguments_variable,
                           LocalValue{LocationReference::Temporary(
                               VisitResult(arguments_type, range),
                               "parameter " + *signature.arguments_variable)},
                           true);
  }

  for (size_t i = 0; i < signature.implicit_count; ++i) {
    const std::string& param_name = signature.parameter_names[i]->value;
    SourcePosition param_pos = signature.parameter_names[i]->pos;
    std::string generated_name = AddParameter(
        i, builtin, &parameters, &parameter_types, &parameter_bindings, true);
    const Type* actual_type = signature.parameter_types.types[i];
    std::vector<const Type*> expected_types;
    if (param_name == "context") {
      csa_ccfile() << "  TNode<NativeContext> " << generated_name
                   << " = UncheckedParameter<NativeContext>("
                   << "Descriptor::kContext);\n";
      csa_ccfile() << "  USE(" << generated_name << ");\n";
      expected_types = {TypeOracle::GetNativeContextType(),
                        TypeOracle::GetContextType()};
    } else if (param_name == "receiver") {
      csa_ccfile()
          << "  TNode<Object> " << generated_name << " = "
          << (builtin->IsVarArgsJavaScript()
                  ? "arguments.GetReceiver()"
                  : "UncheckedParameter<Object>(Descriptor::kReceiver)")
          << ";\n";
      csa_ccfile() << "  USE(" << generated_name << ");\n";
      expected_types = {TypeOracle::GetJSAnyType()};
    } else if (param_name == "newTarget") {
      csa_ccfile() << "  TNode<Object> " << generated_name
                   << " = UncheckedParameter<Object>("
                   << "Descriptor::kJSNewTarget);\n";
      csa_ccfile() << "USE(" << generated_name << ");\n";
      expected_types = {TypeOracle::GetJSAnyType()};
    } else if (param_name == "target") {
      csa_ccfile() << "  TNode<JSFunction> " << generated_name
                   << " = UncheckedParameter<JSFunction>("
                   << "Descriptor::kJSTarget);\n";
      csa_ccfile() << "USE(" << generated_name << ");\n";
      expected_types = {TypeOracle::GetJSFunctionType()};
    } else {
      Error(
          "Unexpected implicit parameter \"", param_name,
          "\" for JavaScript calling convention, "
          "expected \"context\", \"receiver\", \"target\", or \"newTarget\"")
          .Position(param_pos);
      expected_types = {actual_type};
    }
    if (std::find(expected_types.begin(), expected_types.end(),
                  actual_type) == expected_types.end()) {
      Error("According to JavaScript calling convention, expected parameter ",
            param_name, " to have type ", PrintList(expected_types, " or "),
            " but found type ", *actual_type)
          .Position(param_pos);
    }
  }

  for (size_t i = signature.implicit_count;
       i < signature.parameter_names.size(); ++i) {
    const std::string& parameter_name = signature.parameter_names[i]->value;
    const Type* type = signature.types()[i];
    const bool mark_as_used = signature.implicit_count > i;
    std::string var = AddParameter(i, builtin, &parameters, &parameter_types,
                                   &parameter_bindings, mark_as_used);
    csa_ccfile() << "  " << type->GetGeneratedTypeName() << " " << var
                 << " = "
                 << "UncheckedParameter<" << type->GetGeneratedTNodeTypeName()
                 << ">(Descriptor::k" << CamelifyString(parameter_name)
                 << ");\n";
    csa_ccfile() << "  USE(" << var << ");\n";
  }

} else {
  DCHECK(builtin->IsStub())((void) 0);

  for (size_t i = 0; i < signature.parameter_names.size(); ++i) {
    const std::string& parameter_name = signature.parameter_names[i]->value;
    const Type* type = signature.types()[i];
    const bool mark_as_used = signature.implicit_count > i;
    std::string var = AddParameter(i, builtin, &parameters, &parameter_types,
                                   &parameter_bindings, mark_as_used);
    csa_ccfile() << "  " << type->GetGeneratedTypeName() << " " << var
                 << " = "
                 << "UncheckedParameter<" << type->GetGeneratedTNodeTypeName()
                 << ">(Descriptor::k" << CamelifyString(parameter_name)
                 << ");\n";
    csa_ccfile() << "  USE(" << var << ");\n";
  }
}
assembler_ = CfgAssembler(parameter_types);
const Type* body_result = Visit(*builtin->body());
if (body_result != TypeOracle::GetNeverType()) {
  ReportError("control reaches end of builtin, expected return of a value");
}
CSAGenerator csa_generator{assembler().Result(), csa_ccfile(),
                           builtin->kind()};
csa_generator.EmitGraph(parameters);
assembler_ = base::nullopt;
csa_ccfile() << "}\n\n";
725}

727const Type* ImplementationVisitor::Visit(VarDeclarationStatement* stmt) {
BlockBindings<LocalValue> block_bindings(&ValueBindingsManager::Get());
return Visit(stmt, &block_bindings);
730}

732const Type* ImplementationVisitor::Visit(
  VarDeclarationStatement* stmt, BlockBindings<LocalValue>* block_bindings) {
// const qualified variables are required to be initialized properly.
if (stmt->const_qualified && !stmt->initializer) {
  ReportError("local constant \"", stmt->name, "\" is not initialized.");
}

base::Optional<const Type*> type;
if (stmt->type) {
  type = TypeVisitor::ComputeType(*stmt->type);
}
base::Optional<VisitResult> init_result;
if (stmt->initializer) {
  StackScope scope(this);
  init_result = Visit(*stmt->initializer);
  if (type) {
    init_result = GenerateImplicitConvert(*type, *init_result);
  }
  type = init_result->type();
  if ((*type)->IsConstexpr() && !stmt->const_qualified) {
    Error("Use 'const' instead of 'let' for variable '", stmt->name->value,
          "' of constexpr type '", (*type)->ToString(), "'.")
        .Position(stmt->name->pos)
        .Throw();
  }
  init_result = scope.Yield(*init_result);
} else {
  DCHECK(type.has_value())((void) 0);
  if ((*type)->IsConstexpr()) {
    ReportError("constexpr variables need an initializer");
  }
  TypeVector lowered_types = LowerType(*type);
  for (const Type* t : lowered_types) {
    assembler().Emit(PushUninitializedInstruction{TypeOracle::GetTopType(
        "uninitialized variable '" + stmt->name->value + "' of type " +
            t->ToString() + " originally defined at " +
            PositionAsString(stmt->pos),
        t)});
  }
  init_result =
      VisitResult(*type, assembler().TopRange(lowered_types.size()));
}
LocationReference ref = stmt->const_qualified
                            ? LocationReference::Temporary(
                                  *init_result, "const " + stmt->name->value)
                            : LocationReference::VariableAccess(*init_result);
block_bindings->Add(stmt->name, LocalValue{std::move(ref)});
return TypeOracle::GetVoidType();
780}

782const Type* ImplementationVisitor::Visit(TailCallStatement* stmt) {
return Visit(stmt->call, true).type();
784}

786VisitResult ImplementationVisitor::Visit(ConditionalExpression* expr) {
Block* true_block = assembler().NewBlock(assembler().CurrentStack());
Block* false_block = assembler().NewBlock(assembler().CurrentStack());
Block* done_block = assembler().NewBlock();
Block* true_conversion_block = assembler().NewBlock();
GenerateExpressionBranch(expr->condition, true_block, false_block);

VisitResult left;
VisitResult right;

{
  // The code for both paths of the conditional need to be generated first
  // before evaluating the conditional expression because the common type of
  // the result of both the true and false of the condition needs to be known
  // to convert both branches to a common type.
  assembler().Bind(true_block);
  StackScope left_scope(this);
  left = Visit(expr->if_true);
  assembler().Goto(true_conversion_block);

  const Type* common_type;
  {
    assembler().Bind(false_block);
    StackScope right_scope(this);
    right = Visit(expr->if_false);
    common_type = GetCommonType(left.type(), right.type());
    right = right_scope.Yield(GenerateImplicitConvert(common_type, right));
    assembler().Goto(done_block);
  }

  assembler().Bind(true_conversion_block);
  left = left_scope.Yield(GenerateImplicitConvert(common_type, left));
  assembler().Goto(done_block);
}

assembler().Bind(done_block);
CHECK_EQ(left, right)do { bool _cmp = ::v8::base::CmpEQImpl< typename ::v8::base
::pass_value_or_ref<decltype(left)>::type, typename ::v8
::base::pass_value_or_ref<decltype(right)>::type>((left
), (right)); do { if ((__builtin_expect(!!(!(_cmp)), 0))) { V8_Fatal
("Check failed: %s.", "left" " " "==" " " "right"); } } while
 (false); } while (false);
return left;
824}

826VisitResult ImplementationVisitor::Visit(LogicalOrExpression* expr) {
StackScope outer_scope(this);
VisitResult left_result = Visit(expr->left);

if (left_result.type()->IsConstexprBool()) {
  VisitResult right_result = Visit(expr->right);
  if (!right_result.type()->IsConstexprBool()) {
    ReportError(
        "expected type constexpr bool on right-hand side of operator "
        "||");
  }
  return VisitResult(TypeOracle::GetConstexprBoolType(),
                     std::string("(") + left_result.constexpr_value() +
                         " || " + right_result.constexpr_value() + ")");
}

Block* true_block = assembler().NewBlock();
Block* false_block = assembler().NewBlock();
Block* done_block = assembler().NewBlock();

left_result = GenerateImplicitConvert(TypeOracle::GetBoolType(), left_result);
GenerateBranch(left_result, true_block, false_block);

assembler().Bind(true_block);
VisitResult true_result = GenerateBoolConstant(true);
assembler().Goto(done_block);

assembler().Bind(false_block);
VisitResult false_result;
{
  StackScope false_block_scope(this);
  false_result = false_block_scope.Yield(
      GenerateImplicitConvert(TypeOracle::GetBoolType(), Visit(expr->right)));
}
assembler().Goto(done_block);

assembler().Bind(done_block);
DCHECK_EQ(true_result, false_result)((void) 0);
return outer_scope.Yield(true_result);
865}

867VisitResult ImplementationVisitor::Visit(LogicalAndExpression* expr) {
StackScope outer_scope(this);
VisitResult left_result = Visit(expr->left);

if (left_result.type()->IsConstexprBool()) {
  VisitResult right_result = Visit(expr->right);
  if (!right_result.type()->IsConstexprBool()) {
    ReportError(
        "expected type constexpr bool on right-hand side of operator "
        "&&");
  }
  return VisitResult(TypeOracle::GetConstexprBoolType(),
                     std::string("(") + left_result.constexpr_value() +
                         " && " + right_result.constexpr_value() + ")");
}

Block* true_block = assembler().NewBlock();
Block* false_block = assembler().NewBlock();
Block* done_block = assembler().NewBlock();

left_result = GenerateImplicitConvert(TypeOracle::GetBoolType(), left_result);
GenerateBranch(left_result, true_block, false_block);

assembler().Bind(true_block);
VisitResult true_result;
{
  StackScope true_block_scope(this);
  VisitResult right_result = Visit(expr->right);
  if (TryGetSourceForBitfieldExpression(expr->left) != nullptr &&
      TryGetSourceForBitfieldExpression(expr->right) != nullptr &&
      TryGetSourceForBitfieldExpression(expr->left)->value ==
          TryGetSourceForBitfieldExpression(expr->right)->value) {
    Lint(
        "Please use & rather than && when checking multiple bitfield "
        "values, to avoid complexity in generated code.");
  }
  true_result = true_block_scope.Yield(
      GenerateImplicitConvert(TypeOracle::GetBoolType(), right_result));
}
assembler().Goto(done_block);

assembler().Bind(false_block);
VisitResult false_result = GenerateBoolConstant(false);
assembler().Goto(done_block);

assembler().Bind(done_block);
DCHECK_EQ(true_result, false_result)((void) 0);
return outer_scope.Yield(true_result);
915}

917VisitResult ImplementationVisitor::Visit(IncrementDecrementExpression* expr) {
StackScope scope(this);
LocationReference location_ref = GetLocationReference(expr->location);
VisitResult current_value = GenerateFetchFromLocation(location_ref);
VisitResult one = {TypeOracle::GetConstInt31Type(), "1"};
Arguments args;
args.parameters = {current_value, one};
VisitResult assignment_value = GenerateCall(
    expr->op == IncrementDecrementOperator::kIncrement ? "+" : "-", args);
GenerateAssignToLocation(location_ref, assignment_value);
return scope.Yield(expr->postfix ? current_value : assignment_value);
928}

930VisitResult ImplementationVisitor::Visit(AssignmentExpression* expr) {
StackScope scope(this);
LocationReference location_ref = GetLocationReference(expr->location);
VisitResult assignment_value;
if (expr->op) {
  VisitResult location_value = GenerateFetchFromLocation(location_ref);
  assignment_value = Visit(expr->value);
  Arguments args;
  args.parameters = {location_value, assignment_value};
  assignment_value = GenerateCall(*expr->op, args);
  GenerateAssignToLocation(location_ref, assignment_value);
} else {
  assignment_value = Visit(expr->value);
  GenerateAssignToLocation(location_ref, assignment_value);
}
return scope.Yield(assignment_value);
946}

948VisitResult ImplementationVisitor::Visit(FloatingPointLiteralExpression* expr) {
const Type* result_type = TypeOracle::GetConstFloat64Type();
std::stringstream str;
str << std::setprecision(std::numeric_limits<double>::digits10 + 1)
    << expr->value;
return VisitResult{result_type, str.str()};
954}

956VisitResult ImplementationVisitor::Visit(IntegerLiteralExpression* expr) {
const Type* result_type = TypeOracle::GetIntegerLiteralType();
std::stringstream str;
str << "IntegerLiteral("
    << (expr->value.is_negative() ? "true, 0x" : "false, 0x") << std::hex
    << expr->value.absolute_value() << std::dec << "ull)";
return VisitResult{result_type, str.str()};
963}

965VisitResult ImplementationVisitor::Visit(AssumeTypeImpossibleExpression* expr) {
VisitResult result = Visit(expr->expression);
const Type* result_type = SubtractType(
    result.type(), TypeVisitor::ComputeType(expr->excluded_type));
if (result_type->IsNever()) {
  ReportError("unreachable code");
}
CHECK_EQ(LowerType(result_type), TypeVector{result_type})do { bool _cmp = ::v8::base::CmpEQImpl< typename ::v8::base
::pass_value_or_ref<decltype(LowerType(result_type))>::
type, typename ::v8::base::pass_value_or_ref<decltype(TypeVector
{result_type})>::type>((LowerType(result_type)), (TypeVector
{result_type})); do { if ((__builtin_expect(!!(!(_cmp)), 0)))
 { V8_Fatal("Check failed: %s.", "LowerType(result_type)" " "
 "==" " " "TypeVector{result_type}"); } } while (false); } while
 (false);
assembler().Emit(UnsafeCastInstruction{result_type});
result.SetType(result_type);
return result;
976}

978VisitResult ImplementationVisitor::Visit(StringLiteralExpression* expr) {
return VisitResult{
    TypeOracle::GetConstStringType(),
    "\"" + expr->literal.substr(1, expr->literal.size() - 2) + "\""};
982}

984VisitResult ImplementationVisitor::GetBuiltinCode(Builtin* builtin) {
if (builtin->IsExternal() || builtin->kind() != Builtin::kStub) {
  ReportError(
      "creating function pointers is only allowed for internal builtins with "
      "stub linkage");
}
const Type* type = TypeOracle::GetBuiltinPointerType(
    builtin->signature().parameter_types.types,
    builtin->signature().return_type);
assembler().Emit(
    PushBuiltinPointerInstruction{builtin->ExternalName(), type});
return VisitResult(type, assembler().TopRange(1));
996}

998VisitResult ImplementationVisitor::Visit(LocationExpression* expr) {
StackScope scope(this);
return scope.Yield(GenerateFetchFromLocation(GetLocationReference(expr)));
1001}

1003VisitResult ImplementationVisitor::Visit(FieldAccessExpression* expr) {
StackScope scope(this);
LocationReference location = GetLocationReference(expr);
if (location.IsBitFieldAccess()) {
  if (auto* identifier = IdentifierExpression::DynamicCast(expr->object)) {
    bitfield_expressions_[expr] = identifier->name;
  }
}
return scope.Yield(GenerateFetchFromLocation(location));
1012}

1014const Type* ImplementationVisitor::Visit(GotoStatement* stmt) {
Binding<LocalLabel>* label = LookupLabel(stmt->label->value);
size_t parameter_count = label->parameter_types.size();
if (stmt->arguments.size() != parameter_count) {
  ReportError("goto to label has incorrect number of parameters (expected ",
              parameter_count, " found ", stmt->arguments.size(), ")");
}

if (GlobalContext::collect_language_server_data()) {
  LanguageServerData::AddDefinition(stmt->label->pos,
                                    label->declaration_position());
}
if (GlobalContext::collect_kythe_data()) {
  KytheData::AddBindingUse(stmt->label->pos, label);
}

size_t i = 0;
StackRange arguments = assembler().TopRange(0);
for (Expression* e : stmt->arguments) {
  StackScope scope(this);
  VisitResult result = Visit(e);
  const Type* parameter_type = label->parameter_types[i++];
  result = GenerateImplicitConvert(parameter_type, result);
  arguments.Extend(scope.Yield(result).stack_range());
}

assembler().Goto(label->block, arguments.Size());
return TypeOracle::GetNeverType();
1042}

1044const Type* ImplementationVisitor::Visit(IfStatement* stmt) {
bool has_else = stmt->if_false.has_value();

if (stmt->is_constexpr) {
  VisitResult expression_result = Visit(stmt->condition);

  if (!(expression_result.type() == TypeOracle::GetConstexprBoolType())) {
    std::stringstream stream;
    stream << "expression should return type constexpr bool "
           << "but returns type " << *expression_result.type();
    ReportError(stream.str());
  }

  Block* true_block = assembler().NewBlock();
  Block* false_block = assembler().NewBlock();
  Block* done_block = assembler().NewBlock();

  assembler().Emit(ConstexprBranchInstruction{
      expression_result.constexpr_value(), true_block, false_block});

  assembler().Bind(true_block);
  const Type* left_result = Visit(stmt->if_true);
  if (left_result == TypeOracle::GetVoidType()) {
    assembler().Goto(done_block);
  }

  assembler().Bind(false_block);
  const Type* right_result = TypeOracle::GetVoidType();
  if (has_else) {
    right_result = Visit(*stmt->if_false);
  }
  if (right_result == TypeOracle::GetVoidType()) {
    assembler().Goto(done_block);
  }

  if (left_result->IsNever() != right_result->IsNever()) {
    std::stringstream stream;
    stream << "either both or neither branches in a constexpr if statement "
              "must reach their end at"
           << PositionAsString(stmt->pos);
    ReportError(stream.str());
  }

  if (left_result != TypeOracle::GetNeverType()) {
    assembler().Bind(done_block);
  }
  return left_result;
} else {
  Block* true_block = assembler().NewBlock(assembler().CurrentStack(),
                                           IsDeferred(stmt->if_true));
  Block* false_block =
      assembler().NewBlock(assembler().CurrentStack(),
                           stmt->if_false && IsDeferred(*stmt->if_false));
  GenerateExpressionBranch(stmt->condition, true_block, false_block);

  Block* done_block;
  bool live = false;
  if (has_else) {
    done_block = assembler().NewBlock();
  } else {
    done_block = false_block;
    live = true;
  }

  assembler().Bind(true_block);
  {
    const Type* result = Visit(stmt->if_true);
    if (result == TypeOracle::GetVoidType()) {
      live = true;
      assembler().Goto(done_block);
    }
  }

  if (has_else) {
    assembler().Bind(false_block);
    const Type* result = Visit(*stmt->if_false);
    if (result == TypeOracle::GetVoidType()) {
      live = true;
      assembler().Goto(done_block);
    }
  }

  if (live) {
    assembler().Bind(done_block);
  }
  return live ? TypeOracle::GetVoidType() : TypeOracle::GetNeverType();
}
1131}

1133const Type* ImplementationVisitor::Visit(WhileStatement* stmt) {
Block* body_block = assembler().NewBlock(assembler().CurrentStack());
Block* exit_block = assembler().NewBlock(assembler().CurrentStack());

Block* header_block = assembler().NewBlock();
assembler().Goto(header_block);

assembler().Bind(header_block);
GenerateExpressionBranch(stmt->condition, body_block, exit_block);

assembler().Bind(body_block);
{
  BreakContinueActivator activator{exit_block, header_block};
  const Type* body_result = Visit(stmt->body);
  if (body_result != TypeOracle::GetNeverType()) {
    assembler().Goto(header_block);
  }
}

assembler().Bind(exit_block);
return TypeOracle::GetVoidType();
1154}

1156const Type* ImplementationVisitor::Visit(BlockStatement* block) {
BlockBindings<LocalValue> block_bindings(&ValueBindingsManager::Get());
const Type* type = TypeOracle::GetVoidType();
for (Statement* s : block->statements) {
  CurrentSourcePosition::Scope source_position(s->pos);
  if (type->IsNever()) {
    ReportError("statement after non-returning statement");
  }
  if (auto* var_declaration = VarDeclarationStatement::DynamicCast(s)) {
    type = Visit(var_declaration, &block_bindings);
  } else {
    type = Visit(s);
  }
}
return type;
1171}

1173const Type* ImplementationVisitor::Visit(DebugStatement* stmt) {
1174#if defined(DEBUG)
assembler().Emit(PrintConstantStringInstruction{"halting because of '" +
                                                stmt->reason + "' at " +
                                                PositionAsString(stmt->pos)});
1178#endif
assembler().Emit(AbortInstruction{stmt->never_continues
                                      ? AbortInstruction::Kind::kUnreachable
                                      : AbortInstruction::Kind::kDebugBreak});
if (stmt->never_continues) {
  return TypeOracle::GetNeverType();
} else {
  return TypeOracle::GetVoidType();
}
1187}

1189namespace {

1191std::string FormatAssertSource(const std::string& str) {
// Replace all whitespace characters with a space character.
std::string str_no_newlines = str;
std::replace_if(
    str_no_newlines.begin(), str_no_newlines.end(),
    [](unsigned char c) { return isspace(c); }, ' ');

// str might include indentation, squash multiple space characters into one.
std::string result;
std::unique_copy(str_no_newlines.begin(), str_no_newlines.end(),
                 std::back_inserter(result),
                 [](char a, char b) { return a == ' ' && b == ' '; });
return result;
1204}

1206}  // namespace

1208const Type* ImplementationVisitor::Visit(AssertStatement* stmt) {
if (stmt->kind == AssertStatement::AssertKind::kStaticAssert) {
  std::string message =
      "static_assert(" + stmt->source + ") at " + ToString(stmt->pos);
  GenerateCall(QualifiedName({"", TORQUE_INTERNAL_NAMESPACE_STRING},
                             STATIC_ASSERT_MACRO_STRING),
               Arguments{{Visit(stmt->expression),
                          VisitResult(TypeOracle::GetConstexprStringType(),
                                      StringLiteralQuote(message))},
                         {}});
  return TypeOracle::GetVoidType();
}
bool do_check = stmt->kind != AssertStatement::AssertKind::kDcheck ||
                GlobalContext::force_assert_statements();
1222#if defined(DEBUG)
do_check = true;
1224#endif
Block* resume_block;

if (!do_check) {
  Block* unreachable_block = assembler().NewBlock(assembler().CurrentStack());
  resume_block = assembler().NewBlock(assembler().CurrentStack());
  assembler().Goto(resume_block);
  assembler().Bind(unreachable_block);
}

// CSA_DCHECK & co. are not used here on purpose for two reasons. First,
// Torque allows and handles two types of expressions in the if protocol
// automagically, ones that return TNode<BoolT> and those that use the
// BranchIf(..., Label* true, Label* false) idiom. Because the machinery to
// handle this is embedded in the expression handling and to it's not
// possible to make the decision to use CSA_DCHECK or CSA_DCHECK_BRANCH
// isn't trivial up-front. Secondly, on failure, the assert text should be
// the corresponding Torque code, not the -gen.cc code, which would be the
// case when using CSA_DCHECK_XXX.
Block* true_block = assembler().NewBlock(assembler().CurrentStack());
Block* false_block = assembler().NewBlock(assembler().CurrentStack(), true);
GenerateExpressionBranch(stmt->expression, true_block, false_block);

assembler().Bind(false_block);

assembler().Emit(AbortInstruction{
    AbortInstruction::Kind::kAssertionFailure,
    "Torque assert '" + FormatAssertSource(stmt->source) + "' failed"});

assembler().Bind(true_block);

if (!do_check) {
  assembler().Bind(resume_block);
}

return TypeOracle::GetVoidType();
1260}

1262const Type* ImplementationVisitor::Visit(ExpressionStatement* stmt) {
const Type* type = Visit(stmt->expression).type();
return type->IsNever() ? type : TypeOracle::GetVoidType();
1265}

1267const Type* ImplementationVisitor::Visit(ReturnStatement* stmt) {
Callable* current_callable = CurrentCallable::Get();
if (current_callable->signature().return_type->IsNever()) {
  std::stringstream s;
  s << "cannot return from a function with return type never";
  ReportError(s.str());
}
LocalLabel* end =
    current_callable->IsMacro() ? LookupLabel(kMacroEndLabelName) : nullptr;
if (current_callable->HasReturnValue()) {
  if (!stmt->value) {
    std::stringstream s;
    s << "return expression needs to be specified for a return type of "
      << *current_callable->signature().return_type;
    ReportError(s.str());
  }
  VisitResult expression_result = Visit(*stmt->value);
  VisitResult return_result = GenerateImplicitConvert(
      current_callable->signature().return_type, expression_result);
  if (current_callable->IsMacro()) {
    if (return_result.IsOnStack()) {
      StackRange return_value_range =
          GenerateLabelGoto(end, return_result.stack_range());
      SetReturnValue(VisitResult(return_result.type(), return_value_range));
    } else {
      GenerateLabelGoto(end);
      SetReturnValue(return_result);
    }
  } else if (current_callable->IsBuiltin()) {
    assembler().Emit(ReturnInstruction{
        LoweredSlotCount(current_callable->signature().return_type)});
  } else {
    UNREACHABLE()V8_Fatal("unreachable code");
  }
} else {
  if (stmt->value) {
    std::stringstream s;
    s << "return expression can't be specified for a void or never return "
         "type";
    ReportError(s.str());
  }
  GenerateLabelGoto(end);
}
current_callable->IncrementReturns();
return TypeOracle::GetNeverType();
1312}

1314VisitResult ImplementationVisitor::Visit(TryLabelExpression* expr) {
size_t parameter_count = expr->label_block->parameters.names.size();
std::vector<VisitResult> parameters;

Block* label_block = nullptr;
Block* done_block = assembler().NewBlock();
VisitResult try_result;

{
  CurrentSourcePosition::Scope source_position(expr->label_block->pos);
  if (expr->label_block->parameters.has_varargs) {
    ReportError("cannot use ... for label parameters");
  }
  Stack<const Type*> label_input_stack = assembler().CurrentStack();
  TypeVector parameter_types;
  for (size_t i = 0; i < parameter_count; ++i) {
    const Type* type =
        TypeVisitor::ComputeType(expr->label_block->parameters.types[i]);
    parameter_types.push_back(type);
    if (type->IsConstexpr()) {
      ReportError("no constexpr type allowed for label arguments");
    }
    StackRange range = label_input_stack.PushMany(LowerType(type));
    parameters.push_back(VisitResult(type, range));
  }
  label_block = assembler().NewBlock(label_input_stack,
                                     IsDeferred(expr->label_block->body));

  Binding<LocalLabel> label_binding{&LabelBindingsManager::Get(),
                                    expr->label_block->label,
                                    LocalLabel{label_block, parameter_types}};

  // Visit try
  StackScope stack_scope(this);
  try_result = Visit(expr->try_expression);
  if (try_result.type() != TypeOracle::GetNeverType()) {
    try_result = stack_scope.Yield(try_result);
    assembler().Goto(done_block);
  }
}

// Visit and output the code for the label block. If the label block falls
// through, then the try must not return a value. Also, if the try doesn't
// fall through, but the label does, then overall the try-label block
// returns type void.
assembler().Bind(label_block);
const Type* label_result;
{
  BlockBindings<LocalValue> parameter_bindings(&ValueBindingsManager::Get());
  for (size_t i = 0; i < parameter_count; ++i) {
    Identifier* name = expr->label_block->parameters.names[i];
    parameter_bindings.Add(name,
                           LocalValue{LocationReference::Temporary(
                               parameters[i], "parameter " + name->value)});
  }

  label_result = Visit(expr->label_block->body);
}
if (!try_result.type()->IsVoidOrNever() && label_result->IsVoid()) {
  ReportError(
      "otherwise clauses cannot fall through in a non-void expression");
}
if (label_result != TypeOracle::GetNeverType()) {
  assembler().Goto(done_block);
}
if (label_result->IsVoid() && try_result.type()->IsNever()) {
  try_result =
      VisitResult(TypeOracle::GetVoidType(), try_result.stack_range());
}

if (!try_result.type()->IsNever()) {
  assembler().Bind(done_block);
}
return try_result;
1388}

1390VisitResult ImplementationVisitor::Visit(StatementExpression* expr) {
return VisitResult{Visit(expr->statement), assembler().TopRange(0)};
1392}

1394InitializerResults ImplementationVisitor::VisitInitializerResults(
  const ClassType* class_type,
  const std::vector<NameAndExpression>& initializers) {
InitializerResults result;
for (const NameAndExpression& initializer : initializers) {
  result.names.push_back(initializer.name);
  Expression* e = initializer.expression;
  const Field& field = class_type->LookupField(initializer.name->value);
  bool has_index = field.index.has_value();
  if (SpreadExpression* s = SpreadExpression::DynamicCast(e)) {
    if (!has_index) {
      ReportError(
          "spread expressions can only be used to initialize indexed class "
          "fields ('",
          initializer.name->value, "' is not)");
    }
    e = s->spreadee;
  } else if (has_index) {
    ReportError("the indexed class field '", initializer.name->value,
                "' must be initialized with a spread operator");
  }
  result.field_value_map[field.name_and_type.name] = Visit(e);
}
return result;
1418}

1420LocationReference ImplementationVisitor::GenerateFieldReference(
  VisitResult object, const Field& field, const ClassType* class_type,
  bool treat_optional_as_indexed) {
if (field.index.has_value()) {
  LocationReference slice = LocationReference::HeapSlice(
      GenerateCall(class_type->GetSliceMacroName(field), {{object}, {}}));
  if (field.index->optional && !treat_optional_as_indexed) {
    // This field was declared using optional syntax, so any reference to it
    // is implicitly a reference to the first item.
    return GenerateReferenceToItemInHeapSlice(
        slice, {TypeOracle::GetConstInt31Type(), "0"});
  } else {
    return slice;
  }
}
DCHECK(field.offset.has_value())((void) 0);
StackRange result_range = assembler().TopRange(0);
result_range.Extend(GenerateCopy(object).stack_range());
VisitResult offset =
    VisitResult(TypeOracle::GetConstInt31Type(), ToString(*field.offset));
offset = GenerateImplicitConvert(TypeOracle::GetIntPtrType(), offset);
result_range.Extend(offset.stack_range());
const Type* type = TypeOracle::GetReferenceType(field.name_and_type.type,
                                                field.const_qualified);
return LocationReference::HeapReference(VisitResult(type, result_range));
1445}

1447// This is used to generate field references during initialization, where we can
1448// re-use the offsets used for computing the allocation size.
1449LocationReference ImplementationVisitor::GenerateFieldReferenceForInit(
  VisitResult object, const Field& field,
  const LayoutForInitialization& layout) {
StackRange result_range = assembler().TopRange(0);
result_range.Extend(GenerateCopy(object).stack_range());
VisitResult offset = GenerateImplicitConvert(
    TypeOracle::GetIntPtrType(), layout.offsets.at(field.name_and_type.name));
result_range.Extend(offset.stack_range());
if (field.index) {
  VisitResult length =
      GenerateCopy(layout.array_lengths.at(field.name_and_type.name));
  result_range.Extend(length.stack_range());
  const Type* slice_type =
      TypeOracle::GetMutableSliceType(field.name_and_type.type);
  return LocationReference::HeapSlice(VisitResult(slice_type, result_range));
} else {
  // Const fields are writable during initialization.
  VisitResult heap_reference(
      TypeOracle::GetMutableReferenceType(field.name_and_type.type),
      result_range);
  return LocationReference::HeapReference(heap_reference);
}
1471}

1473void ImplementationVisitor::InitializeClass(
  const ClassType* class_type, VisitResult allocate_result,
  const InitializerResults& initializer_results,
  const LayoutForInitialization& layout) {
if (const ClassType* super = class_type->GetSuperClass()) {
  InitializeClass(super, allocate_result, initializer_results, layout);
}

for (Field f : class_type->fields()) {
  VisitResult initializer_value =
      initializer_results.field_value_map.at(f.name_and_type.name);
  LocationReference field =
      GenerateFieldReferenceForInit(allocate_result, f, layout);
  if (f.index) {
    DCHECK(field.IsHeapSlice())((void) 0);
    VisitResult slice = field.GetVisitResult();
    GenerateCall(QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING},
                               "InitializeFieldsFromIterator"),
                 {{slice, initializer_value}, {}});
  } else {
    GenerateAssignToLocation(field, initializer_value);
  }
}
1496}

1498VisitResult ImplementationVisitor::GenerateArrayLength(
  Expression* array_length, Namespace* nspace,
  const std::map<std::string, LocalValue>& bindings) {
StackScope stack_scope(this);
CurrentSourcePosition::Scope pos_scope(array_length->pos);
// Switch to the namespace where the class was declared.
CurrentScope::Scope current_scope_scope(nspace);
// Reset local bindings and install local binding for the preceding fields.
BindingsManagersScope bindings_managers_scope;
BlockBindings<LocalValue> field_bindings(&ValueBindingsManager::Get());
for (auto& p : bindings) {
  field_bindings.Add(p.first, LocalValue{p.second}, true);
}
VisitResult length = Visit(array_length);
VisitResult converted_length =
    GenerateCall("Convert", Arguments{{length}, {}},
                 {TypeOracle::GetIntPtrType(), length.type()}, false);
return stack_scope.Yield(converted_length);
1516}

1518VisitResult ImplementationVisitor::GenerateArrayLength(VisitResult object,
                                                     const Field& field) {
DCHECK(field.index)((void) 0);

StackScope stack_scope(this);
const ClassType* class_type = *object.type()->ClassSupertype();
std::map<std::string, LocalValue> bindings;
bool before_current = true;
for (Field f : class_type->ComputeAllFields()) {
  if (field.name_and_type.name == f.name_and_type.name) {
    before_current = false;
  }
  // We can't generate field references eagerly here, because some preceding
  // fields might be optional, and attempting to get a reference to an
  // optional field can crash the program if the field isn't present.
  // Instead, we use the lazy form of LocalValue to only generate field
  // references if they are used in the length expression.
  bindings.insert(
      {f.name_and_type.name,
       f.const_qualified
           ? (before_current
                  ? LocalValue{[=]() {
                      return GenerateFieldReference(object, f, class_type);
                    }}
                  : LocalValue("Array lengths may only refer to fields "
                               "defined earlier"))
           : LocalValue(
                 "Non-const fields cannot be used for array lengths.")});
}
return stack_scope.Yield(
    GenerateArrayLength(field.index->expr, class_type->nspace(), bindings));
1549}

1551VisitResult ImplementationVisitor::GenerateArrayLength(
  const ClassType* class_type, const InitializerResults& initializer_results,
  const Field& field) {
DCHECK(field.index)((void) 0);

StackScope stack_scope(this);
std::map<std::string, LocalValue> bindings;
for (Field f : class_type->ComputeAllFields()) {
  if (f.index) break;
  const std::string& fieldname = f.name_and_type.name;
  VisitResult value = initializer_results.field_value_map.at(fieldname);
  bindings.insert(
      {fieldname,
       f.const_qualified
           ? LocalValue{LocationReference::Temporary(
                 value, "initial field " + fieldname)}
           : LocalValue(
                 "Non-const fields cannot be used for array lengths.")});
}
return stack_scope.Yield(
    GenerateArrayLength(field.index->expr, class_type->nspace(), bindings));
1572}

1574LayoutForInitialization ImplementationVisitor::GenerateLayoutForInitialization(
  const ClassType* class_type,
  const InitializerResults& initializer_results) {
LayoutForInitialization layout;
VisitResult offset;
for (Field f : class_type->ComputeAllFields()) {
  if (f.offset.has_value()) {
    offset =
        VisitResult(TypeOracle::GetConstInt31Type(), ToString(*f.offset));
  }
  layout.offsets[f.name_and_type.name] = offset;
  if (f.index) {
    size_t element_size;
    std::string element_size_string;
    std::tie(element_size, element_size_string) =
        *SizeOf(f.name_and_type.type);
    VisitResult array_element_size =
        VisitResult(TypeOracle::GetConstInt31Type(), element_size_string);
    VisitResult array_length =
        GenerateArrayLength(class_type, initializer_results, f);
    layout.array_lengths[f.name_and_type.name] = array_length;
    Arguments arguments;
    arguments.parameters = {offset, array_length, array_element_size};
    offset = GenerateCall(QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING},
                                        "AddIndexedFieldSizeToObjectSize"),
                          arguments);
  } else {
    DCHECK(f.offset.has_value())((void) 0);
  }
}
if (class_type->size().SingleValue()) {
  layout.size = VisitResult(TypeOracle::GetConstInt31Type(),
                            ToString(*class_type->size().SingleValue()));
} else {
  layout.size = offset;
}
if ((size_t{1} << class_type->size().AlignmentLog2()) <
    TargetArchitecture::TaggedSize()) {
  Arguments arguments;
  arguments.parameters = {layout.size};
  layout.size = GenerateCall(
      QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING}, "AlignTagged"),
      arguments);
}
return layout;
1619}

1621VisitResult ImplementationVisitor::Visit(NewExpression* expr) {
StackScope stack_scope(this);
const Type* type = TypeVisitor::ComputeType(expr->type);
const ClassType* class_type = ClassType::DynamicCast(type);
if (class_type == nullptr) {
  ReportError("type for new expression must be a class, \"", *type,
              "\" is not");
}

if (!class_type->AllowInstantiation()) {
  // Classes that are only used for testing should never be instantiated.
  ReportError(*class_type,
              " cannot be allocated with new (it's used for testing)");
}

InitializerResults initializer_results =
    VisitInitializerResults(class_type, expr->initializers);

const Field& map_field = class_type->LookupField("map");
if (*map_field.offset != 0) {
  ReportError("class initializers must have a map as first parameter");
}
const std::map<std::string, VisitResult>& initializer_fields =
    initializer_results.field_value_map;
auto it_object_map = initializer_fields.find(map_field.name_and_type.name);
VisitResult object_map;
if (class_type->IsExtern()) {
  if (it_object_map == initializer_fields.end()) {
    ReportError("Constructor for ", class_type->name(),
                " needs Map argument!");
  }
  object_map = it_object_map->second;
} else {
  if (it_object_map != initializer_fields.end()) {
    ReportError(
        "Constructor for ", class_type->name(),
        " must not specify Map argument; it is automatically inserted.");
  }
  Arguments get_struct_map_arguments;
  get_struct_map_arguments.parameters.push_back(
      VisitResult(TypeOracle::GetConstexprInstanceTypeType(),
                  CapifyStringWithUnderscores(class_type->name()) + "_TYPE"));
  object_map = GenerateCall(
      QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING}, "GetInstanceTypeMap"),
      get_struct_map_arguments, {}, false);
  CurrentSourcePosition::Scope current_pos(expr->pos);
  initializer_results.names.insert(initializer_results.names.begin(),
                                   MakeNode<Identifier>("map"));
  initializer_results.field_value_map[map_field.name_and_type.name] =
      object_map;
}

CheckInitializersWellformed(class_type->name(),
                            class_type->ComputeAllFields(),
                            expr->initializers, !class_type->IsExtern());

LayoutForInitialization layout =
    GenerateLayoutForInitialization(class_type, initializer_results);

Arguments allocate_arguments;
allocate_arguments.parameters.push_back(layout.size);
allocate_arguments.parameters.push_back(object_map);
allocate_arguments.parameters.push_back(
    GenerateBoolConstant(expr->pretenured));
VisitResult allocate_result = GenerateCall(
    QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING}, "AllocateFromNew"),
    allocate_arguments, {class_type}, false);
DCHECK(allocate_result.IsOnStack())((void) 0);

InitializeClass(class_type, allocate_result, initializer_results, layout);

return stack_scope.Yield(GenerateCall(
    "%RawDownCast", Arguments{{allocate_result}, {}}, {class_type}));
1694}

1696const Type* ImplementationVisitor::Visit(BreakStatement* stmt) {
base::Optional<Binding<LocalLabel>*> break_label =
    TryLookupLabel(kBreakLabelName);
if (!break_label) {
  ReportError("break used outside of loop");
}
assembler().Goto((*break_label)->block);
return TypeOracle::GetNeverType();
1704}

1706const Type* ImplementationVisitor::Visit(ContinueStatement* stmt) {
base::Optional<Binding<LocalLabel>*> continue_label =
    TryLookupLabel(kContinueLabelName);
if (!continue_label) {
  ReportError("continue used outside of loop");
}
assembler().Goto((*continue_label)->block);
return TypeOracle::GetNeverType();
1714}

1716const Type* ImplementationVisitor::Visit(ForLoopStatement* stmt) {
BlockBindings<LocalValue> loop_bindings(&ValueBindingsManager::Get());

if (stmt->var_declaration) Visit(*stmt->var_declaration, &loop_bindings);

Block* body_block = assembler().NewBlock(assembler().CurrentStack());
Block* exit_block = assembler().NewBlock(assembler().CurrentStack());

Block* header_block = assembler().NewBlock();
assembler().Goto(header_block);
assembler().Bind(header_block);

// The continue label is where "continue" statements jump to. If no action
// expression is provided, we jump directly to the header.
Block* continue_block = header_block;

// The action label is only needed when an action expression was provided.
Block* action_block = nullptr;
if (stmt->action) {
  action_block = assembler().NewBlock();

  // The action expression needs to be executed on a continue.
  continue_block = action_block;
}

if (stmt->test) {
  GenerateExpressionBranch(*stmt->test, body_block, exit_block);
} else {
  assembler().Goto(body_block);
}

assembler().Bind(body_block);
{
  BreakContinueActivator activator(exit_block, continue_block);
  const Type* body_result = Visit(stmt->body);
  if (body_result != TypeOracle::GetNeverType()) {
    assembler().Goto(continue_block);
  }
}

if (stmt->action) {
  assembler().Bind(action_block);
  const Type* action_result = Visit(*stmt->action);
  if (action_result != TypeOracle::GetNeverType()) {
    assembler().Goto(header_block);
  }
}

assembler().Bind(exit_block);
return TypeOracle::GetVoidType();
1766}

1768VisitResult ImplementationVisitor::Visit(SpreadExpression* expr) {
ReportError(
    "spread operators are only currently supported in indexed class field "
    "initialization expressions");
1772}

1774void ImplementationVisitor::GenerateImplementation(const std::string& dir) {
for (SourceId file : SourceFileMap::AllSources()) {
  std::string base_filename =
      dir + "/" + SourceFileMap::PathFromV8RootWithoutExtension(file);
  GlobalContext::PerFileStreams& streams =
      GlobalContext::GeneratedPerFile(file);

  std::string csa_cc = streams.csa_ccfile.str();
  // Insert missing builtin includes where the marker is.
  {
    auto pos = csa_cc.find(BuiltinIncludesMarker);
    CHECK_NE(pos, std::string::npos)do { bool _cmp = ::v8::base::CmpNEImpl< typename ::v8::base
::pass_value_or_ref<decltype(pos)>::type, typename ::v8
::base::pass_value_or_ref<decltype(std::string::npos)>::
type>((pos), (std::string::npos)); do { if ((__builtin_expect
(!!(!(_cmp)), 0))) { V8_Fatal("Check failed: %s.", "pos" " " "!="
 " " "std::string::npos"); } } while (false); } while (false);
    std::string includes;
    for (const SourceId& include : streams.required_builtin_includes) {
      std::string include_file =
          SourceFileMap::PathFromV8RootWithoutExtension(include);
      includes += "#include \"torque-generated/";
      includes += include_file;
      includes += "-tq-csa.h\"\n";
    }
    csa_cc.replace(pos, strlen(BuiltinIncludesMarker), std::move(includes));
  }

  // TODO(torque-builder): Pass file directly.
  WriteFile(base_filename + "-tq-csa.cc", std::move(csa_cc));
  WriteFile(base_filename + "-tq-csa.h", streams.csa_headerfile.str());
  WriteFile(base_filename + "-tq.inc",
            streams.class_definition_headerfile.str());
  WriteFile(
      base_filename + "-tq-inl.inc",
      streams.class_definition_inline_headerfile_macro_declarations.str() +
          streams.class_definition_inline_headerfile_macro_definitions.str() +
          streams.class_definition_inline_headerfile.str());
  WriteFile(base_filename + "-tq.cc", streams.class_definition_ccfile.str());
}

WriteFile(dir + "/debug-macros.h", debug_macros_h_.str());
WriteFile(dir + "/debug-macros.cc", debug_macros_cc_.str());
1812}

1814cpp::Function ImplementationVisitor::GenerateMacroFunctionDeclaration(
  Macro* macro) {
return GenerateFunction(nullptr,
                        output_type_ == OutputType::kCC
                            ? macro->CCName()
                            : output_type_ == OutputType::kCCDebug
                                  ? macro->CCDebugName()
                                  : macro->ExternalName(),
                        macro->signature(), macro->parameter_names());
1823}

1825cpp::Function ImplementationVisitor::GenerateFunction(
  cpp::Class* owner, const std::string& name, const Signature& signature,
  const NameVector& parameter_names, bool pass_code_assembler_state,
  std::vector<std::string>* generated_parameter_names) {
cpp::Function f(owner, name);
f.SetInline(output_type_ == OutputType::kCC);

// Set return type.
// TODO(torque-builder): Consider an overload of SetReturnType that handles
// this.
if (signature.return_type->IsVoidOrNever()) {
  f.SetReturnType("void");
} else if (output_type_ == OutputType::kCCDebug) {
  f.SetReturnType(std::string("Value<") +
                  signature.return_type->GetDebugType() + ">");
} else if (output_type_ == OutputType::kCC) {
  f.SetReturnType(signature.return_type->GetRuntimeType());
} else {
  DCHECK_EQ(output_type_, OutputType::kCSA)((void) 0);
  f.SetReturnType(signature.return_type->GetGeneratedTypeName());
}

bool ignore_first_parameter = true;
if (output_type_ == OutputType::kCCDebug) {
  f.AddParameter("d::MemoryAccessor", "accessor");
} else if (output_type_ == OutputType::kCSA && pass_code_assembler_state) {
  f.AddParameter("compiler::CodeAssemblerState*", "state_");
} else {
  ignore_first_parameter = false;
}

// TODO(torque-builder): Consider an overload for AddParameter that handles
// this.
DCHECK_GE(signature.types().size(), parameter_names.size())((void) 0);
for (std::size_t i = 0; i < signature.types().size(); ++i) {
  const Type* parameter_type = signature.types()[i];
  std::string type;
  if (output_type_ == OutputType::kCC) {
    type = parameter_type->GetRuntimeType();
  } else if (output_type_ == OutputType::kCCDebug) {
    type = parameter_type->GetDebugType();
  } else {
    DCHECK_EQ(output_type_, OutputType::kCSA)((void) 0);
    type = parameter_type->GetGeneratedTypeName();
  }
  f.AddParameter(std::move(type),
                 ExternalParameterName(i < parameter_names.size()
                                           ? parameter_names[i]->value
                                           : std::to_string(i)));
}

for (const LabelDeclaration& label_info : signature.labels) {
  if (output_type_ == OutputType::kCC ||
      output_type_ == OutputType::kCCDebug) {
    ReportError("Macros that generate runtime code can't have label exits");
  }
  f.AddParameter("compiler::CodeAssemblerLabel*",
                 ExternalLabelName(label_info.name->value));
  size_t i = 0;
  for (const Type* type : label_info.types) {
    std::string generated_type_name;
    if (type->StructSupertype()) {
      generated_type_name = "\n#error no structs allowed in labels\n";
    } else {
      generated_type_name = "compiler::TypedCodeAssemblerVariable<";
      generated_type_name += type->GetGeneratedTNodeTypeName();
      generated_type_name += ">*";
    }
    f.AddParameter(generated_type_name,
                   ExternalLabelParameterName(label_info.name->value, i));
    ++i;
  }
}

if (generated_parameter_names) {
  *generated_parameter_names = f.GetParameterNames();
  if (ignore_first_parameter) {
    DCHECK(!generated_parameter_names->empty())((void) 0);
    generated_parameter_names->erase(generated_parameter_names->begin());
  }
}
return f;
1907}

1909namespace {

1911void FailCallableLookup(
  const std::string& reason, const QualifiedName& name,
  const TypeVector& parameter_types,
  const std::vector<Binding<LocalLabel>*>& labels,
  const std::vector<Signature>& candidates,
  const std::vector<std::pair<GenericCallable*, std::string>>
      inapplicable_generics) {
std::stringstream stream;
stream << "\n" << reason << ": \n  " << name << "(" << parameter_types << ")";
if (labels.size() != 0) {
  stream << " labels ";
  for (size_t i = 0; i < labels.size(); ++i) {
    stream << labels[i]->name() << "(" << labels[i]->parameter_types << ")";
  }
}
stream << "\ncandidates are:";
for (const Signature& signature : candidates) {
  stream << "\n  " << name;
  PrintSignature(stream, signature, false);
}
if (inapplicable_generics.size() != 0) {
  stream << "\nfailed to instantiate all of these generic declarations:";
  for (auto& failure : inapplicable_generics) {
    GenericCallable* generic = failure.first;
    const std::string& fail_reason = failure.second;
    stream << "\n  " << generic->name() << " defined at "
           << PositionAsString(generic->Position()) << ":\n    "
           << fail_reason << "\n";
  }
}
ReportError(stream.str());
1942}

1944Callable* GetOrCreateSpecialization(
  const SpecializationKey<GenericCallable>& key) {
if (base::Optional<Callable*> specialization =
        key.generic->GetSpecialization(key.specialized_types)) {
  return *specialization;
}
return DeclarationVisitor::SpecializeImplicit(key);
1951}

1953}  // namespace

1955base::Optional<Binding<LocalValue>*> ImplementationVisitor::TryLookupLocalValue(
  const std::string& name) {
return ValueBindingsManager::Get().TryLookup(name);
1958}

1960base::Optional<Binding<LocalLabel>*> ImplementationVisitor::TryLookupLabel(
  const std::string& name) {
return LabelBindingsManager::Get().TryLookup(name);
1963}

1965Binding<LocalLabel>* ImplementationVisitor::LookupLabel(
  const std::string& name) {
base::Optional<Binding<LocalLabel>*> label = TryLookupLabel(name);
if (!label) ReportError("cannot find label ", name);
return *label;
1970}

1972Block* ImplementationVisitor::LookupSimpleLabel(const std::string& name) {
LocalLabel* label = LookupLabel(name);
if (!label->parameter_types.empty()) {
  ReportError("label ", name,
              "was expected to have no parameters, but has parameters (",
              label->parameter_types, ")");
}
return label->block;
1980}

1982// Try to lookup a callable with the provided argument types. Do not report
1983// an error if no matching callable was found, but return false instead.
1984// This is used to test the presence of overloaded field accessors.
1985bool ImplementationVisitor::TestLookupCallable(
  const QualifiedName& name, const TypeVector& parameter_types) {
return LookupCallable(name, Declarations::TryLookup(name), parameter_types,
                      {}, {}, true) != nullptr;
1989}

1991TypeArgumentInference ImplementationVisitor::InferSpecializationTypes(
  GenericCallable* generic, const TypeVector& explicit_specialization_types,
  const TypeVector& explicit_arguments) {
std::vector<base::Optional<const Type*>> all_arguments;
const ParameterList& parameters = generic->declaration()->parameters;
for (size_t i = 0; i < parameters.implicit_count; ++i) {
  base::Optional<Binding<LocalValue>*> val =
      TryLookupLocalValue(parameters.names[i]->value);
  all_arguments.push_back(
      val ? (*val)->GetLocationReference(*val).ReferencedType()
          : base::nullopt);
}
for (const Type* explicit_argument : explicit_arguments) {
  all_arguments.push_back(explicit_argument);
}
return generic->InferSpecializationTypes(explicit_specialization_types,
                                         all_arguments);
2008}

2010template <class Container>
2011Callable* ImplementationVisitor::LookupCallable(
  const QualifiedName& name, const Container& declaration_container,
  const TypeVector& parameter_types,
  const std::vector<Binding<LocalLabel>*>& labels,
  const TypeVector& specialization_types, bool silence_errors) {
Callable* result = nullptr;

std::vector<Declarable*> overloads;
std::vector<Signature> overload_signatures;
std::vector<std::pair<GenericCallable*, std::string>> inapplicable_generics;
for (auto* declarable : declaration_container) {
  if (GenericCallable* generic = GenericCallable::DynamicCast(declarable)) {
    TypeArgumentInference inference = InferSpecializationTypes(
        generic, specialization_types, parameter_types);
    if (inference.HasFailed()) {
      inapplicable_generics.push_back(
          std::make_pair(generic, inference.GetFailureReason()));
      continue;
    }
    overloads.push_back(generic);
    overload_signatures.push_back(
        DeclarationVisitor::MakeSpecializedSignature(
            SpecializationKey<GenericCallable>{generic,
                                               inference.GetResult()}));
  } else if (Callable* callable = Callable::DynamicCast(declarable)) {
    overloads.push_back(callable);
    overload_signatures.push_back(callable->signature());
  }
}
// Indices of candidates in overloads/overload_signatures.
std::vector<size_t> candidates;
for (size_t i = 0; i < overloads.size(); ++i) {
  const Signature& signature = overload_signatures[i];
  if (IsCompatibleSignature(signature, parameter_types, labels.size())) {
    candidates.push_back(i);
  }
}

if (overloads.empty() && inapplicable_generics.empty()) {
  if (silence_errors) return nullptr;
  std::stringstream stream;
  stream << "no matching declaration found for " << name;
  ReportError(stream.str());
} else if (candidates.empty()) {
  if (silence_errors) return nullptr;
  FailCallableLookup("cannot find suitable callable with name", name,
                     parameter_types, labels, overload_signatures,
                     inapplicable_generics);
}

auto is_better_candidate = [&](size_t a, size_t b) {
  return ParameterDifference(overload_signatures[a].GetExplicitTypes(),
                             parameter_types)
      .StrictlyBetterThan(ParameterDifference(
          overload_signatures[b].GetExplicitTypes(), parameter_types));
};

size_t best = *std::min_element(candidates.begin(), candidates.end(),
                                is_better_candidate);
// This check is contained in libstdc++'s std::min_element.
DCHECK(!is_better_candidate(best, best))((void) 0);
for (size_t candidate : candidates) {
  if (candidate != best && !is_better_candidate(best, candidate)) {
    std::vector<Signature> candidate_signatures;
    candidate_signatures.reserve(candidates.size());
    for (size_t i : candidates) {
      candidate_signatures.push_back(overload_signatures[i]);
    }
    FailCallableLookup("ambiguous callable ", name, parameter_types, labels,
                       candidate_signatures, inapplicable_generics);
  }
}

if (GenericCallable* generic =
        GenericCallable::DynamicCast(overloads[best])) {
  TypeArgumentInference inference = InferSpecializationTypes(
      generic, specialization_types, parameter_types);
  result = GetOrCreateSpecialization(
      SpecializationKey<GenericCallable>{generic, inference.GetResult()});
} else {
  result = Callable::cast(overloads[best]);
}

size_t caller_size = parameter_types.size();
size_t callee_size =
    result->signature().types().size() - result->signature().implicit_count;
if (caller_size != callee_size &&
    !result->signature().parameter_types.var_args) {
  std::stringstream stream;
  stream << "parameter count mismatch calling " << *result << " - expected "
         << std::to_string(callee_size) << ", found "
         << std::to_string(caller_size);
  ReportError(stream.str());
}

return result;
2107}

2109template <class Container>
2110Callable* ImplementationVisitor::LookupCallable(
  const QualifiedName& name, const Container& declaration_container,
  const Arguments& arguments, const TypeVector& specialization_types) {
return LookupCallable(name, declaration_container,
                      arguments.parameters.ComputeTypeVector(),
                      arguments.labels, specialization_types);
2116}

2118Method* ImplementationVisitor::LookupMethod(
  const std::string& name, const AggregateType* receiver_type,
  const Arguments& arguments, const TypeVector& specialization_types) {
TypeVector types(arguments.parameters.ComputeTypeVector());
types.insert(types.begin(), receiver_type);
return Method::cast(LookupCallable({{}, name}, receiver_type->Methods(name),
                                   types, arguments.labels,
                                   specialization_types));
2126}

2128const Type* ImplementationVisitor::GetCommonType(const Type* left,
                                               const Type* right) {
const Type* common_type;
if (IsAssignableFrom(left, right)) {
  common_type = left;
} else if (IsAssignableFrom(right, left)) {
  common_type = right;
} else {
  common_type = TypeOracle::GetUnionType(left, right);
}
common_type = common_type->NonConstexprVersion();
return common_type;
2140}

2142VisitResult ImplementationVisitor::GenerateCopy(const VisitResult& to_copy) {
if (to_copy.IsOnStack()) {
  return VisitResult(to_copy.type(),
                     assembler().Peek(to_copy.stack_range(), to_copy.type()));
}
return to_copy;
2148}

2150VisitResult ImplementationVisitor::Visit(StructExpression* expr) {
StackScope stack_scope(this);

auto& initializers = expr->initializers;
std::vector<VisitResult> values;
std::vector<const Type*> term_argument_types;
values.reserve(initializers.size());
term_argument_types.reserve(initializers.size());

// Compute values and types of all initializer arguments
for (const NameAndExpression& initializer : initializers) {
  VisitResult value = Visit(initializer.expression);
  values.push_back(value);
  term_argument_types.push_back(value.type());
}

// Compute and check struct type from given struct name and argument types
const Type* type = TypeVisitor::ComputeTypeForStructExpression(
    expr->type, term_argument_types);
if (const auto* struct_type = StructType::DynamicCast(type)) {
  CheckInitializersWellformed(struct_type->name(), struct_type->fields(),
                              initializers);

  // Implicitly convert values and thereby build the struct on the stack
  StackRange struct_range = assembler().TopRange(0);
  auto& fields = struct_type->fields();
  for (size_t i = 0; i < values.size(); i++) {
    values[i] =
        GenerateImplicitConvert(fields[i].name_and_type.type, values[i]);
    struct_range.Extend(values[i].stack_range());
  }

  return stack_scope.Yield(VisitResult(struct_type, struct_range));
} else {
  const auto* bitfield_struct_type = BitFieldStructType::cast(type);
  CheckInitializersWellformed(bitfield_struct_type->name(),
                              bitfield_struct_type->fields(), initializers);

  // Create a zero and cast it to the desired bitfield struct type.
  VisitResult result{TypeOracle::GetConstInt32Type(), "0"};
  result = GenerateImplicitConvert(TypeOracle::GetInt32Type(), result);
  result = GenerateCall("Unsigned", Arguments{{result}, {}}, {});
  result = GenerateCall("%RawDownCast", Arguments{{result}, {}},
                        {bitfield_struct_type});

  // Set each field in the result. If these fields are constexpr, then all of
  // this initialization will end up reduced to a single value during TurboFan
  // optimization.
  auto& fields = bitfield_struct_type->fields();
  for (size_t i = 0; i < values.size(); i++) {
    values[i] =
        GenerateImplicitConvert(fields[i].name_and_type.type, values[i]);
    result = GenerateSetBitField(bitfield_struct_type, fields[i], result,
                                 values[i], /*starts_as_zero=*/true);
  }

  return stack_scope.Yield(result);
}
2208}

2210VisitResult ImplementationVisitor::GenerateSetBitField(
  const Type* bitfield_struct_type, const BitField& bitfield,
  VisitResult bitfield_struct, VisitResult value, bool starts_as_zero) {
GenerateCopy(bitfield_struct);
GenerateCopy(value);
assembler().Emit(
    StoreBitFieldInstruction{bitfield_struct_type, bitfield, starts_as_zero});
return VisitResult(bitfield_struct_type, assembler().TopRange(1));
2218}

2220LocationReference ImplementationVisitor::GetLocationReference(
  Expression* location) {
switch (location->kind) {
  case AstNode::Kind::kIdentifierExpression:
    return GetLocationReference(static_cast<IdentifierExpression*>(location));
  case AstNode::Kind::kFieldAccessExpression:
    return GetLocationReference(
        static_cast<FieldAccessExpression*>(location));
  case AstNode::Kind::kElementAccessExpression:
    return GetLocationReference(
        static_cast<ElementAccessExpression*>(location));
  case AstNode::Kind::kDereferenceExpression:
    return GetLocationReference(
        static_cast<DereferenceExpression*>(location));
  default:
    return LocationReference::Temporary(Visit(location), "expression");
}
2237}

2239LocationReference ImplementationVisitor::GetLocationReference(
  FieldAccessExpression* expr) {
return GenerateFieldAccess(GetLocationReference(expr->object),
                           expr->field->value, false, expr->field->pos);
2243}

2245LocationReference ImplementationVisitor::GenerateFieldAccess(
  LocationReference reference, const std::string& fieldname,
  bool ignore_stuct_field_constness, base::Optional<SourcePosition> pos) {
if (reference.IsVariableAccess() &&
    reference.variable().type()->StructSupertype()) {
  const StructType* type = *reference.variable().type()->StructSupertype();
  const Field& field = type->LookupField(fieldname);
  if (GlobalContext::collect_language_server_data() && pos.has_value()) {
    LanguageServerData::AddDefinition(*pos, field.pos);
  }
  if (GlobalContext::collect_kythe_data() && pos.has_value()) {
    KytheData::AddClassFieldUse(*pos, &field);
  }
  if (field.const_qualified) {
    VisitResult t_value = ProjectStructField(reference.variable(), fieldname);
    return LocationReference::Temporary(
        t_value, "for constant field '" + field.name_and_type.name + "'");
  } else {
    return LocationReference::VariableAccess(
        ProjectStructField(reference.variable(), fieldname));
  }
}
if (reference.IsTemporary() &&
    reference.temporary().type()->StructSupertype()) {
  if (GlobalContext::collect_language_server_data() && pos.has_value()) {
    const StructType* type = *reference.temporary().type()->StructSupertype();
    const Field& field = type->LookupField(fieldname);
    LanguageServerData::AddDefinition(*pos, field.pos);
  }
  return LocationReference::Temporary(
      ProjectStructField(reference.temporary(), fieldname),
      reference.temporary_description());
}
if (base::Optional<const Type*> referenced_type =
        reference.ReferencedType()) {
  if ((*referenced_type)->IsBitFieldStructType()) {
    const BitFieldStructType* bitfield_struct =
        BitFieldStructType::cast(*referenced_type);
    const BitField& field = bitfield_struct->LookupField(fieldname);
    return LocationReference::BitFieldAccess(reference, field);
  }
  if (const auto type_wrapped_in_smi = Type::MatchUnaryGeneric(
          (*referenced_type), TypeOracle::GetSmiTaggedGeneric())) {
    const BitFieldStructType* bitfield_struct =
        BitFieldStructType::DynamicCast(*type_wrapped_in_smi);
    if (bitfield_struct == nullptr) {
      ReportError(
          "When a value of type SmiTagged<T> is used in a field access "
          "expression, T is expected to be a bitfield struct type. Instead, "
          "T "
          "is ",
          **type_wrapped_in_smi);
    }
    const BitField& field = bitfield_struct->LookupField(fieldname);
    return LocationReference::BitFieldAccess(reference, field);
  }
}
if (reference.IsHeapReference()) {
  VisitResult ref = reference.heap_reference();
  bool is_const;
  auto generic_type =
      TypeOracle::MatchReferenceGeneric(ref.type(), &is_const);
  if (!generic_type) {
    ReportError(
        "Left-hand side of field access expression is marked as a reference "
        "but is not of type Reference<...>. Found type: ",
        ref.type()->ToString());
  }
  if (auto struct_type = (*generic_type)->StructSupertype()) {
    const Field& field = (*struct_type)->LookupField(fieldname);
    // Update the Reference's type to refer to the field type within the
    // struct.
    ref.SetType(TypeOracle::GetReferenceType(
        field.name_and_type.type,
        is_const ||
            (field.const_qualified && !ignore_stuct_field_constness)));
    if (!field.offset.has_value()) {
      Error("accessing field with unknown offset").Throw();
    }
    if (*field.offset != 0) {
      // Copy the Reference struct up the stack and update the new copy's
      // |offset| value to point to the struct field.
      StackScope scope(this);
      ref = GenerateCopy(ref);
      VisitResult ref_offset = ProjectStructField(ref, "offset");
      VisitResult struct_offset{
          TypeOracle::GetIntPtrType()->ConstexprVersion(),
          std::to_string(*field.offset)};
      VisitResult updated_offset =
          GenerateCall("+", Arguments{{ref_offset, struct_offset}, {}});
      assembler().Poke(ref_offset.stack_range(), updated_offset.stack_range(),
                       ref_offset.type());
      ref = scope.Yield(ref);
    }
    return LocationReference::HeapReference(ref);
  }
}
VisitResult object_result = GenerateFetchFromLocation(reference);
if (base::Optional<const ClassType*> class_type =
        object_result.type()->ClassSupertype()) {
  // This is a hack to distinguish the situation where we want to use
  // overloaded field accessors from when we want to create a reference.
  bool has_explicit_overloads = TestLookupCallable(
      QualifiedName{"." + fieldname}, {object_result.type()});
  if ((*class_type)->HasField(fieldname) && !has_explicit_overloads) {
    const Field& field = (*class_type)->LookupField(fieldname);
    if (GlobalContext::collect_language_server_data() && pos.has_value()) {
      LanguageServerData::AddDefinition(*pos, field.pos);
    }
    if (GlobalContext::collect_kythe_data()) {
      KytheData::AddClassFieldUse(*pos, &field);
    }
    return GenerateFieldReference(object_result, field, *class_type);
  }
}
return LocationReference::FieldAccess(object_result, fieldname);
2361}

2363LocationReference ImplementationVisitor::GetLocationReference(
  ElementAccessExpression* expr) {
LocationReference reference = GetLocationReference(expr->array);
VisitResult index = Visit(expr->index);
if (reference.IsHeapSlice()) {
  return GenerateReferenceToItemInHeapSlice(reference, index);
} else {
  return LocationReference::ArrayAccess(GenerateFetchFromLocation(reference),
                                        index);
}
2373}

2375LocationReference ImplementationVisitor::GenerateReferenceToItemInHeapSlice(
  LocationReference slice, VisitResult index) {
DCHECK(slice.IsHeapSlice())((void) 0);
Arguments arguments{{index}, {}};
const StructType* slice_type = *slice.heap_slice().type()->StructSupertype();
Method* method = LookupMethod("AtIndex", slice_type, arguments, {});
// The reference has to be treated like a normal value when calling methods
// on the underlying slice implementation.
LocationReference slice_value =
    LocationReference::Temporary(slice.GetVisitResult(), "slice as value");
return LocationReference::HeapReference(
    GenerateCall(method, std::move(slice_value), arguments, {}, false));
2387}

2389LocationReference ImplementationVisitor::GetLocationReference(
  IdentifierExpression* expr) {
if (expr->namespace_qualification.empty()) {
  if (base::Optional<Binding<LocalValue>*> value =
          TryLookupLocalValue(expr->name->value)) {
    if (GlobalContext::collect_language_server_data()) {
      LanguageServerData::AddDefinition(expr->name->pos,
                                        (*value)->declaration_position());
    }
    if (GlobalContext::collect_kythe_data()) {
      if (!expr->IsThis()) {
        DCHECK_EQ(expr->name->pos.end.column - expr->name->pos.start.column,((void) 0)
                  expr->name->value.length())((void) 0);
        KytheData::AddBindingUse(expr->name->pos, *value);
      }
    }
    if (expr->generic_arguments.size() != 0) {
      ReportError("cannot have generic parameters on local name ",
                  expr->name);
    }
    return (*value)->GetLocationReference(*value);
  }
}

if (expr->IsThis()) {
  ReportError("\"this\" cannot be qualified");
}
QualifiedName name =
    QualifiedName(expr->namespace_qualification, expr->name->value);
if (base::Optional<Builtin*> builtin = Declarations::TryLookupBuiltin(name)) {
  if (GlobalContext::collect_language_server_data()) {
    LanguageServerData::AddDefinition(expr->name->pos,
                                      (*builtin)->Position());
  }
  // TODO(v8:12261): Consider collecting KytheData here.
  return LocationReference::Temporary(GetBuiltinCode(*builtin),
                                      "builtin " + expr->name->value);
}
if (expr->generic_arguments.size() != 0) {
  GenericCallable* generic = Declarations::LookupUniqueGeneric(name);
  Callable* specialization =
      GetOrCreateSpecialization(SpecializationKey<GenericCallable>{
          generic, TypeVisitor::ComputeTypeVector(expr->generic_arguments)});
  if (Builtin* builtin = Builtin::DynamicCast(specialization)) {
    DCHECK(!builtin->IsExternal())((void) 0);
    return LocationReference::Temporary(GetBuiltinCode(builtin),
                                        "builtin " + expr->name->value);
  } else {
    ReportError("cannot create function pointer for non-builtin ",
                generic->name());
  }
}
Value* value = Declarations::LookupValue(name);
CHECK(value->Position().source.IsValid())do { if ((__builtin_expect(!!(!(value->Position().source.IsValid
())), 0))) { V8_Fatal("Check failed: %s.", "value->Position().source.IsValid()"
); } } while (false);
if (auto stream = CurrentFileStreams::Get()) {
  stream->required_builtin_includes.insert(value->Position().source);
}
if (GlobalContext::collect_language_server_data()) {
  LanguageServerData::AddDefinition(expr->name->pos, value->name()->pos);
}
if (auto* constant = NamespaceConstant::DynamicCast(value)) {
  if (GlobalContext::collect_kythe_data()) {
    KytheData::AddConstantUse(expr->name->pos, constant);
  }
  if (constant->type()->IsConstexpr()) {
    return LocationReference::Temporary(
        VisitResult(constant->type(), constant->external_name() + "(state_)"),
        "namespace constant " + expr->name->value);
  }
  assembler().Emit(NamespaceConstantInstruction{constant});
  StackRange stack_range =
      assembler().TopRange(LoweredSlotCount(constant->type()));
  return LocationReference::Temporary(
      VisitResult(constant->type(), stack_range),
      "namespace constant " + expr->name->value);
}
ExternConstant* constant = ExternConstant::cast(value);
if (GlobalContext::collect_kythe_data()) {
  KytheData::AddConstantUse(expr->name->pos, constant);
}
return LocationReference::Temporary(constant->value(),
                                    "extern value " + expr->name->value);
2471}

2473LocationReference ImplementationVisitor::GetLocationReference(
  DereferenceExpression* expr) {
VisitResult ref = Visit(expr->reference);
if (!TypeOracle::MatchReferenceGeneric(ref.type())) {
  Error("Operator * expects a reference type but found a value of type ",
        *ref.type())
      .Throw();
}
return LocationReference::HeapReference(ref);
2482}

2484VisitResult ImplementationVisitor::GenerateFetchFromLocation(
  const LocationReference& reference) {
if (reference.IsTemporary()) {
  return GenerateCopy(reference.temporary());
} else if (reference.IsVariableAccess()) {
  return GenerateCopy(reference.variable());
} else if (reference.IsHeapReference()) {
  const Type* referenced_type = *reference.ReferencedType();
  if (referenced_type == TypeOracle::GetFloat64OrHoleType()) {
    return GenerateCall(QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING},
                                      "LoadFloat64OrHole"),
                        Arguments{{reference.heap_reference()}, {}});
  } else if (auto struct_type = referenced_type->StructSupertype()) {
    StackRange result_range = assembler().TopRange(0);
    for (const Field& field : (*struct_type)->fields()) {
      StackScope scope(this);
      const std::string& fieldname = field.name_and_type.name;
      VisitResult field_value = scope.Yield(GenerateFetchFromLocation(
          GenerateFieldAccess(reference, fieldname)));
      result_range.Extend(field_value.stack_range());
    }
    return VisitResult(referenced_type, result_range);
  } else {
    GenerateCopy(reference.heap_reference());
    assembler().Emit(LoadReferenceInstruction{referenced_type});
    DCHECK_EQ(1, LoweredSlotCount(referenced_type))((void) 0);
    return VisitResult(referenced_type, assembler().TopRange(1));
  }
} else if (reference.IsBitFieldAccess()) {
  // First fetch the bitfield struct, then get the bits out of it.
  VisitResult bit_field_struct =
      GenerateFetchFromLocation(reference.bit_field_struct_location());
  assembler().Emit(LoadBitFieldInstruction{bit_field_struct.type(),
                                           reference.bit_field()});
  return VisitResult(*reference.ReferencedType(), assembler().TopRange(1));
} else {
  if (reference.IsHeapSlice()) {
    ReportError(
        "fetching a value directly from an indexed field isn't allowed");
  }
  DCHECK(reference.IsCallAccess())((void) 0);
  return GenerateCall(reference.eval_function(),
                      Arguments{reference.call_arguments(), {}});
}
2528}

2530void ImplementationVisitor::GenerateAssignToLocation(
  const LocationReference& reference, const VisitResult& assignment_value) {
if (reference.IsCallAccess()) {
  Arguments arguments{reference.call_arguments(), {}};
  arguments.parameters.push_back(assignment_value);
  GenerateCall(reference.assign_function(), arguments);
} else if (reference.IsVariableAccess()) {
  VisitResult variable = reference.variable();
  VisitResult converted_value =
      GenerateImplicitConvert(variable.type(), assignment_value);
  assembler().Poke(variable.stack_range(), converted_value.stack_range(),
                   variable.type());

  // Local variables are detected by the existence of a binding. Assignment
  // to local variables is recorded to support lint errors.
  if (reference.binding()) {
    (*reference.binding())->SetWritten();
  }
} else if (reference.IsHeapSlice()) {
  ReportError("assigning a value directly to an indexed field isn't allowed");
} else if (reference.IsHeapReference()) {
  const Type* referenced_type = *reference.ReferencedType();
  if (reference.IsConst()) {
    Error("cannot assign to const value of type ", *referenced_type).Throw();
  }
  if (referenced_type == TypeOracle::GetFloat64OrHoleType()) {
    GenerateCall(
        QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING},
                      "StoreFloat64OrHole"),
        Arguments{{reference.heap_reference(), assignment_value}, {}});
  } else if (auto struct_type = referenced_type->StructSupertype()) {
    if (!assignment_value.type()->IsSubtypeOf(referenced_type)) {
      ReportError("Cannot assign to ", *referenced_type,
                  " with value of type ", *assignment_value.type());
    }
    for (const Field& field : (*struct_type)->fields()) {
      const std::string& fieldname = field.name_and_type.name;
      // Allow assignment of structs even if they contain const fields.
      // Const on struct fields just disallows direct writes to them.
      bool ignore_stuct_field_constness = true;
      GenerateAssignToLocation(
          GenerateFieldAccess(reference, fieldname,
                              ignore_stuct_field_constness),
          ProjectStructField(assignment_value, fieldname));
    }
  } else {
    GenerateCopy(reference.heap_reference());
    VisitResult converted_assignment_value =
        GenerateImplicitConvert(referenced_type, assignment_value);
    if (referenced_type == TypeOracle::GetFloat64Type()) {
      VisitResult silenced_float_value = GenerateCall(
          "Float64SilenceNaN", Arguments{{assignment_value}, {}});
      assembler().Poke(converted_assignment_value.stack_range(),
                       silenced_float_value.stack_range(), referenced_type);
    }
    assembler().Emit(StoreReferenceInstruction{referenced_type});
  }
} else if (reference.IsBitFieldAccess()) {
  // First fetch the bitfield struct, then set the updated bits, then store
  // it back to where we found it.
  VisitResult bit_field_struct =
      GenerateFetchFromLocation(reference.bit_field_struct_location());
  VisitResult converted_value =
      GenerateImplicitConvert(*reference.ReferencedType(), assignment_value);
  VisitResult updated_bit_field_struct =
      GenerateSetBitField(bit_field_struct.type(), reference.bit_field(),
                          bit_field_struct, converted_value);
  GenerateAssignToLocation(reference.bit_field_struct_location(),
                           updated_bit_field_struct);
} else {
  DCHECK(reference.IsTemporary())((void) 0);
  ReportError("cannot assign to const-bound or temporary ",
              reference.temporary_description());
}
2604}

2606VisitResult ImplementationVisitor::GeneratePointerCall(
  Expression* callee, const Arguments& arguments, bool is_tailcall) {
StackScope scope(this);
TypeVector parameter_types(arguments.parameters.ComputeTypeVector());
VisitResult callee_result = Visit(callee);
if (!callee_result.type()->IsBuiltinPointerType()) {
  std::stringstream stream;
  stream << "Expected a function pointer type but found "
         << *callee_result.type();
  ReportError(stream.str());
}
const BuiltinPointerType* type =
    BuiltinPointerType::cast(callee_result.type());

if (type->parameter_types().size() != parameter_types.size()) {
  std::stringstream stream;
  stream << "parameter count mismatch calling function pointer with Type: "
         << *type << " - expected "
         << std::to_string(type->parameter_types().size()) << ", found "
         << std::to_string(parameter_types.size());
  ReportError(stream.str());
}

ParameterTypes types{type->parameter_types(), false};
Signature sig;
sig.parameter_types = types;
if (!IsCompatibleSignature(sig, parameter_types, 0)) {
  std::stringstream stream;
  stream << "parameters do not match function pointer signature. Expected: ("
         << type->parameter_types() << ") but got: (" << parameter_types
         << ")";
  ReportError(stream.str());
}

callee_result = GenerateCopy(callee_result);
StackRange arg_range = assembler().TopRange(0);
for (size_t current = 0; current < arguments.parameters.size(); ++current) {
  const Type* to_type = type->parameter_types()[current];
  arg_range.Extend(
      GenerateImplicitConvert(to_type, arguments.parameters[current])
          .stack_range());
}

assembler().Emit(
    CallBuiltinPointerInstruction{is_tailcall, type, arg_range.Size()});

if (is_tailcall) {
  return VisitResult::NeverResult();
}
DCHECK_EQ(1, LoweredSlotCount(type->return_type()))((void) 0);
return scope.Yield(VisitResult(type->return_type(), assembler().TopRange(1)));
2657}

2659void ImplementationVisitor::AddCallParameter(
  Callable* callable, VisitResult parameter, const Type* parameter_type,
  std::vector<VisitResult>* converted_arguments, StackRange* argument_range,
  std::vector<std::string>* constexpr_arguments, bool inline_macro) {
VisitResult converted;
if ((converted_arguments->size() < callable->signature().implicit_count) &&
    parameter.type()->IsTopType()) {
  converted = GenerateCopy(parameter);
} else {
  converted = GenerateImplicitConvert(parameter_type, parameter);
}
converted_arguments->push_back(converted);
if (!inline_macro) {
  if (converted.IsOnStack()) {
    argument_range->Extend(converted.stack_range());
  } else {
    constexpr_arguments->push_back(converted.constexpr_value());
  }
}
2678}

2680namespace {
2681std::pair<std::string, std::string> GetClassInstanceTypeRange(
  const ClassType* class_type) {
std::pair<std::string, std::string> result;
if (class_type->InstanceTypeRange()) {
  auto instance_type_range = *class_type->InstanceTypeRange();
  std::string instance_type_string_first =
      "static_cast<InstanceType>(" +
      std::to_string(instance_type_range.first) + ")";
  std::string instance_type_string_second =
      "static_cast<InstanceType>(" +
      std::to_string(instance_type_range.second) + ")";
  result =
      std::make_pair(instance_type_string_first, instance_type_string_second);
} else {
  ReportError(
      "%Min/MaxInstanceType must take a class type that is either a string "
      "or has a generated instance type range");
}
return result;
2700}
2701}  // namespace

2703VisitResult ImplementationVisitor::GenerateCall(
  Callable* callable, base::Optional<LocationReference> this_reference,
  Arguments arguments, const TypeVector& specialization_types,
  bool is_tailcall) {
CHECK(callable->Position().source.IsValid())do { if ((__builtin_expect(!!(!(callable->Position().source
.IsValid())), 0))) { V8_Fatal("Check failed: %s.", "callable->Position().source.IsValid()"
); } } while (false);
if (auto stream = CurrentFileStreams::Get()) {
  stream->required_builtin_includes.insert(callable->Position().source);
}

const Type* return_type = callable->signature().return_type;

if (is_tailcall) {
  if (Builtin* builtin = Builtin::DynamicCast(CurrentCallable::Get())) {
    const Type* outer_return_type = builtin->signature().return_type;
    if (!return_type->IsSubtypeOf(outer_return_type)) {
      Error("Cannot tailcall, type of result is ", *return_type,
            " but should be a subtype of ", *outer_return_type, ".");
    }
  } else {
    Error("Tail calls are only allowed from builtins");
  }
}

bool inline_macro = callable->ShouldBeInlined(output_type_);
std::vector<VisitResult> implicit_arguments;
for (size_t i = 0; i < callable->signature().implicit_count; ++i) {
  std::string implicit_name = callable->signature().parameter_names[i]->value;
  base::Optional<Binding<LocalValue>*> val =
      TryLookupLocalValue(implicit_name);
  if (val) {
    implicit_arguments.push_back(
        GenerateFetchFromLocation((*val)->GetLocationReference(*val)));
  } else {
    VisitResult unititialized = VisitResult::TopTypeResult(
        "implicit parameter '" + implicit_name +
            "' is not defined when invoking " + callable->ReadableName() +
            " at " + PositionAsString(CurrentSourcePosition::Get()),
        callable->signature().parameter_types.types[i]);
    implicit_arguments.push_back(unititialized);
  }
  const Type* type = implicit_arguments.back().type();
  if (const TopType* top_type = TopType::DynamicCast(type)) {
    if (!callable->IsMacro() || callable->IsExternal()) {
      ReportError(
          "unititialized implicit parameters can only be passed to "
          "Torque-defined macros: the ",
          top_type->reason());
    }
    inline_macro = true;
  }
}

std::vector<VisitResult> converted_arguments;
StackRange argument_range = assembler().TopRange(0);
std::vector<std::string> constexpr_arguments;

size_t current = 0;
for (; current < callable->signature().implicit_count; ++current) {
  AddCallParameter(callable, implicit_arguments[current],
                   callable->signature().parameter_types.types[current],
                   &converted_arguments, &argument_range,
                   &constexpr_arguments, inline_macro);
}

if (this_reference) {
  DCHECK(callable->IsMethod())((void) 0);
  Method* method = Method::cast(callable);
  // By now, the this reference should either be a variable, a temporary or
  // a Slice. In either case the fetch of the VisitResult should succeed.
  VisitResult this_value = this_reference->GetVisitResult();
  if (inline_macro) {
    if (!this_value.type()->IsSubtypeOf(method->aggregate_type())) {
      ReportError("this parameter must be a subtype of ",
                  *method->aggregate_type(), " but it is of type ",
                  *this_value.type());
    }
  } else {
    AddCallParameter(callable, this_value, method->aggregate_type(),
                     &converted_arguments, &argument_range,
                     &constexpr_arguments, inline_macro);
  }
  ++current;
}

for (auto arg : arguments.parameters) {
  const Type* to_type = (current >= callable->signature().types().size())
                            ? TypeOracle::GetObjectType()
                            : callable->signature().types()[current++];
  AddCallParameter(callable, arg, to_type, &converted_arguments,
                   &argument_range, &constexpr_arguments, inline_macro);
}

size_t label_count = callable->signature().labels.size();
if (label_count != arguments.labels.size()) {
  std::stringstream s;
  s << "unexpected number of otherwise labels for "
    << callable->ReadableName() << " (expected "
    << std::to_string(label_count) << " found "
    << std::to_string(arguments.labels.size()) << ")";
  ReportError(s.str());
}

if (callable->IsTransitioning()) {
  if (!CurrentCallable::Get()->IsTransitioning()) {
    std::stringstream s;
    s << *CurrentCallable::Get()
      << " isn't marked transitioning but calls the transitioning "
      << *callable;
    ReportError(s.str());
  }
}

if (auto* builtin = Builtin::DynamicCast(callable)) {
  base::Optional<Block*> catch_block = GetCatchBlock();
  assembler().Emit(CallBuiltinInstruction{
      is_tailcall, builtin, argument_range.Size(), catch_block});
  GenerateCatchBlock(catch_block);
  if (is_tailcall) {
    return VisitResult::NeverResult();
  } else {
    size_t slot_count = LoweredSlotCount(return_type);
    if (builtin->IsStub()) {
      if (slot_count < 1 || slot_count > 2) {
        ReportError(
            "Builtin with stub linkage is expected to return one or two "
            "values but returns ",
            slot_count);
      }
    } else {
      if (slot_count != 1) {
        ReportError(
            "Builtin with JS linkage is expected to return one value but "
            "returns ",
            slot_count);
      }
    }
    return VisitResult(return_type, assembler().TopRange(slot_count));
  }
} else if (auto* macro = Macro::DynamicCast(callable)) {
  if (is_tailcall) {
    ReportError("can't tail call a macro");
  }

  macro->SetUsed();

  // If we're currently generating a C++ macro and it's calling another macro,
  // then we need to make sure that we also generate C++ code for the called
  // macro within the same -inl.inc file.
  if ((output_type_ == OutputType::kCC ||
       output_type_ == OutputType::kCCDebug) &&
      !inline_macro) {
    if (auto* torque_macro = TorqueMacro::DynamicCast(macro)) {
      auto* streams = CurrentFileStreams::Get();
      SourceId file = streams ? streams->file : SourceId::Invalid();
      GlobalContext::EnsureInCCOutputList(torque_macro, file);
    }
  }

  // TODO(torque-builder): Consider a function builder here.
  if (return_type->IsConstexpr()) {
    DCHECK_EQ(0, arguments.labels.size())((void) 0);
    std::stringstream result;
    result << "(";
    bool first = true;
    switch (output_type_) {
      case OutputType::kCSA: {
        if (auto* extern_macro = ExternMacro::DynamicCast(macro)) {
          result << extern_macro->external_assembler_name() << "(state_)."
                 << extern_macro->ExternalName() << "(";
        } else {
          result << macro->ExternalName() << "(state_";
          first = false;
        }
        break;
      }
      case OutputType::kCC: {
        auto* extern_macro = ExternMacro::DynamicCast(macro);
        CHECK_NOT_NULL(extern_macro)do { if ((__builtin_expect(!!(!((extern_macro) != nullptr)), 0
))) { V8_Fatal("Check failed: %s.", "(extern_macro) != nullptr"
); } } while (false);
        result << extern_macro->CCName() << "(";
        break;
      }
      case OutputType::kCCDebug: {
        auto* extern_macro = ExternMacro::DynamicCast(macro);
        CHECK_NOT_NULL(extern_macro)do { if ((__builtin_expect(!!(!((extern_macro) != nullptr)), 0
))) { V8_Fatal("Check failed: %s.", "(extern_macro) != nullptr"
); } } while (false);
        result << extern_macro->CCDebugName() << "(accessor";
        first = false;
        break;
      }
    }
    for (VisitResult arg : converted_arguments) {
      DCHECK(!arg.IsOnStack())((void) 0);
      if (!first) {
        result << ", ";
      }
      first = false;
      result << arg.constexpr_value();
    }
    result << "))";
    return VisitResult(return_type, result.str());
  } else if (inline_macro) {
    std::vector<Block*> label_blocks;
    for (Binding<LocalLabel>* label : arguments.labels) {
      label_blocks.push_back(label->block);
    }
    return InlineMacro(macro, this_reference, converted_arguments,
                       label_blocks);
  } else if (arguments.labels.empty() &&
             return_type != TypeOracle::GetNeverType()) {
    base::Optional<Block*> catch_block = GetCatchBlock();
    assembler().Emit(
        CallCsaMacroInstruction{macro, constexpr_arguments, catch_block});
    GenerateCatchBlock(catch_block);
    size_t return_slot_count = LoweredSlotCount(return_type);
    return VisitResult(return_type, assembler().TopRange(return_slot_count));
  } else {
    base::Optional<Block*> return_continuation;
    if (return_type != TypeOracle::GetNeverType()) {
      return_continuation = assembler().NewBlock();
    }

    std::vector<Block*> label_blocks;

    for (size_t i = 0; i < label_count; ++i) {
      label_blocks.push_back(assembler().NewBlock());
    }
    base::Optional<Block*> catch_block = GetCatchBlock();
    assembler().Emit(CallCsaMacroAndBranchInstruction{
        macro, constexpr_arguments, return_continuation, label_blocks,
        catch_block});
    GenerateCatchBlock(catch_block);

    for (size_t i = 0; i < label_count; ++i) {
      Binding<LocalLabel>* label = arguments.labels[i];
      size_t callee_label_parameters =
          callable->signature().labels[i].types.size();
      if (label->parameter_types.size() != callee_label_parameters) {
        std::stringstream s;
        s << "label " << label->name()
          << " doesn't have the right number of parameters (found "
          << std::to_string(label->parameter_types.size()) << " expected "
          << std::to_string(callee_label_parameters) << ")";
        ReportError(s.str());
      }
      assembler().Bind(label_blocks[i]);
      assembler().Goto(
          label->block,
          LowerParameterTypes(callable->signature().labels[i].types).size());

      size_t j = 0;
      for (auto t : callable->signature().labels[i].types) {
        const Type* parameter_type = label->parameter_types[j];
        if (!t->IsSubtypeOf(parameter_type)) {
          ReportError("mismatch of label parameters (label expects ",
                      *parameter_type, " but macro produces ", *t,
                      " for parameter ", i + 1, ")");
        }
        j++;
      }
    }

    if (return_continuation) {
      assembler().Bind(*return_continuation);
      size_t return_slot_count = LoweredSlotCount(return_type);
      return VisitResult(return_type,
                         assembler().TopRange(return_slot_count));
    } else {
      return VisitResult::NeverResult();
    }
  }
} else if (auto* runtime_function = RuntimeFunction::DynamicCast(callable)) {
  base::Optional<Block*> catch_block = GetCatchBlock();
  assembler().Emit(CallRuntimeInstruction{
      is_tailcall, runtime_function, argument_range.Size(), catch_block});
  GenerateCatchBlock(catch_block);
  if (is_tailcall || return_type == TypeOracle::GetNeverType()) {
    return VisitResult::NeverResult();
  } else {
    size_t slot_count = LoweredSlotCount(return_type);
    DCHECK_LE(slot_count, 1)((void) 0);
    // TODO(turbofan): Actually, runtime functions have to return a value, so
    // we should assert slot_count == 1 here.
    return VisitResult(return_type, assembler().TopRange(slot_count));
  }
} else if (auto* intrinsic = Intrinsic::DynamicCast(callable)) {
  if (intrinsic->ExternalName() == "%SizeOf") {
    if (specialization_types.size() != 1) {
      ReportError("%SizeOf must take a single type parameter");
    }
    const Type* type = specialization_types[0];
    std::string size_string;
    if (base::Optional<std::tuple<size_t, std::string>> size = SizeOf(type)) {
      size_string = std::get<1>(*size);
    } else {
      Error("size of ", *type, " is not known.");
    }
    return VisitResult(return_type, size_string);
  } else if (intrinsic->ExternalName() == "%ClassHasMapConstant") {
    const Type* type = specialization_types[0];
    const ClassType* class_type = ClassType::DynamicCast(type);
    if (!class_type) {
      ReportError("%ClassHasMapConstant must take a class type parameter");
    }
    // If the class isn't actually used as the parameter to a TNode,
    // then we can't rely on the class existing in C++ or being of the same
    // type (e.g. it could be a template), so don't use the template CSA
    // machinery for accessing the class' map.
    if (class_type->name() != class_type->GetGeneratedTNodeTypeName()) {
      return VisitResult(return_type, std::string("false"));
    } else {
      return VisitResult(
          return_type,
          std::string("CodeStubAssembler(state_).ClassHasMapConstant<") +
              class_type->name() + ">()");
    }
  } else if (intrinsic->ExternalName() == "%MinInstanceType") {
    if (specialization_types.size() != 1) {
      ReportError("%MinInstanceType must take a single type parameter");
    }
    const Type* type = specialization_types[0];
    const ClassType* class_type = ClassType::DynamicCast(type);
    if (!class_type) {
      ReportError("%MinInstanceType must take a class type parameter");
    }
    std::pair<std::string, std::string> instance_types =
        GetClassInstanceTypeRange(class_type);
    return VisitResult(return_type, instance_types.first);
  } else if (intrinsic->ExternalName() == "%MaxInstanceType") {
    if (specialization_types.size() != 1) {
      ReportError("%MaxInstanceType must take a single type parameter");
    }
    const Type* type = specialization_types[0];
    const ClassType* class_type = ClassType::DynamicCast(type);
    if (!class_type) {
      ReportError("%MaxInstanceType must take a class type parameter");
    }
    std::pair<std::string, std::string> instance_types =
        GetClassInstanceTypeRange(class_type);
    return VisitResult(return_type, instance_types.second);
  } else if (intrinsic->ExternalName() == "%RawConstexprCast") {
    if (intrinsic->signature().parameter_types.types.size() != 1 ||
        constexpr_arguments.size() != 1) {
      ReportError(
          "%RawConstexprCast must take a single parameter with constexpr "
          "type");
    }
    if (!return_type->IsConstexpr()) {
      std::stringstream s;
      s << *return_type
        << " return type for %RawConstexprCast is not constexpr";
      ReportError(s.str());
    }
    std::stringstream result;
    result << "static_cast<" << return_type->GetGeneratedTypeName() << ">(";
    result << constexpr_arguments[0];
    result << ")";
    return VisitResult(return_type, result.str());
  } else if (intrinsic->ExternalName() == "%IndexedFieldLength") {
    const Type* type = specialization_types[0];
    const ClassType* class_type = ClassType::DynamicCast(type);
    if (!class_type) {
      ReportError("%IndexedFieldLength must take a class type parameter");
    }
    const Field& field =
        class_type->LookupField(StringLiteralUnquote(constexpr_arguments[0]));
    return GenerateArrayLength(VisitResult(type, argument_range), field);
  } else if (intrinsic->ExternalName() == "%MakeLazy") {
    if (specialization_types[0]->IsStructType()) {
      ReportError("%MakeLazy can't use macros that return structs");
    }
    std::string getter_name = StringLiteralUnquote(constexpr_arguments[0]);

    // Normally the parser would split namespace names for us, but we
    // sidestepped it by putting the macro name in a string literal.
    QualifiedName qualified_getter_name = QualifiedName::Parse(getter_name);

    // converted_arguments contains all of the arguments to %MakeLazy. We're
    // looking for a function that takes all but the first.
    Arguments arguments_to_getter;
    arguments_to_getter.parameters.insert(
        arguments_to_getter.parameters.begin(),
        converted_arguments.begin() + 1, converted_arguments.end());

    Callable* callable_macro = LookupCallable(
        qualified_getter_name, Declarations::Lookup(qualified_getter_name),
        arguments_to_getter, {});
    Macro* getter = Macro::DynamicCast(callable_macro);
    if (!getter || getter->IsMethod()) {
      ReportError(
          "%MakeLazy expects a macro, not builtin or other type of callable");
    }
    if (!getter->signature().labels.empty()) {
      ReportError("%MakeLazy requires a macro with no labels");
    }
    if (!getter->signature().return_type->IsSubtypeOf(
            specialization_types[0])) {
      ReportError("%MakeLazy expected return type ", *specialization_types[0],
                  " but found ", *getter->signature().return_type);
    }
    if (getter->signature().implicit_count > 0) {
      ReportError("Implicit parameters are not yet supported in %MakeLazy");
    }

    getter->SetUsed();  // Prevent warnings about unused macros.

    // Now that we've looked up the getter macro, we have to convert the
    // arguments again, so that, for example, constexpr arguments can be
    // coerced to non-constexpr types and put on the stack.

    std::vector<VisitResult> converted_arguments_for_getter;
    StackRange argument_range_for_getter = assembler().TopRange(0);
    std::vector<std::string> constexpr_arguments_for_getter;

    size_t arg_count = 0;
    for (auto arg : arguments_to_getter.parameters) {
      DCHECK_LT(arg_count, getter->signature().types().size())((void) 0);
      const Type* to_type = getter->signature().types()[arg_count++];
      AddCallParameter(getter, arg, to_type, &converted_arguments_for_getter,
                       &argument_range_for_getter,
                       &constexpr_arguments_for_getter,
                       /*inline_macro=*/false);
    }

    // Now that the arguments are prepared, emit the instruction that consumes
    // them.
    assembler().Emit(MakeLazyNodeInstruction{getter, return_type,
                                             constexpr_arguments_for_getter});
    return VisitResult(return_type, assembler().TopRange(1));
  } else if (intrinsic->ExternalName() == "%FieldSlice") {
    const Type* type = specialization_types[0];
    const ClassType* class_type = ClassType::DynamicCast(type);
    if (!class_type) {
      ReportError("The first type parameter to %FieldSlice must be a class");
    }
    const Field& field =
        class_type->LookupField(StringLiteralUnquote(constexpr_arguments[0]));
    const Type* expected_slice_type =
        field.const_qualified
            ? TypeOracle::GetConstSliceType(field.name_and_type.type)
            : TypeOracle::GetMutableSliceType(field.name_and_type.type);
    const Type* declared_slice_type = specialization_types[1];
    if (expected_slice_type != declared_slice_type) {
      Error(
          "The second type parameter to %FieldSlice must be the precise "
          "slice type for the named field");
    }
    LocationReference ref = GenerateFieldReference(
        VisitResult(type, argument_range), field, class_type,
        /*treat_optional_as_indexed=*/true);
    if (!ref.IsHeapSlice()) {
      ReportError("%FieldSlice expected an indexed or optional field");
    }
    return ref.heap_slice();
  } else {
    assembler().Emit(CallIntrinsicInstruction{intrinsic, specialization_types,
                                              constexpr_arguments});
    size_t return_slot_count =
        LoweredSlotCount(intrinsic->signature().return_type);
    return VisitResult(return_type, assembler().TopRange(return_slot_count));
  }
} else {
  UNREACHABLE()V8_Fatal("unreachable code");
}
3165}

3167VisitResult ImplementationVisitor::GenerateCall(
  const QualifiedName& callable_name, Arguments arguments,
  const TypeVector& specialization_types, bool is_tailcall) {
Callable* callable =
    LookupCallable(callable_name, Declarations::Lookup(callable_name),
                   arguments, specialization_types);
return GenerateCall(callable, base::nullopt, arguments, specialization_types,
                    is_tailcall);
3175}

3177VisitResult ImplementationVisitor::Visit(CallExpression* expr,
                                       bool is_tailcall) {
StackScope scope(this);

if (expr->callee->name->value == "&" && expr->arguments.size() == 1) {
  if (auto* loc_expr = LocationExpression::DynamicCast(expr->arguments[0])) {
    LocationReference ref = GetLocationReference(loc_expr);
    if (ref.IsHeapReference()) return scope.Yield(ref.heap_reference());
    if (ref.IsHeapSlice()) return scope.Yield(ref.heap_slice());
  }
  ReportError("Unable to create a heap reference.");
}

Arguments arguments;
QualifiedName name = QualifiedName(expr->callee->namespace_qualification,
                                   expr->callee->name->value);
TypeVector specialization_types =
    TypeVisitor::ComputeTypeVector(expr->callee->generic_arguments);
bool has_template_arguments = !specialization_types.empty();
for (Expression* arg : expr->arguments)
  arguments.parameters.push_back(Visit(arg));
arguments.labels = LabelsFromIdentifiers(expr->labels);
if (!has_template_arguments && name.namespace_qualification.empty() &&
    TryLookupLocalValue(name.name)) {
  return scope.Yield(
      GeneratePointerCall(expr->callee, arguments, is_tailcall));
} else {
  if (GlobalContext::collect_language_server_data()) {
    Callable* callable = LookupCallable(name, Declarations::Lookup(name),
                                        arguments, specialization_types);
    LanguageServerData::AddDefinition(expr->callee->name->pos,
                                      callable->IdentifierPosition());
  }
  if (GlobalContext::collect_kythe_data()) {
    Callable* callable = LookupCallable(name, Declarations::Lookup(name),
                                        arguments, specialization_types);
    Callable* caller = CurrentCallable::Get();
    KytheData::AddCall(caller, expr->callee->name->pos, callable);
  }
  if (expr->callee->name->value == "!" && arguments.parameters.size() == 1) {
    PropagateBitfieldMark(expr->arguments[0], expr);
  }
  if (expr->callee->name->value == "==" && arguments.parameters.size() == 2) {
    if (arguments.parameters[0].type()->IsConstexpr()) {
      PropagateBitfieldMark(expr->arguments[1], expr);
    } else if (arguments.parameters[1].type()->IsConstexpr()) {
      PropagateBitfieldMark(expr->arguments[0], expr);
    }
  }
  return scope.Yield(
      GenerateCall(name, arguments, specialization_types, is_tailcall));
}
3229}

3231VisitResult ImplementationVisitor::Visit(CallMethodExpression* expr) {
StackScope scope(this);
Arguments arguments;
std::string method_name = expr->method->name->value;
TypeVector specialization_types =
    TypeVisitor::ComputeTypeVector(expr->method->generic_arguments);
LocationReference target = GetLocationReference(expr->target);
if (!target.IsVariableAccess()) {
  VisitResult result = GenerateFetchFromLocation(target);
  target = LocationReference::Temporary(result, "this parameter");
}
const AggregateType* target_type =
    (*target.ReferencedType())->AggregateSupertype().value_or(nullptr);
if (!target_type) {
  ReportError("target of method call not a struct or class type");
}
for (Expression* arg : expr->arguments) {
  arguments.parameters.push_back(Visit(arg));
}
arguments.labels = LabelsFromIdentifiers(expr->labels);
TypeVector argument_types = arguments.parameters.ComputeTypeVector();
DCHECK_EQ(expr->method->namespace_qualification.size(), 0)((void) 0);
QualifiedName qualified_name = QualifiedName(method_name);
Callable* callable = LookupMethod(method_name, target_type, arguments, {});
if (GlobalContext::collect_language_server_data()) {
  LanguageServerData::AddDefinition(expr->method->name->pos,
                                    callable->IdentifierPosition());
}
if (GlobalContext::collect_kythe_data()) {
  Callable* caller = CurrentCallable::Get();
  KytheData::AddCall(caller, expr->method->name->pos, callable);
}
return scope.Yield(GenerateCall(callable, target, arguments, {}, false));
3264}

3266VisitResult ImplementationVisitor::Visit(IntrinsicCallExpression* expr) {
StackScope scope(this);
Arguments arguments;
TypeVector specialization_types =
    TypeVisitor::ComputeTypeVector(expr->generic_arguments);
for (Expression* arg : expr->arguments)
  arguments.parameters.push_back(Visit(arg));
return scope.Yield(
    GenerateCall(expr->name->value, arguments, specialization_types, false));
3275}

3277void ImplementationVisitor::GenerateBranch(const VisitResult& condition,
                                         Block* true_block,
                                         Block* false_block) {
DCHECK_EQ(condition,((void) 0)
          VisitResult(TypeOracle::GetBoolType(), assembler().TopRange(1)))((void) 0);
assembler().Branch(true_block, false_block);
3283}

3285VisitResult ImplementationVisitor::GenerateBoolConstant(bool constant) {
return GenerateImplicitConvert(TypeOracle::GetBoolType(),
                               VisitResult(TypeOracle::GetConstexprBoolType(),
                                           constant ? "true" : "false"));
3289}

3291void ImplementationVisitor::GenerateExpressionBranch(Expression* expression,
                                                   Block* true_block,
                                                   Block* false_block) {
StackScope stack_scope(this);
VisitResult expression_result = this->Visit(expression);
expression_result = stack_scope.Yield(
    GenerateImplicitConvert(TypeOracle::GetBoolType(), expression_result));
GenerateBranch(expression_result, true_block, false_block);
3299}

3301VisitResult ImplementationVisitor::GenerateImplicitConvert(
  const Type* destination_type, VisitResult source) {
StackScope scope(this);
if (source.type() == TypeOracle::GetNeverType()) {
  ReportError("it is not allowed to use a value of type never");
}

if (destination_type == source.type()) {
  return scope.Yield(GenerateCopy(source));
}

if (auto from = TypeOracle::ImplicitlyConvertableFrom(destination_type,
                                                      source.type())) {
  return scope.Yield(GenerateCall(kFromConstexprMacroName,
                                  Arguments{{source}, {}},
                                  {destination_type, *from}, false));
} else if (IsAssignableFrom(destination_type, source.type())) {
  source.SetType(destination_type);
  return scope.Yield(GenerateCopy(source));
} else {
  std::stringstream s;
  if (const TopType* top_type = TopType::DynamicCast(source.type())) {
    s << "undefined expression of type " << *destination_type << ": the "
      << top_type->reason();
  } else {
    s << "cannot use expression of type " << *source.type()
      << " as a value of type " << *destination_type;
  }
  ReportError(s.str());
}
3331}

3333StackRange ImplementationVisitor::GenerateLabelGoto(
  LocalLabel* label, base::Optional<StackRange> arguments) {
return assembler().Goto(label->block, arguments ? arguments->Size() : 0);
3336}

3338std::vector<Binding<LocalLabel>*> ImplementationVisitor::LabelsFromIdentifiers(
  const std::vector<Identifier*>& names) {
std::vector<Binding<LocalLabel>*> result;
result.reserve(names.size());
for (const auto& name : names) {
  Binding<LocalLabel>* label = LookupLabel(name->value);
  result.push_back(label);

  // Link up labels in "otherwise" part of the call expression with
  // either the label in the signature of the calling macro or the label
  // block ofa surrounding "try".
  if (GlobalContext::collect_language_server_data()) {
    LanguageServerData::AddDefinition(name->pos,
                                      label->declaration_position());
  }
  // TODO(v8:12261): Might have to track KytheData here.
}
return result;
3356}

3358StackRange ImplementationVisitor::LowerParameter(
  const Type* type, const std::string& parameter_name,
  Stack<std::string>* lowered_parameters) {
if (base::Optional<const StructType*> struct_type = type->StructSupertype()) {
  StackRange range = lowered_parameters->TopRange(0);
  for (auto& field : (*struct_type)->fields()) {
    StackRange parameter_range = LowerParameter(
        field.name_and_type.type,
        parameter_name + "." + field.name_and_type.name, lowered_parameters);
    range.Extend(parameter_range);
  }
  return range;
} else {
  lowered_parameters->Push(parameter_name);
  return lowered_parameters->TopRange(1);
}
3374}

3376void ImplementationVisitor::LowerLabelParameter(
  const Type* type, const std::string& parameter_name,
  std::vector<std::string>* lowered_parameters) {
if (base::Optional<const StructType*> struct_type = type->StructSupertype()) {
  for (auto& field : (*struct_type)->fields()) {
    LowerLabelParameter(
        field.name_and_type.type,
        "&((*" + parameter_name + ")." + field.name_and_type.name + ")",
        lowered_parameters);
  }
} else {
  lowered_parameters->push_back(parameter_name);
}
3389}

3391std::string ImplementationVisitor::ExternalLabelName(
  const std::string& label_name) {
return "label_" + label_name;
3394}

3396std::string ImplementationVisitor::ExternalLabelParameterName(
  const std::string& label_name, size_t i) {
return "label_" + label_name + "_parameter_" + std::to_string(i);
3399}

3401std::string ImplementationVisitor::ExternalParameterName(
  const std::string& name) {
return std::string("p_") + name;
3404}

3406DEFINE_CONTEXTUAL_VARIABLE(ImplementationVisitor::ValueBindingsManager)template <> ImplementationVisitor::ValueBindingsManager
::Scope*& ContextualVariableTop<ImplementationVisitor::
ValueBindingsManager>() { static thread_local ImplementationVisitor
::ValueBindingsManager::Scope* top = nullptr; return top; }
3407DEFINE_CONTEXTUAL_VARIABLE(ImplementationVisitor::LabelBindingsManager)template <> ImplementationVisitor::LabelBindingsManager
::Scope*& ContextualVariableTop<ImplementationVisitor::
LabelBindingsManager>() { static thread_local ImplementationVisitor
::LabelBindingsManager::Scope* top = nullptr; return top; }
3408DEFINE_CONTEXTUAL_VARIABLE(ImplementationVisitor::CurrentCallable)template <> ImplementationVisitor::CurrentCallable::Scope
*& ContextualVariableTop<ImplementationVisitor::CurrentCallable
>() { static thread_local ImplementationVisitor::CurrentCallable
::Scope* top = nullptr; return top; }
3409DEFINE_CONTEXTUAL_VARIABLE(ImplementationVisitor::CurrentFileStreams)template <> ImplementationVisitor::CurrentFileStreams::
Scope*& ContextualVariableTop<ImplementationVisitor::CurrentFileStreams
>() { static thread_local ImplementationVisitor::CurrentFileStreams
::Scope* top = nullptr; return top; }
3410DEFINE_CONTEXTUAL_VARIABLE(ImplementationVisitor::CurrentReturnValue)template <> ImplementationVisitor::CurrentReturnValue::
Scope*& ContextualVariableTop<ImplementationVisitor::CurrentReturnValue
>() { static thread_local ImplementationVisitor::CurrentReturnValue
::Scope* top = nullptr; return top; }

3412bool IsCompatibleSignature(const Signature& sig, const TypeVector& types,
                         size_t label_count) {
auto i = sig.parameter_types.types.begin() + sig.implicit_count;
if ((sig.parameter_types.types.size() - sig.implicit_count) > types.size())
  return false;
if (sig.labels.size() != label_count) return false;
for (auto current : types) {
  if (i == sig.parameter_types.types.end()) {
    if (!sig.parameter_types.var_args) return false;
    if (!IsAssignableFrom(TypeOracle::GetObjectType(), current)) return false;
  } else {
    if (!IsAssignableFrom(*i++, current)) return false;
  }
}
return true;
3427}

3429base::Optional<Block*> ImplementationVisitor::GetCatchBlock() {
base::Optional<Block*> catch_block;
if (base::Optional<Binding<LocalLabel>*> catch_handler =
        TryLookupLabel(kCatchLabelName)) {
  catch_block = assembler().NewBlock(base::nullopt, true);
}
return catch_block;
3436}

3438void ImplementationVisitor::GenerateCatchBlock(
  base::Optional<Block*> catch_block) {
if (catch_block) {
  base::Optional<Binding<LocalLabel>*> catch_handler =
      TryLookupLabel(kCatchLabelName);
  // Reset the local scopes to prevent the macro calls below from using the
  // current catch handler.
  BindingsManagersScope bindings_managers_scope;
  if (assembler().CurrentBlockIsComplete()) {
    assembler().Bind(*catch_block);
    GenerateCall(QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING},
                               "GetAndResetPendingMessage"),
                 Arguments{{}, {}}, {}, false);
    assembler().Goto((*catch_handler)->block, 2);
  } else {
    CfgAssemblerScopedTemporaryBlock temp(&assembler(), *catch_block);
    GenerateCall(QualifiedName({TORQUE_INTERNAL_NAMESPACE_STRING},
                               "GetAndResetPendingMessage"),
                 Arguments{{}, {}}, {}, false);
    assembler().Goto((*catch_handler)->block, 2);
  }
}
3460}
3461void ImplementationVisitor::VisitAllDeclarables() {
CurrentCallable::Scope current_callable(nullptr);
const std::vector<std::unique_ptr<Declarable>>& all_declarables =
    GlobalContext::AllDeclarables();

// This has to be an index-based loop because all_declarables can be extended
// during the loop.
for (size_t i = 0; i < all_declarables.size(); ++i) {
  try {
    Visit(all_declarables[i].get());
  } catch (TorqueAbortCompilation&) {
    // Recover from compile errors here. The error is recorded already.
  }
}

// Do the same for macros which generate C++ code.
output_type_ = OutputType::kCC;
const std::vector<std::pair<TorqueMacro*, SourceId>>& cc_macros =
    GlobalContext::AllMacrosForCCOutput();
for (size_t i = 0; i < cc_macros.size(); ++i) {
  try {
    Visit(static_cast<Declarable*>(cc_macros[i].first), cc_macros[i].second);
  } catch (TorqueAbortCompilation&) {
    // Recover from compile errors here. The error is recorded already.
  }
}

// Do the same for macros which generate C++ debug code.
// The set of macros is the same as C++ macros.
output_type_ = OutputType::kCCDebug;
for (size_t i = 0; i < cc_macros.size(); ++i) {
  try {
    Visit(static_cast<Declarable*>(cc_macros[i].first), cc_macros[i].second);
  } catch (TorqueAbortCompilation&) {
    // Recover from compile errors here. The error is recorded already.
  }
}
output_type_ = OutputType::kCSA;
3499}

3501void ImplementationVisitor::Visit(Declarable* declarable,
                                base::Optional<SourceId> file) {
CurrentScope::Scope current_scope(declarable->ParentScope());
CurrentSourcePosition::Scope current_source_position(declarable->Position());
CurrentFileStreams::Scope current_file_streams(
    &GlobalContext::GeneratedPerFile(file ? *file
                                          : declarable->Position().source));
if (Callable* callable = Callable::DynamicCast(declarable)) {
  if (!callable->ShouldGenerateExternalCode(output_type_))
    CurrentFileStreams::Get() = nullptr;
}
switch (declarable->kind()) {
  case Declarable::kExternMacro:
    return Visit(ExternMacro::cast(declarable));
  case Declarable::kTorqueMacro:
    return Visit(TorqueMacro::cast(declarable));
  case Declarable::kMethod:
    return Visit(Method::cast(declarable));
  case Declarable::kBuiltin:
    return Visit(Builtin::cast(declarable));
  case Declarable::kTypeAlias:
    return Visit(TypeAlias::cast(declarable));
  case Declarable::kNamespaceConstant:
    return Visit(NamespaceConstant::cast(declarable));
  case Declarable::kRuntimeFunction:
  case Declarable::kIntrinsic:
  case Declarable::kExternConstant:
  case Declarable::kNamespace:
  case Declarable::kGenericCallable:
  case Declarable::kGenericType:
    return;
}
3533}

3535std::string MachineTypeString(const Type* type) {
if (type->IsSubtypeOf(TypeOracle::GetSmiType())) {
  return "MachineType::TaggedSigned()";
}
if (type->IsSubtypeOf(TypeOracle::GetHeapObjectType())) {
  return "MachineType::TaggedPointer()";
}
if (type->IsSubtypeOf(TypeOracle::GetTaggedType())) {
  return "MachineType::AnyTagged()";
}
return "MachineTypeOf<" + type->GetGeneratedTNodeTypeName() + ">::value";
3546}

3548void ImplementationVisitor::GenerateBuiltinDefinitionsAndInterfaceDescriptors(
  const std::string& output_directory) {
std::stringstream builtin_definitions;
std::string builtin_definitions_file_name = "builtin-definitions.h";

// This file contains plain interface descriptor definitions and has to be
// included in the middle of interface-descriptors.h. Thus it is not a normal
// header file and uses the .inc suffix instead of the .h suffix.
std::stringstream interface_descriptors;
std::string interface_descriptors_file_name = "interface-descriptors.inc";
{
  IncludeGuardScope builtin_definitions_include_guard(
      builtin_definitions, builtin_definitions_file_name);

  builtin_definitions
      << "\n"
         "#define BUILTIN_LIST_FROM_TORQUE(CPP, TFJ, TFC, TFS, TFH, "
         "ASM) "
         "\\\n";
  for (auto& declarable : GlobalContext::AllDeclarables()) {
    Builtin* builtin = Builtin::DynamicCast(declarable.get());
    if (!builtin || builtin->IsExternal()) continue;
    if (builtin->IsStub()) {
      builtin_definitions << "TFC(" << builtin->ExternalName() << ", "
                          << builtin->ExternalName();
      std::string descriptor_name = builtin->ExternalName() + "Descriptor";
      bool has_context_parameter = builtin->signature().HasContextParameter();
      size_t kFirstNonContextParameter = has_context_parameter ? 1 : 0;
      TypeVector return_types = LowerType(builtin->signature().return_type);

      interface_descriptors << "class " << descriptor_name
                            << " : public StaticCallInterfaceDescriptor<"
                            << descriptor_name << "> {\n";

      interface_descriptors << " public:\n";

      if (has_context_parameter) {
        interface_descriptors << "  DEFINE_RESULT_AND_PARAMETERS(";
      } else {
        interface_descriptors << "  DEFINE_RESULT_AND_PARAMETERS_NO_CONTEXT(";
      }
      interface_descriptors << return_types.size();
      for (size_t i = kFirstNonContextParameter;
           i < builtin->parameter_names().size(); ++i) {
        Identifier* parameter = builtin->parameter_names()[i];
        interface_descriptors << ", k" << CamelifyString(parameter->value);
      }
      interface_descriptors << ")\n";

      interface_descriptors << "  DEFINE_RESULT_AND_PARAMETER_TYPES(";
      PrintCommaSeparatedList(interface_descriptors, return_types,
                              MachineTypeString);
      for (size_t i = kFirstNonContextParameter;
           i < builtin->parameter_names().size(); ++i) {
        const Type* type = builtin->signature().parameter_types.types[i];
        interface_descriptors << ", " << MachineTypeString(type);
      }
      interface_descriptors << ")\n";

      interface_descriptors << "  DECLARE_DEFAULT_DESCRIPTOR("
                            << descriptor_name << ")\n";
      interface_descriptors << "};\n\n";
    } else {
      builtin_definitions << "TFJ(" << builtin->ExternalName();
      if (builtin->IsVarArgsJavaScript()) {
        builtin_definitions << ", kDontAdaptArgumentsSentinel";
      } else {
        DCHECK(builtin->IsFixedArgsJavaScript())((void) 0);
        // FixedArg javascript builtins need to offer the parameter
        // count.
        int parameter_count =
            static_cast<int>(builtin->signature().ExplicitCount());
        builtin_definitions << ", JSParameterCount(" << parameter_count
                            << ")";
        // And the receiver is explicitly declared.
        builtin_definitions << ", kReceiver";
        for (size_t i = builtin->signature().implicit_count;
             i < builtin->parameter_names().size(); ++i) {
          Identifier* parameter = builtin->parameter_names()[i];
          builtin_definitions << ", k" << CamelifyString(parameter->value);
        }
      }
    }
    builtin_definitions << ") \\\n";
  }
  builtin_definitions << "\n";

  builtin_definitions
      << "#define TORQUE_FUNCTION_POINTER_TYPE_TO_BUILTIN_MAP(V) \\\n";
  for (const BuiltinPointerType* type :
       TypeOracle::AllBuiltinPointerTypes()) {
    Builtin* example_builtin =
        Declarations::FindSomeInternalBuiltinWithType(type);
    if (!example_builtin) {
      CurrentSourcePosition::Scope current_source_position(
          SourcePosition{CurrentSourceFile::Get(), LineAndColumn::Invalid(),
                         LineAndColumn::Invalid()});
      ReportError("unable to find any builtin with type \"", *type, "\"");
    }
    builtin_definitions << "  V(" << type->function_pointer_type_id() << ","
                        << example_builtin->ExternalName() << ")\\\n";
  }
  builtin_definitions << "\n";
}
WriteFile(output_directory + "/" + builtin_definitions_file_name,
          builtin_definitions.str());
WriteFile(output_directory + "/" + interface_descriptors_file_name,
          interface_descriptors.str());
3656}

3658namespace {

3660enum class FieldSectionType : uint32_t {
kNoSection = 0,
kWeakSection = 1 << 0,
kStrongSection = 2 << 0,
kScalarSection = 3 << 0
3665};

3667bool IsPointerSection(FieldSectionType type) {
return type == FieldSectionType::kWeakSection ||
       type == FieldSectionType::kStrongSection;
3670}

3672using FieldSections = base::Flags<FieldSectionType>;

3674std::string ToString(FieldSectionType type) {
switch (type) {
  case FieldSectionType::kNoSection:
    return "NoSection";
  case FieldSectionType::kWeakSection:
    return "WeakFields";
  case FieldSectionType::kStrongSection:
    return "StrongFields";
  case FieldSectionType::kScalarSection:
    return "ScalarFields";
}
UNREACHABLE()V8_Fatal("unreachable code");
3686}

3688class FieldOffsetsGenerator {
public:
explicit FieldOffsetsGenerator(const ClassType* type) : type_(type) {}

virtual void WriteField(const Field& f, const std::string& size_string) = 0;
virtual void WriteFieldOffsetGetter(const Field& f) = 0;
virtual void WriteMarker(const std::string& marker) = 0;

virtual ~FieldOffsetsGenerator() { CHECK(is_finished_)do { if ((__builtin_expect(!!(!(is_finished_)), 0))) { V8_Fatal
("Check failed: %s.", "is_finished_"); } } while (false); }

void RecordOffsetFor(const Field& f) {
  CHECK(!is_finished_)do { if ((__builtin_expect(!!(!(!is_finished_)), 0))) { V8_Fatal
("Check failed: %s.", "!is_finished_"); } } while (false);
  UpdateSection(f);

  // Emit kHeaderSize before any indexed field.
  if (f.index.has_value() && !header_size_emitted_) {
    WriteMarker("kHeaderSize");
    header_size_emitted_ = true;
  }

  // We don't know statically how much space an indexed field takes, so report
  // it as zero.
  std::string size_string = "0";
  if (!f.index.has_value()) {
    size_t field_size;
    std::tie(field_size, size_string) = f.GetFieldSizeInformation();
  }
  if (f.offset.has_value()) {
    WriteField(f, size_string);
  } else {
    WriteFieldOffsetGetter(f);
  }
}

void Finish() {
  End(current_section_);
  if (!(completed_sections_ & FieldSectionType::kWeakSection)) {
    Begin(FieldSectionType::kWeakSection);
    End(FieldSectionType::kWeakSection);
  }
  if (!(completed_sections_ & FieldSectionType::kStrongSection)) {
    Begin(FieldSectionType::kStrongSection);
    End(FieldSectionType::kStrongSection);
  }
  is_finished_ = true;

  // In the presence of indexed fields, we already emitted kHeaderSize before
  // the indexed field.
  if (!type_->IsShape() && !header_size_emitted_) {
    WriteMarker("kHeaderSize");
  }
  if (!type_->IsAbstract() && type_->HasStaticSize()) {
    WriteMarker("kSize");
  }
}

protected:
const ClassType* type_;

private:
FieldSectionType GetSectionFor(const Field& f) {
  const Type* field_type = f.name_and_type.type;
  if (field_type == TypeOracle::GetVoidType()) {
    // Allow void type for marker constants of size zero.
    return current_section_;
  }
  StructType::Classification struct_contents =
      StructType::ClassificationFlag::kEmpty;
  if (auto field_as_struct = field_type->StructSupertype()) {
    struct_contents = (*field_as_struct)->ClassifyContents();
  }
  if ((struct_contents & StructType::ClassificationFlag::kStrongTagged) &&
      (struct_contents & StructType::ClassificationFlag::kWeakTagged)) {
    // It's okay for a struct to contain both strong and weak data. We'll just
    // treat the whole thing as weak. This is required for DescriptorEntry.
    struct_contents &= ~StructType::Classification(
        StructType::ClassificationFlag::kStrongTagged);
  }
  bool struct_contains_tagged_fields =
      (struct_contents & StructType::ClassificationFlag::kStrongTagged) ||
      (struct_contents & StructType::ClassificationFlag::kWeakTagged);
  if (struct_contains_tagged_fields &&
      (struct_contents & StructType::ClassificationFlag::kUntagged)) {
    // We can't declare what section a struct goes in if it has multiple
    // categories of data within.
    Error(
        "Classes do not support fields which are structs containing both "
        "tagged and untagged data.")
        .Position(f.pos);
  }
  if ((field_type->IsSubtypeOf(TypeOracle::GetStrongTaggedType()) ||
       struct_contents == StructType::ClassificationFlag::kStrongTagged) &&
      !f.custom_weak_marking) {
    return FieldSectionType::kStrongSection;
  } else if (field_type->IsSubtypeOf(TypeOracle::GetTaggedType()) ||
             struct_contains_tagged_fields) {
    return FieldSectionType::kWeakSection;
  } else {
    return FieldSectionType::kScalarSection;
  }
}
void UpdateSection(const Field& f) {
  FieldSectionType type = GetSectionFor(f);
  if (current_section_ == type) return;
  if (IsPointerSection(type)) {
    if (completed_sections_ & type) {
      std::stringstream s;
      s << "cannot declare field " << f.name_and_type.name << " in class "
        << type_->name() << ", because section " << ToString(type)
        << " to which it belongs has already been finished.";
      Error(s.str()).Position(f.pos);
    }
  }
  End(current_section_);
  current_section_ = type;
  Begin(current_section_);
}
void Begin(FieldSectionType type) {
  DCHECK(type != FieldSectionType::kNoSection)((void) 0);
  if (!IsPointerSection(type)) return;
  WriteMarker("kStartOf" + ToString(type) + "Offset");
}
void End(FieldSectionType type) {
  if (!IsPointerSection(type)) return;
  completed_sections_ |= type;
  WriteMarker("kEndOf" + ToString(type) + "Offset");
}

FieldSectionType current_section_ = FieldSectionType::kNoSection;
FieldSections completed_sections_ = FieldSectionType::kNoSection;
bool is_finished_ = false;
bool header_size_emitted_ = false;
3820};

3822void GenerateClassExport(const ClassType* type, std::ostream& header,
                       std::ostream& inl_header) {
const ClassType* super = type->GetSuperClass();
std::string parent = "TorqueGenerated" + type->name() + "<" + type->name() +
                     ", " + super->name() + ">";
header << "class " << type->name() << " : public " << parent << " {\n";
header << " public:\n";
if (type->ShouldGenerateBodyDescriptor()) {
  header << "  class BodyDescriptor;\n";
}
header << "  TQ_OBJECT_CONSTRUCTORS(" << type->name() << ")\n";
header << "};\n\n";
inl_header << "TQ_OBJECT_CONSTRUCTORS_IMPL(" << type->name() << ")\n";
3835}

3837}  // namespace

3839void ImplementationVisitor::GenerateVisitorLists(
  const std::string& output_directory) {
std::stringstream header;
std::string file_name = "visitor-lists.h";
{
  IncludeGuardScope include_guard(header, file_name);

  header << "#define TORQUE_INSTANCE_TYPE_TO_BODY_DESCRIPTOR_LIST(V)\\\n";
  for (const ClassType* type : TypeOracle::GetClasses()) {
    if (type->ShouldGenerateBodyDescriptor() && type->OwnInstanceType()) {
      std::string type_name =
          CapifyStringWithUnderscores(type->name()) + "_TYPE";
      header << "V(" << type_name << "," << type->name() << ")\\\n";
    }
  }
  header << "\n";

  header << "#define TORQUE_DATA_ONLY_VISITOR_ID_LIST(V)\\\n";
  for (const ClassType* type : TypeOracle::GetClasses()) {
    if (type->ShouldGenerateBodyDescriptor() && type->HasNoPointerSlots()) {
      header << "V(" << type->name() << ")\\\n";
    }
  }
  header << "\n";

  header << "#define TORQUE_POINTER_VISITOR_ID_LIST(V)\\\n";
  for (const ClassType* type : TypeOracle::GetClasses()) {
    if (type->ShouldGenerateBodyDescriptor() && !type->HasNoPointerSlots()) {
      header << "V(" << type->name() << ")\\\n";
    }
  }
  header << "\n";
}
const std::string output_header_path = output_directory + "/" + file_name;
WriteFile(output_header_path, header.str());
3874}

3876void ImplementationVisitor::GenerateBitFields(
  const std::string& output_directory) {
std::stringstream header;
std::string file_name = "bit-fields.h";
{
  IncludeGuardScope include_guard(header, file_name);
  header << "#include \"src/base/bit-field.h\"\n\n";
  NamespaceScope namespaces(header, {"v8", "internal"});

  for (const auto& type : TypeOracle::GetBitFieldStructTypes()) {
    bool all_single_bits = true;  // Track whether every field is one bit.

    header << "#define DEFINE_TORQUE_GENERATED_"
           << CapifyStringWithUnderscores(type->name()) << "() \\\n";
    std::string type_name = type->GetConstexprGeneratedTypeName();
    for (const auto& field : type->fields()) {
      const char* suffix = field.num_bits == 1 ? "Bit" : "Bits";
      all_single_bits = all_single_bits && field.num_bits == 1;
      std::string field_type_name =
          field.name_and_type.type->GetConstexprGeneratedTypeName();
      header << "  using " << CamelifyString(field.name_and_type.name)
             << suffix << " = base::BitField<" << field_type_name << ", "
             << field.offset << ", " << field.num_bits << ", " << type_name
             << ">; \\\n";
    }

    // If every field is one bit, we can also generate a convenient enum.
    if (all_single_bits) {
      header << "  enum Flag: " << type_name << " { \\\n";
      header << "    kNone = 0, \\\n";
      for (const auto& field : type->fields()) {
        header << "    k" << CamelifyString(field.name_and_type.name) << " = "
               << type_name << "{1} << " << field.offset << ", \\\n";
      }
      header << "  }; \\\n";
      header << "  using Flags = base::Flags<Flag>; \\\n";
      header << "  static constexpr int kFlagCount = "
             << type->fields().size() << "; \\\n";
    }

    header << "\n";
  }
}
const std::string output_header_path = output_directory + "/" + file_name;
WriteFile(output_header_path, header.str());
3921}

3923namespace {

3925class ClassFieldOffsetGenerator : public FieldOffsetsGenerator {
public:
ClassFieldOffsetGenerator(std::ostream& header, std::ostream& inline_header,
                          const ClassType* type, std::string gen_name,
                          const ClassType* parent)
    : FieldOffsetsGenerator(type),
      hdr_(header),
      inl_(inline_header),
      previous_field_end_((parent && parent->IsShape()) ? "P::kSize"
                                                        : "P::kHeaderSize"),
      gen_name_(gen_name) {}

void WriteField(const Field& f, const std::string& size_string) override {
  hdr_ << "  // " << f.pos << "\n";
  std::string field = "k" + CamelifyString(f.name_and_type.name) + "Offset";
  std::string field_end = field + "End";
  hdr_ << "  static constexpr int " << field << " = " << previous_field_end_
       << ";\n";
  hdr_ << "  static constexpr int " << field_end << " = " << field << " + "
       << size_string << " - 1;\n";
  previous_field_end_ = field_end + " + 1";
}

void WriteFieldOffsetGetter(const Field& f) override {
  // A static constexpr int is more convenient than a getter if the offset is
  // known.
  DCHECK(!f.offset.has_value())((void) 0);

  std::string function_name = CamelifyString(f.name_and_type.name) + "Offset";

  std::vector<cpp::TemplateParameter> params = {cpp::TemplateParameter("D"),
                                                cpp::TemplateParameter("P")};
  cpp::Class owner(std::move(params), gen_name_);

  auto getter = cpp::Function::DefaultGetter("int", &owner, function_name);
  getter.PrintDeclaration(hdr_);
  getter.PrintDefinition(inl_, [&](std::ostream& stream) {
    // Item 1 in a flattened slice is the offset.
    stream << "  return static_cast<int>(std::get<1>("
           << Callable::PrefixNameForCCOutput(type_->GetSliceMacroName(f))
           << "(*static_cast<const D*>(this))));\n";
  });
}
void WriteMarker(const std::string& marker) override {
  hdr_ << "  static constexpr int " << marker << " = " << previous_field_end_
       << ";\n";
}

private:
std::ostream& hdr_;
std::ostream& inl_;
std::string previous_field_end_;
std::string gen_name_;
3978};

3980class CppClassGenerator {
public:
CppClassGenerator(const ClassType* type, std::ostream& header,
                  std::ostream& inl_header, std::ostream& impl)
    : type_(type),
      super_(type->GetSuperClass()),
      name_(type->name()),
      gen_name_("TorqueGenerated" + name_),
      gen_name_T_(gen_name_ + "<D, P>"),
      gen_name_I_(gen_name_ + "<" + name_ + ", " + super_->name() + ">"),
      hdr_(header),
      inl_(inl_header),
      impl_(impl) {}
const std::string template_decl() const {
  return "template <class D, class P>";
}

void GenerateClass();
void GenerateCppObjectDefinitionAsserts();

private:
SourcePosition Position();

void GenerateClassConstructors();

// Generates getter and setter runtime member functions for the given class
// field. Traverses depth-first through any nested struct fields to generate
// accessors for them also; struct_fields represents the stack of currently
// active struct fields.
void GenerateFieldAccessors(const Field& class_field,
                            std::vector<const Field*>& struct_fields);
void EmitLoadFieldStatement(std::ostream& stream, const Field& class_field,
                            std::vector<const Field*>& struct_fields);
void EmitStoreFieldStatement(std::ostream& stream, const Field& class_field,
                             std::vector<const Field*>& struct_fields);

void GenerateClassCasts();

std::string GetFieldOffsetForAccessor(const Field& f);

// Gets the C++ type name that should be used in accessors for referring to
// the value of a class field.
std::string GetTypeNameForAccessor(const Field& f);

bool CanContainHeapObjects(const Type* t);

const ClassType* type_;
const ClassType* super_;
const std::string name_;
const std::string gen_name_;
const std::string gen_name_T_;
const std::string gen_name_I_;
std::ostream& hdr_;
std::ostream& inl_;
std::ostream& impl_;
4035};

4037base::Optional<std::vector<Field>> GetOrderedUniqueIndexFields(
  const ClassType& type) {
std::vector<Field> result;
std::set<std::string> index_names;
for (const Field& field : type.ComputeAllFields()) {
  if (field.index) {
    auto name_and_type = ExtractSimpleFieldArraySize(type, field.index->expr);
    if (!name_and_type) {
      return base::nullopt;
    }
    index_names.insert(name_and_type->name);
  }
}

for (const Field& field : type.ComputeAllFields()) {
  if (index_names.count(field.name_and_type.name) != 0) {
    result.push_back(field);
  }
}

return result;
4058}

4060void CppClassGenerator::GenerateClass() {
// Is<name>_NonInline(HeapObject)
if (!type_->IsShape()) {
  cpp::Function f("Is"s + name_ + "_NonInline");
  f.SetDescription("Alias for HeapObject::Is"s + name_ +
                   "() that avoids inlining.");
  f.SetExport(true);
  f.SetReturnType("bool");
  f.AddParameter("HeapObject", "o");

  f.PrintDeclaration(hdr_);
  hdr_ << "\n";
  f.PrintDefinition(impl_, [&](std::ostream& stream) {
    stream << "  return o.Is" << name_ << "();\n";
  });
}
hdr_ << "// Definition " << Position() << "\n";
hdr_ << template_decl() << "\n";
hdr_ << "class " << gen_name_ << " : public P {\n";
hdr_ << "  static_assert(\n"
     << "      std::is_same<" << name_ << ", D>::value,\n"
     << "      \"Use this class as direct base for " << name_ << ".\");\n";
hdr_ << "  static_assert(\n"
     << "      std::is_same<" << super_->name() << ", P>::value,\n"
     << "      \"Pass in " << super_->name()
     << " as second template parameter for " << gen_name_ << ".\");\n\n";
hdr_ << " public: \n";
hdr_ << "  using Super = P;\n";
hdr_ << "  using TorqueGeneratedClass = " << gen_name_ << "<D,P>;\n\n";
if (!type_->ShouldExport() && !type_->IsExtern()) {
  hdr_ << " protected: // not extern or @export\n";
}
for (const Field& f : type_->fields()) {
  CurrentSourcePosition::Scope scope(f.pos);
  std::vector<const Field*> struct_fields;
  GenerateFieldAccessors(f, struct_fields);
}
if (!type_->ShouldExport() && !type_->IsExtern()) {
  hdr_ << " public:\n";
}

GenerateClassCasts();

std::vector<cpp::TemplateParameter> templateArgs = {
    cpp::TemplateParameter("D"), cpp::TemplateParameter("P")};
cpp::Class c(std::move(templateArgs), gen_name_);

if (type_->ShouldGeneratePrint()) {
  hdr_ << "  DECL_PRINTER(" << name_ << ")\n\n";
}

if (type_->ShouldGenerateVerify()) {
  IfDefScope hdr_scope(hdr_, "VERIFY_HEAP");
  // V8_EXPORT_PRIVATE void Verify(Isolate*);
  cpp::Function f(&c, name_ + "Verify");
  f.SetExport();
  f.SetReturnType("void");
  f.AddParameter("Isolate*", "isolate");
  f.PrintDeclaration(hdr_);

  IfDefScope impl_scope(impl_, "VERIFY_HEAP");
  impl_ << "\ntemplate <>\n";
  impl_ << "void " << gen_name_I_ << "::" << name_
        << "Verify(Isolate* isolate) {\n";
  impl_ << "  TorqueGeneratedClassVerifiers::" << name_ << "Verify(" << name_
        << "::cast(*this), "
           "isolate);\n";
  impl_ << "}\n\n";
  impl_ << "\n";
}

hdr_ << "\n";
ClassFieldOffsetGenerator g(hdr_, inl_, type_, gen_name_,
                            type_->GetSuperClass());
for (auto f : type_->fields()) {
  CurrentSourcePosition::Scope scope(f.pos);
  g.RecordOffsetFor(f);
}
g.Finish();
hdr_ << "\n";

auto index_fields = GetOrderedUniqueIndexFields(*type_);

if (!index_fields.has_value()) {
  hdr_ << "  // SizeFor implementations not generated due to complex array "
          "lengths\n\n";

  const Field& last_field = type_->LastField();
  std::string last_field_item_size =
      std::get<1>(*SizeOf(last_field.name_and_type.type));

  // int AllocatedSize() const
  {
    cpp::Function f =
        cpp::Function::DefaultGetter("int", &c, "AllocatedSize");
    f.PrintDeclaration(hdr_);

    f.PrintDefinition(inl_, [&](std::ostream& stream) {
      stream << "  auto slice = "
             << Callable::PrefixNameForCCOutput(
                    type_->GetSliceMacroName(last_field))
             << "(*static_cast<const D*>(this));\n";
      stream << "  return static_cast<int>(std::get<1>(slice)) + "
             << last_field_item_size
             << " * static_cast<int>(std::get<2>(slice));\n";
    });
  }
} else if (type_->ShouldGenerateBodyDescriptor() ||
           (!type_->IsAbstract() &&
            !type_->IsSubtypeOf(TypeOracle::GetJSObjectType()))) {
  cpp::Function f(&c, "SizeFor");
  f.SetReturnType("int32_t");
  f.SetFlags(cpp::Function::kStatic | cpp::Function::kConstexpr |
             cpp::Function::kV8Inline);
  for (const Field& field : *index_fields) {
    f.AddParameter("int", field.name_and_type.name);
  }
  f.PrintInlineDefinition(hdr_, [&](std::ostream& stream) {
    if (index_fields->empty()) {
      stream << "    DCHECK(kHeaderSize == kSize && kHeaderSize == "
             << *type_->size().SingleValue() << ");\n";
    }
    stream << "    int32_t size = kHeaderSize;\n";
    for (const Field& field : type_->ComputeAllFields()) {
      if (field.index) {
        auto index_name_and_type =
            *ExtractSimpleFieldArraySize(*type_, field.index->expr);
        stream << "    size += " << index_name_and_type.name << " * "
               << std::get<0>(field.GetFieldSizeInformation()) << ";\n";
      }
    }
    if (type_->size().Alignment() < TargetArchitecture::TaggedSize()) {
      stream << "    size = OBJECT_POINTER_ALIGN(size);\n";
    }
    stream << "    return size;\n";
  });

  // V8_INLINE int32_t AllocatedSize() const
  {
    cpp::Function allocated_size_f =
        cpp::Function::DefaultGetter("int32_t", &c, "AllocatedSize");
    allocated_size_f.SetFlag(cpp::Function::kV8Inline);
    allocated_size_f.PrintInlineDefinition(hdr_, [&](std::ostream& stream) {
      stream << "    return SizeFor(";
      bool first = true;
      for (auto field : *index_fields) {
        if (!first) stream << ", ";
        stream << "this->" << field.name_and_type.name << "()";
        first = false;
      }
      stream << ");\n";
    });
  }
}

hdr_ << "  friend class Factory;\n\n";

GenerateClassConstructors();

hdr_ << "};\n\n";

if (type_->ShouldGenerateFullClassDefinition()) {
  // If this class extends from another class which is defined in the same tq
  // file, and that other class doesn't generate a full class definition, then
  // the resulting .inc file would be uncompilable due to ordering
  // requirements: the generated file must go before the hand-written
  // definition of the base class, but it must also go after that same
  // hand-written definition.
  base::Optional<const ClassType*> parent = type_->parent()->ClassSupertype();
  while (parent) {
    if ((*parent)->ShouldGenerateCppClassDefinitions() &&
        !(*parent)->ShouldGenerateFullClassDefinition() &&
        (*parent)->AttributedToFile() == type_->AttributedToFile()) {
      Error("Exported ", *type_,
            " cannot be in the same file as its parent extern ", **parent);
    }
    parent = (*parent)->parent()->ClassSupertype();
  }

  GenerateClassExport(type_, hdr_, inl_);
}
4241}

4243void CppClassGenerator::GenerateCppObjectDefinitionAsserts() {
hdr_ << "// Definition " << Position() << "\n"
     << template_decl() << "\n"
     << "class " << gen_name_ << "Asserts {\n";

ClassFieldOffsetGenerator g(hdr_, inl_, type_, gen_name_,
                            type_->GetSuperClass());
for (auto f : type_->fields()) {
  CurrentSourcePosition::Scope scope(f.pos);
  g.RecordOffsetFor(f);
}
g.Finish();
hdr_ << "\n";

for (auto f : type_->fields()) {
  std::string field = "k" + CamelifyString(f.name_and_type.name) + "Offset";
  std::string type = f.name_and_type.type->SimpleName();
  hdr_ << "  static_assert(" << field << " == D::" << field << ",\n"
       << "                \"Values of " << name_ << "::" << field
       << " defined in Torque and C++ do not match\");\n"
       << "  static_assert(StaticStringsEqual(\"" << type << "\", D::k"
       << CamelifyString(f.name_and_type.name) << "TqFieldType),\n"
       << "                \"Types of " << name_ << "::" << field
       << " specified in Torque and C++ do not match\");\n";
}
hdr_ << "  static_assert(kSize == D::kSize);\n";

hdr_ << "};\n\n";
4271}

4273void CppClassGenerator::GenerateClassCasts() {
cpp::Class owner({cpp::TemplateParameter("D"), cpp::TemplateParameter("P")},
                 gen_name_);
cpp::Function f(&owner, "cast");
f.SetFlags(cpp::Function::kV8Inline | cpp::Function::kStatic);
f.SetReturnType("D");
f.AddParameter("Object", "object");

// V8_INLINE static D cast(Object)
f.PrintDeclaration(hdr_);
f.PrintDefinition(inl_, [](std::ostream& stream) {
  stream << "    return D(object.ptr());\n";
});
// V8_INLINE static D unchecked_cast(Object)
f.SetName("unchecked_cast");
f.PrintInlineDefinition(hdr_, [](std::ostream& stream) {
  stream << "    return bit_cast<D>(object);\n";
});
4291}

4293SourcePosition CppClassGenerator::Position() { return type_->GetPosition(); }

4295void CppClassGenerator::GenerateClassConstructors() {
const ClassType* typecheck_type = type_;
while (typecheck_type->IsShape()) {
  typecheck_type = typecheck_type->GetSuperClass();

  // Shapes have already been checked earlier to inherit from JSObject, so we
  // should have found an appropriate type.
  DCHECK(typecheck_type)((void) 0);
}

hdr_ << "  template <class DAlias = D>\n";
hdr_ << "  constexpr " << gen_name_ << "() : P() {\n";
hdr_ << "    static_assert(\n";
hdr_ << "        std::is_base_of<" << gen_name_ << ", DAlias>::value,\n";
hdr_ << "        \"class " << gen_name_
     << " should be used as direct base for " << name_ << ".\");\n";
hdr_ << "  }\n\n";

hdr_ << " protected:\n";
hdr_ << "  inline explicit " << gen_name_ << "(Address ptr);\n";
hdr_ << "  // Special-purpose constructor for subclasses that have fast "
        "paths where\n";
hdr_ << "  // their ptr() is a Smi.\n";
hdr_ << "  inline explicit " << gen_name_
     << "(Address ptr, HeapObject::AllowInlineSmiStorage allow_smi);\n";

inl_ << "template<class D, class P>\n";
inl_ << "inline " << gen_name_T_ << "::" << gen_name_ << "(Address ptr)\n";
inl_ << "    : P(ptr) {\n";
inl_ << "  SLOW_DCHECK(Is" << typecheck_type->name()
     << "_NonInline(*this));\n";
inl_ << "}\n";

inl_ << "template<class D, class P>\n";
inl_ << "inline " << gen_name_T_ << "::" << gen_name_
     << "(Address ptr, HeapObject::AllowInlineSmiStorage allow_smi)\n";
inl_ << "    : P(ptr, allow_smi) {\n";
inl_ << "  SLOW_DCHECK("
     << "(allow_smi == HeapObject::AllowInlineSmiStorage::kAllowBeingASmi"
        " && this->IsSmi()) || Is"
     << typecheck_type->name() << "_NonInline(*this));\n";
inl_ << "}\n";
4337}

4339namespace {
4340std::string GenerateRuntimeTypeCheck(const Type* type,
                                   const std::string& value) {
bool maybe_object = !type->IsSubtypeOf(TypeOracle::GetStrongTaggedType());
std::stringstream type_check;
bool at_start = true;
// If weak pointers are allowed, then start by checking for a cleared value.
if (maybe_object) {
  type_check << value << ".IsCleared()";
  at_start = false;
}
for (const TypeChecker& runtime_type : type->GetTypeCheckers()) {
  if (!at_start) type_check << " || ";
  at_start = false;
  if (maybe_object) {
    bool strong = runtime_type.weak_ref_to.empty();
    if (strong && runtime_type.type == WEAK_HEAP_OBJECT) {
      // Rather than a generic Weak<T>, this is the basic type WeakHeapObject.
      // We can't validate anything more about the type of the object pointed
      // to, so just check that it's weak.
      type_check << value << ".IsWeak()";
    } else {
      type_check << "(" << (strong ? "!" : "") << value << ".IsWeak() && "
                 << value << ".GetHeapObjectOrSmi().Is"
                 << (strong ? runtime_type.type : runtime_type.weak_ref_to)
                 << "())";
    }
  } else {
    type_check << value << ".Is" << runtime_type.type << "()";
  }
}
return type_check.str();
4371}

4373void GenerateBoundsDCheck(std::ostream& os, const std::string& index,
                        const ClassType* type, const Field& f) {
os << "  DCHECK_GE(" << index << ", 0);\n";
std::string length_expression;
if (base::Optional<NameAndType> array_length =
        ExtractSimpleFieldArraySize(*type, f.index->expr)) {
  length_expression = "this ->" + array_length->name + "()";
} else {
  // The length is element 2 in the flattened field slice.
  length_expression =
      "static_cast<int>(std::get<2>(" +
      Callable::PrefixNameForCCOutput(type->GetSliceMacroName(f)) +
      "(*static_cast<const D*>(this))))";
}
os << "  DCHECK_LT(" << index << ", " << length_expression << ");\n";
4388}

4390bool CanGenerateFieldAccessors(const Type* field_type) {
// float64_or_hole should be treated like float64. For now, we don't need it.
// TODO(v8:10391) Generate accessors for external pointers.
return field_type != TypeOracle::GetVoidType() &&
       field_type != TypeOracle::GetFloat64OrHoleType() &&
       !field_type->IsSubtypeOf(TypeOracle::GetExternalPointerType());
4396}
4397}  // namespace

4399// TODO(sigurds): Keep in sync with DECL_ACCESSORS and ACCESSORS macro.
4400void CppClassGenerator::GenerateFieldAccessors(
  const Field& class_field, std::vector<const Field*>& struct_fields) {
const Field& innermost_field =
    struct_fields.empty() ? class_field : *struct_fields.back();
const Type* field_type = innermost_field.name_and_type.type;
if (!CanGenerateFieldAccessors(field_type)) return;

if (const StructType* struct_type = StructType::DynamicCast(field_type)) {
  struct_fields.resize(struct_fields.size() + 1);
  for (const Field& struct_field : struct_type->fields()) {
    struct_fields[struct_fields.size() - 1] = &struct_field;
    GenerateFieldAccessors(class_field, struct_fields);
  }
  struct_fields.resize(struct_fields.size() - 1);
  return;
}

bool indexed = class_field.index && !class_field.index->optional;
std::string type_name = GetTypeNameForAccessor(innermost_field);
bool can_contain_heap_objects = CanContainHeapObjects(field_type);

// Assemble an accessor name by accumulating together all of the nested field
// names.
std::string name = class_field.name_and_type.name;
for (const Field* nested_struct_field : struct_fields) {
  name += "_" + nested_struct_field->name_and_type.name;
}

// Generate declarations in header.
if (can_contain_heap_objects && !field_type->IsClassType() &&
    !field_type->IsStructType() &&
    field_type != TypeOracle::GetObjectType()) {
  hdr_ << "  // Torque type: " << field_type->ToString() << "\n";
}

std::vector<cpp::TemplateParameter> templateParameters = {
    cpp::TemplateParameter("D"), cpp::TemplateParameter("P")};
cpp::Class owner(std::move(templateParameters), gen_name_);

// getter
{
  auto getter = cpp::Function::DefaultGetter(type_name, &owner, name);
  if (indexed) {
    getter.AddParameter("int", "i");
  }
  const char* tag_argument;
  switch (class_field.read_synchronization) {
    case FieldSynchronization::kNone:
      tag_argument = "";
      break;
    case FieldSynchronization::kRelaxed:
      getter.AddParameter("RelaxedLoadTag");
      tag_argument = ", kRelaxedLoad";
      break;
    case FieldSynchronization::kAcquireRelease:
      getter.AddParameter("AcquireLoadTag");
      tag_argument = ", kAcquireLoad";
      break;
  }

  getter.PrintDeclaration(hdr_);

  // For tagged data, generate the extra getter that derives an
  // PtrComprCageBase from the current object's pointer.
  if (can_contain_heap_objects) {
    getter.PrintDefinition(inl_, [&](auto& stream) {
      stream
          << "  PtrComprCageBase cage_base = GetPtrComprCageBase(*this);\n";
      stream << "  return " << gen_name_ << "::" << name << "(cage_base"
             << (indexed ? ", i" : "") << tag_argument << ");\n";
    });

    getter.InsertParameter(0, "PtrComprCageBase", "cage_base");
    getter.PrintDeclaration(hdr_);
  }

  getter.PrintDefinition(inl_, [&](auto& stream) {
    stream << "  " << type_name << " value;\n";
    EmitLoadFieldStatement(stream, class_field, struct_fields);
    stream << "  return value;\n";
  });
}

// setter
{
  auto setter = cpp::Function::DefaultSetter(
      &owner, std::string("set_") + name, type_name, "value");
  if (indexed) {
    setter.InsertParameter(0, "int", "i");
  }
  switch (class_field.write_synchronization) {
    case FieldSynchronization::kNone:
      break;
    case FieldSynchronization::kRelaxed:
      setter.AddParameter("RelaxedStoreTag");
      break;
    case FieldSynchronization::kAcquireRelease:
      setter.AddParameter("ReleaseStoreTag");
      break;
  }
  if (can_contain_heap_objects) {
    setter.AddParameter("WriteBarrierMode", "mode", "UPDATE_WRITE_BARRIER");
  }
  setter.PrintDeclaration(hdr_);

  setter.PrintDefinition(inl_, [&](auto& stream) {
    EmitStoreFieldStatement(stream, class_field, struct_fields);
  });
}

hdr_ << "\n";
4511}

4513std::string CppClassGenerator::GetFieldOffsetForAccessor(const Field& f) {
if (f.offset.has_value()) {
  return "k" + CamelifyString(f.name_and_type.name) + "Offset";
}
return CamelifyString(f.name_and_type.name) + "Offset()";
4518}

4520std::string CppClassGenerator::GetTypeNameForAccessor(const Field& f) {
const Type* field_type = f.name_and_type.type;
if (!field_type->IsSubtypeOf(TypeOracle::GetTaggedType())) {
  const Type* constexpr_version = field_type->ConstexprVersion();
  if (!constexpr_version) {
    Error("Field accessor for ", type_->name(), ":: ", f.name_and_type.name,
          " cannot be generated because its type ", *field_type,
          " is neither a subclass of Object nor does the type have a "
          "constexpr "
          "version.")
        .Position(f.pos)
        .Throw();
  }
  return constexpr_version->GetGeneratedTypeName();
}
if (field_type->IsSubtypeOf(TypeOracle::GetSmiType())) {
  // Follow the convention to create Smi accessors with type int.
  return "int";
}
return field_type->UnhandlifiedCppTypeName();
4540}

4542bool CppClassGenerator::CanContainHeapObjects(const Type* t) {
return t->IsSubtypeOf(TypeOracle::GetTaggedType()) &&
       !t->IsSubtypeOf(TypeOracle::GetSmiType());
4545}

4547void CppClassGenerator::EmitLoadFieldStatement(
  std::ostream& stream, const Field& class_field,
  std::vector<const Field*>& struct_fields) {
const Field& innermost_field =
    struct_fields.empty() ? class_field : *struct_fields.back();
const Type* field_type = innermost_field.name_and_type.type;
std::string type_name = GetTypeNameForAccessor(innermost_field);
const std::string class_field_size =
    std::get<1>(class_field.GetFieldSizeInformation());

// field_offset contains both the offset from the beginning of the object to
// the class field and the combined offsets of any nested struct fields
// within, but not the index adjustment.
std::string field_offset = GetFieldOffsetForAccessor(class_field);
for (const Field* nested_struct_field : struct_fields) {
  field_offset += " + " + std::to_string(*nested_struct_field->offset);
}

std::string offset = field_offset;
if (class_field.index) {
  const char* index = class_field.index->optional ? "0" : "i";
  GenerateBoundsDCheck(stream, index, type_, class_field);
  stream << "  int offset = " << field_offset << " + " << index << " * "
         << class_field_size << ";\n";
  offset = "offset";
}

stream << "  value = ";

if (!field_type->IsSubtypeOf(TypeOracle::GetTaggedType())) {
  if (class_field.read_synchronization ==
      FieldSynchronization::kAcquireRelease) {
    ReportError("Torque doesn't support @cppAcquireRead on untagged data");
  } else if (class_field.read_synchronization ==
             FieldSynchronization::kRelaxed) {
    ReportError("Torque doesn't support @cppRelaxedRead on untagged data");
  }
  stream << "this->template ReadField<" << type_name << ">(" << offset
         << ");\n";
} else {
  const char* load;
  switch (class_field.read_synchronization) {
    case FieldSynchronization::kNone:
      load = "load";
      break;
    case FieldSynchronization::kRelaxed:
      load = "Relaxed_Load";
      break;
    case FieldSynchronization::kAcquireRelease:
      load = "Acquire_Load";
      break;
  }
  bool is_smi = field_type->IsSubtypeOf(TypeOracle::GetSmiType());
  const std::string load_type = is_smi ? "Smi" : type_name;
  const char* postfix = is_smi ? ".value()" : "";
  const char* optional_cage_base = is_smi ? "" : "cage_base, ";

  stream << "TaggedField<" << load_type << ">::" << load << "("
         << optional_cage_base << "*this, " << offset << ")" << postfix
         << ";\n";
}

if (CanContainHeapObjects(field_type)) {
  stream << "  DCHECK(" << GenerateRuntimeTypeCheck(field_type, "value")
         << ");\n";
}
4613}

4615void CppClassGenerator::EmitStoreFieldStatement(
  std::ostream& stream, const Field& class_field,
  std::vector<const Field*>& struct_fields) {
const Field& innermost_field =
    struct_fields.empty() ? class_field : *struct_fields.back();
const Type* field_type = innermost_field.name_and_type.type;
std::string type_name = GetTypeNameForAccessor(innermost_field);
const std::string class_field_size =
    std::get<1>(class_field.GetFieldSizeInformation());

// field_offset contains both the offset from the beginning of the object to
// the class field and the combined offsets of any nested struct fields
// within, but not the index adjustment.
std::string field_offset = GetFieldOffsetForAccessor(class_field);
for (const Field* nested_struct_field : struct_fields) {
  field_offset += " + " + std::to_string(*nested_struct_field->offset);
}

std::string offset = field_offset;
if (class_field.index) {
  const char* index = class_field.index->optional ? "0" : "i";
  GenerateBoundsDCheck(stream, index, type_, class_field);
  stream << "  int offset = " << field_offset << " + " << index << " * "
         << class_field_size << ";\n";
  offset = "offset";
}

if (!field_type->IsSubtypeOf(TypeOracle::GetTaggedType())) {
  stream << "  this->template WriteField<" << type_name << ">(" << offset
         << ", value);\n";
} else {
  bool strong_pointer = field_type->IsSubtypeOf(TypeOracle::GetObjectType());
  bool is_smi = field_type->IsSubtypeOf(TypeOracle::GetSmiType());
  const char* write_macro;
  if (!strong_pointer) {
    if (class_field.write_synchronization ==
        FieldSynchronization::kAcquireRelease) {
      ReportError("Torque doesn't support @releaseWrite on weak fields");
    }
    write_macro = "RELAXED_WRITE_WEAK_FIELD";
  } else {
    switch (class_field.write_synchronization) {
      case FieldSynchronization::kNone:
        write_macro = "WRITE_FIELD";
        break;
      case FieldSynchronization::kRelaxed:
        write_macro = "RELAXED_WRITE_FIELD";
        break;
      case FieldSynchronization::kAcquireRelease:
        write_macro = "RELEASE_WRITE_FIELD";
        break;
    }
  }
  const std::string value_to_write = is_smi ? "Smi::FromInt(value)" : "value";

  if (!is_smi) {
    stream << "  SLOW_DCHECK("
           << GenerateRuntimeTypeCheck(field_type, "value") << ");\n";
  }
  stream << "  " << write_macro << "(*this, " << offset << ", "
         << value_to_write << ");\n";
  if (!is_smi) {
    const char* write_barrier = strong_pointer
                                    ? "CONDITIONAL_WRITE_BARRIER"
                                    : "CONDITIONAL_WEAK_WRITE_BARRIER";
    stream << "  " << write_barrier << "(*this, " << offset
           << ", value, mode);\n";
  }
}
4684}

4686void GenerateStructLayoutDescription(std::ostream& header,
                                   const StructType* type) {
header << "struct TorqueGenerated" << CamelifyString(type->name())
       << "Offsets {\n";
for (const Field& field : type->fields()) {
  header << "  static constexpr int k"
         << CamelifyString(field.name_and_type.name)
         << "Offset = " << *field.offset << ";\n";
}
header << "  static constexpr int kSize = " << type->PackedSize() << ";\n";
header << "};\n\n";
4697}

4699}  // namespace

4701void ImplementationVisitor::GenerateClassDefinitions(
  const std::string& output_directory) {
std::stringstream factory_header;
std::stringstream factory_impl;
std::string factory_basename = "factory";

std::stringstream forward_declarations;
std::string forward_declarations_filename = "class-forward-declarations.h";

{
  factory_impl << "#include \"src/heap/factory-base.h\"\n";
  factory_impl << "#include \"src/heap/factory-base-inl.h\"\n";
  factory_impl << "#include \"src/heap/heap.h\"\n";
  factory_impl << "#include \"src/heap/heap-inl.h\"\n";
  factory_impl << "#include \"src/execution/isolate.h\"\n";
  factory_impl << "#include "
                  "\"src/objects/all-objects-inl.h\"\n\n";
  NamespaceScope factory_impl_namespaces(factory_impl, {"v8", "internal"});
  factory_impl << "\n";

  IncludeGuardScope include_guard(forward_declarations,
                                  forward_declarations_filename);
  NamespaceScope forward_declarations_namespaces(forward_declarations,
                                                 {"v8", "internal"});

  std::set<const StructType*, TypeLess> structs_used_in_classes;

  // Emit forward declarations.
  for (const ClassType* type : TypeOracle::GetClasses()) {
    CurrentSourcePosition::Scope position_activator(type->GetPosition());
    auto& streams = GlobalContext::GeneratedPerFile(type->AttributedToFile());
    std::ostream& header = streams.class_definition_headerfile;
    std::string name = type->ShouldGenerateCppClassDefinitions()
                           ? type->name()
                           : type->GetGeneratedTNodeTypeName();
    if (type->ShouldGenerateCppClassDefinitions()) {
      header << "class " << name << ";\n";
    }
    forward_declarations << "class " << name << ";\n";
  }

  for (const ClassType* type : TypeOracle::GetClasses()) {
    CurrentSourcePosition::Scope position_activator(type->GetPosition());
    auto& streams = GlobalContext::GeneratedPerFile(type->AttributedToFile());
    std::ostream& header = streams.class_definition_headerfile;
    std::ostream& inline_header = streams.class_definition_inline_headerfile;
    std::ostream& implementation = streams.class_definition_ccfile;

    if (type->ShouldGenerateCppClassDefinitions()) {
      CppClassGenerator g(type, header, inline_header, implementation);
      g.GenerateClass();
    } else if (type->ShouldGenerateCppObjectDefinitionAsserts()) {
      CppClassGenerator g(type, header, inline_header, implementation);
      g.GenerateCppObjectDefinitionAsserts();
    }
    for (const Field& f : type->fields()) {
      const Type* field_type = f.name_and_type.type;
      if (auto field_as_struct = field_type->StructSupertype()) {
        structs_used_in_classes.insert(*field_as_struct);
      }
    }
    if (type->ShouldGenerateFactoryFunction()) {
      std::string return_type = type->HandlifiedCppTypeName();
      std::string function_name = "New" + type->name();
      std::stringstream parameters;
      for (const Field& f : type->ComputeAllFields()) {
        if (f.name_and_type.name == "map") continue;
        if (!f.index) {
          std::string type_string =
              f.name_and_type.type->HandlifiedCppTypeName();
          parameters << type_string << " " << f.name_and_type.name << ", ";
        }
      }
      parameters << "AllocationType allocation_type";

      factory_header << return_type << " " << function_name << "("
                     << parameters.str() << ");\n";
      factory_impl << "template <typename Impl>\n";
      factory_impl << return_type
                   << " TorqueGeneratedFactory<Impl>::" << function_name
                   << "(" << parameters.str() << ") {\n";

      factory_impl << " int size = ";
      const ClassType* super = type->GetSuperClass();
      std::string gen_name = "TorqueGenerated" + type->name();
      std::string gen_name_T =
          gen_name + "<" + type->name() + ", " + super->name() + ">";
      factory_impl << gen_name_T << "::SizeFor(";

      bool first = true;
      auto index_fields = GetOrderedUniqueIndexFields(*type);
      CHECK(index_fields.has_value())do { if ((__builtin_expect(!!(!(index_fields.has_value())), 0
))) { V8_Fatal("Check failed: %s.", "index_fields.has_value()"
); } } while (false);
      for (auto index_field : *index_fields) {
        if (!first) {
          factory_impl << ", ";
        }
        factory_impl << index_field.name_and_type.name;
        first = false;
      }

      factory_impl << ");\n";
      factory_impl << "  Map map = factory()->read_only_roots()."
                   << SnakeifyString(type->name()) << "_map();";
      factory_impl << "  HeapObject raw_object =\n";
      factory_impl << "    factory()->AllocateRawWithImmortalMap(size, "
                      "allocation_type, map);\n";
      factory_impl << "  " << type->UnhandlifiedCppTypeName()
                   << " result = " << type->UnhandlifiedCppTypeName()
                   << "::cast(raw_object);\n";
      factory_impl << "  DisallowGarbageCollection no_gc;";
      factory_impl << "  WriteBarrierMode write_barrier_mode =\n"
                   << "     allocation_type == AllocationType::kYoung\n"
                   << "     ? SKIP_WRITE_BARRIER : UPDATE_WRITE_BARRIER;\n"
                   << "  USE(write_barrier_mode);\n";

      for (const Field& f : type->ComputeAllFields()) {
        if (f.name_and_type.name == "map") continue;
        if (!f.index) {
          factory_impl << "  result.TorqueGeneratedClass::set_"
                       << SnakeifyString(f.name_and_type.name) << "(";
          if (f.name_and_type.type->IsSubtypeOf(
                  TypeOracle::GetTaggedType()) &&
              !f.name_and_type.type->IsSubtypeOf(TypeOracle::GetSmiType())) {
            factory_impl << "*" << f.name_and_type.name
                         << ", write_barrier_mode";
          } else {
            factory_impl << f.name_and_type.name;
          }
          factory_impl << ");\n";
        }
      }

      factory_impl << "  return handle(result, factory()->isolate());\n";
      factory_impl << "}\n\n";

      factory_impl << "template EXPORT_TEMPLATE_DEFINE(V8_EXPORT_PRIVATE) "
                   << return_type
                   << "TorqueGeneratedFactory<Factory>::" << function_name
                   << "(" << parameters.str() << ");\n";
      factory_impl << "template EXPORT_TEMPLATE_DEFINE(V8_EXPORT_PRIVATE) "
                   << return_type << "TorqueGeneratedFactory<LocalFactory>::"
                   << function_name << "(" << parameters.str() << ");\n";

      factory_impl << "\n\n";
    }
  }

  for (const StructType* type : structs_used_in_classes) {
    CurrentSourcePosition::Scope position_activator(type->GetPosition());
    std::ostream& header =
        GlobalContext::GeneratedPerFile(type->GetPosition().source)
            .class_definition_headerfile;
    if (type != TypeOracle::GetFloat64OrHoleType()) {
      GenerateStructLayoutDescription(header, type);
    }
  }
}
WriteFile(output_directory + "/" + factory_basename + ".inc",
          factory_header.str());
WriteFile(output_directory + "/" + factory_basename + ".cc",
          factory_impl.str());
WriteFile(output_directory + "/" + forward_declarations_filename,
          forward_declarations.str());
4864}

4866namespace {
4867void GeneratePrintDefinitionsForClass(std::ostream& impl, const ClassType* type,
                                    const std::string& gen_name,
                                    const std::string& gen_name_T,
                                    const std::string template_params) {
impl << template_params << "\n";
impl << "void " << gen_name_T << "::" << type->name()
     << "Print(std::ostream& os) {\n";
impl << "  this->PrintHeader(os, \"" << type->name() << "\");\n";
auto hierarchy = type->GetHierarchy();
std::map<std::string, const AggregateType*> field_names;
for (const AggregateType* aggregate_type : hierarchy) {
  for (const Field& f : aggregate_type->fields()) {
    if (f.name_and_type.name == "map" || f.index.has_value() ||
        !CanGenerateFieldAccessors(f.name_and_type.type)) {
      continue;
    }
    std::string getter = f.name_and_type.name;
    if (aggregate_type != type) {
      // We must call getters directly on the class that provided them,
      // because a subclass could have hidden them.
      getter = aggregate_type->name() + "::TorqueGeneratedClass::" + getter;
    }
    if (f.name_and_type.type->IsSubtypeOf(TypeOracle::GetSmiType()) ||
        !f.name_and_type.type->IsSubtypeOf(TypeOracle::GetTaggedType())) {
      impl << "  os << \"\\n - " << f.name_and_type.name << ": \" << ";
      if (f.name_and_type.type->StructSupertype()) {
        // TODO(turbofan): Print struct fields too.
        impl << "\" <struct field printing still unimplemented>\";\n";
      } else {
        impl << "this->" << getter;
        switch (f.read_synchronization) {
          case FieldSynchronization::kNone:
            impl << "();\n";
            break;
          case FieldSynchronization::kRelaxed:
            impl << "(kRelaxedLoad);\n";
            break;
          case FieldSynchronization::kAcquireRelease:
            impl << "(kAcquireLoad);\n";
            break;
        }
      }
    } else {
      impl << "  os << \"\\n - " << f.name_and_type.name << ": \" << "
           << "Brief(this->" << getter;
      switch (f.read_synchronization) {
        case FieldSynchronization::kNone:
          impl << "());\n";
          break;
        case FieldSynchronization::kRelaxed:
          impl << "(kRelaxedLoad));\n";
          break;
        case FieldSynchronization::kAcquireRelease:
          impl << "(kAcquireLoad));\n";
          break;
      }
    }
  }
}
impl << "  os << '\\n';\n";
impl << "}\n\n";
4928}
4929}  // namespace

4931void ImplementationVisitor::GeneratePrintDefinitions(
  const std::string& output_directory) {
std::stringstream impl;
std::string file_name = "objects-printer.cc";
{
  IfDefScope object_print(impl, "OBJECT_PRINT");

  impl << "#include <iosfwd>\n\n";
  impl << "#include \"src/objects/all-objects-inl.h\"\n\n";

  NamespaceScope impl_namespaces(impl, {"v8", "internal"});

  for (const ClassType* type : TypeOracle::GetClasses()) {
    if (!type->ShouldGeneratePrint()) continue;
    DCHECK(type->ShouldGenerateCppClassDefinitions())((void) 0);
    const ClassType* super = type->GetSuperClass();
    std::string gen_name = "TorqueGenerated" + type->name();
    std::string gen_name_T =
        gen_name + "<" + type->name() + ", " + super->name() + ">";
    std::string template_decl = "template <>";
    GeneratePrintDefinitionsForClass(impl, type, gen_name, gen_name_T,
                                     template_decl);
  }
}

std::string new_contents(impl.str());
WriteFile(output_directory + "/" + file_name, new_contents);
4958}

4960base::Optional<std::string> MatchSimpleBodyDescriptor(const ClassType* type) {
std::vector<ObjectSlotKind> slots = type->ComputeHeaderSlotKinds();
if (!type->HasStaticSize()) {
  slots.push_back(*type->ComputeArraySlotKind());
}

// Skip the map slot.
size_t i = 1;
while (i < slots.size() && slots[i] == ObjectSlotKind::kNoPointer) ++i;
if (i == slots.size()) return "DataOnlyBodyDescriptor";
bool has_weak_pointers = false;
size_t start_index = i;
for (; i < slots.size(); ++i) {
  if (slots[i] == ObjectSlotKind::kStrongPointer) {
    continue;
  } else if (slots[i] == ObjectSlotKind::kMaybeObjectPointer) {
    has_weak_pointers = true;
  } else if (slots[i] == ObjectSlotKind::kNoPointer) {
    break;
  } else {
    return base::nullopt;
  }
}
size_t end_index = i;
for (; i < slots.size(); ++i) {
  if (slots[i] != ObjectSlotKind::kNoPointer) return base::nullopt;
}
size_t start_offset = start_index * TargetArchitecture::TaggedSize();
size_t end_offset = end_index * TargetArchitecture::TaggedSize();
// We pick a suffix-range body descriptor even in cases where the object size
// is fixed, to reduce the amount of code executed for object visitation.
if (end_index == slots.size()) {
  return ToString("SuffixRange", has_weak_pointers ? "Weak" : "",
                  "BodyDescriptor<", start_offset, ">");
}
if (!has_weak_pointers) {
  return ToString("FixedRangeBodyDescriptor<", start_offset, ", ", end_offset,
                  ">");
}
return base::nullopt;
5000}

5002void ImplementationVisitor::GenerateBodyDescriptors(
  const std::string& output_directory) {
std::string file_name = "objects-body-descriptors-inl.inc";
std::stringstream h_contents;

  for (const ClassType* type : TypeOracle::GetClasses()) {
    std::string name = type->name();
    if (!type->ShouldGenerateBodyDescriptor()) continue;

    bool has_array_fields = !type->HasStaticSize();
    std::vector<ObjectSlotKind> header_slot_kinds =
        type->ComputeHeaderSlotKinds();
    base::Optional<ObjectSlotKind> array_slot_kind =
        type->ComputeArraySlotKind();
    DCHECK_EQ(has_array_fields, array_slot_kind.has_value())((void) 0);

    h_contents << "class " << name << "::BodyDescriptor final : public ";
    if (auto descriptor_name = MatchSimpleBodyDescriptor(type)) {
      h_contents << *descriptor_name << " {\n";
      h_contents << " public:\n";
    } else {
      h_contents << "BodyDescriptorBase {\n";
      h_contents << " public:\n";

      h_contents << "  static bool IsValidSlot(Map map, HeapObject obj, int "
                    "offset) {\n";
      if (has_array_fields) {
        h_contents << "    if (offset < kHeaderSize) {\n";
      }
      h_contents << "      bool valid_slots[] = {";
      for (ObjectSlotKind slot : header_slot_kinds) {
        h_contents << (slot != ObjectSlotKind::kNoPointer ? "1" : "0") << ",";
      }
      h_contents << "};\n"
                 << "      return valid_slots[static_cast<unsigned "
                    "int>(offset)/kTaggedSize];\n";
      if (has_array_fields) {
        h_contents << "    }\n";
        bool array_is_tagged = *array_slot_kind != ObjectSlotKind::kNoPointer;
        h_contents << "    return " << (array_is_tagged ? "true" : "false")
                   << ";\n";
      }
      h_contents << "  }\n\n";

      h_contents << "  template <typename ObjectVisitor>\n";
      h_contents
          << "  static inline void IterateBody(Map map, HeapObject obj, "
             "int object_size, ObjectVisitor* v) {\n";

      std::vector<ObjectSlotKind> slots = std::move(header_slot_kinds);
      if (has_array_fields) slots.push_back(*array_slot_kind);

      // Skip the map slot.
      slots.erase(slots.begin());
      size_t start_offset = TargetArchitecture::TaggedSize();

      size_t end_offset = start_offset;
      ObjectSlotKind section_kind;
      for (size_t i = 0; i <= slots.size(); ++i) {
        base::Optional<ObjectSlotKind> next_section_kind;
        bool finished_section = false;
        if (i == 0) {
          next_section_kind = slots[i];
        } else if (i < slots.size()) {
          if (auto combined = Combine(section_kind, slots[i])) {
            next_section_kind = *combined;
          } else {
            next_section_kind = slots[i];
            finished_section = true;
          }
        } else {
          finished_section = true;
        }
        if (finished_section) {
          bool is_array_slot = i == slots.size() && has_array_fields;
          bool multiple_slots =
              is_array_slot ||
              (end_offset - start_offset > TargetArchitecture::TaggedSize());
          base::Optional<std::string> iterate_command;
          switch (section_kind) {
            case ObjectSlotKind::kStrongPointer:
              iterate_command = "IteratePointer";
              break;
            case ObjectSlotKind::kMaybeObjectPointer:
              iterate_command = "IterateMaybeWeakPointer";
              break;
            case ObjectSlotKind::kCustomWeakPointer:
              iterate_command = "IterateCustomWeakPointer";
              break;
            case ObjectSlotKind::kNoPointer:
              break;
          }
          if (iterate_command) {
            if (multiple_slots) *iterate_command += "s";
            h_contents << "    " << *iterate_command << "(obj, "
                       << start_offset;
            if (multiple_slots) {
              h_contents << ", "
                         << (i == slots.size() ? "object_size"
                                               : std::to_string(end_offset));
            }
            h_contents << ", v);\n";
          }
          start_offset = end_offset;
        }
        if (i < slots.size()) section_kind = *next_section_kind;
        end_offset += TargetArchitecture::TaggedSize();
      }

      h_contents << "  }\n\n";
    }

    h_contents
        << "  static inline int SizeOf(Map map, HeapObject raw_object) {\n";
    if (type->size().SingleValue()) {
      h_contents << "    return " << *type->size().SingleValue() << ";\n";
    } else {
      // We use an unchecked_cast here because this is used for concurrent
      // marking, where we shouldn't re-read the map.
      h_contents << "    return " << name
                 << "::unchecked_cast(raw_object).AllocatedSize();\n";
    }
    h_contents << "  }\n\n";

    h_contents << "};\n";
  }

  WriteFile(output_directory + "/" + file_name, h_contents.str());
5130}

5132namespace {

5134// Generate verification code for a single piece of class data, which might be
5135// nested within a struct or might be a single element in an indexed field (or
5136// both).
5137void GenerateFieldValueVerifier(const std::string& class_name, bool indexed,
                              std::string offset, const Field& leaf_field,
                              std::string indexed_field_size,
                              std::ostream& cc_contents, bool is_map) {
const Type* field_type = leaf_field.name_and_type.type;

bool maybe_object =
    !field_type->IsSubtypeOf(TypeOracle::GetStrongTaggedType());
const char* object_type = maybe_object ? "MaybeObject" : "Object";
const char* verify_fn =
    maybe_object ? "VerifyMaybeObjectPointer" : "VerifyPointer";
if (indexed) {
  offset += " + i * " + indexed_field_size;
}
// Name the local var based on the field name for nicer CHECK output.
const std::string value = leaf_field.name_and_type.name + "__value";

// Read the field.
if (is_map) {
  cc_contents << "    " << object_type << " " << value << " = o.map();\n";
} else {
  cc_contents << "    " << object_type << " " << value << " = TaggedField<"
              << object_type << ">::load(o, " << offset << ");\n";
}

// Call VerifyPointer or VerifyMaybeObjectPointer on it.
cc_contents << "    " << object_type << "::" << verify_fn << "(isolate, "
            << value << ");\n";

// Check that the value is of an appropriate type. We can skip this part for
// the Object type because it would not check anything beyond what we already
// checked with VerifyPointer.
if (field_type != TypeOracle::GetObjectType()) {
  cc_contents << "    CHECK(" << GenerateRuntimeTypeCheck(field_type, value)
              << ");\n";
}
5173}

5175void GenerateClassFieldVerifier(const std::string& class_name,
                              const ClassType& class_type, const Field& f,
                              std::ostream& h_contents,
                              std::ostream& cc_contents) {
const Type* field_type = f.name_and_type.type;

// We only verify tagged types, not raw numbers or pointers. Structs
// consisting of tagged types are also included.
if (!field_type->IsSubtypeOf(TypeOracle::GetTaggedType()) &&
    !field_type->StructSupertype())
  return;
if (field_type == TypeOracle::GetFloat64OrHoleType()) return;
// Do not verify if the field may be uninitialized.
if (TypeOracle::GetUninitializedType()->IsSubtypeOf(field_type)) return;

std::string field_start_offset;
if (f.index) {
  field_start_offset = f.name_and_type.name + "__offset";
  std::string length = f.name_and_type.name + "__length";
  cc_contents << "  intptr_t " << field_start_offset << ", " << length
              << ";\n";
  cc_contents << "  std::tie(std::ignore, " << field_start_offset << ", "
              << length << ") = "
              << Callable::PrefixNameForCCOutput(
                     class_type.GetSliceMacroName(f))
              << "(o);\n";

  // Slices use intptr, but TaggedField<T>.load() uses int, so verify that
  // such a cast is valid.
  cc_contents << "  CHECK_EQ(" << field_start_offset << ", static_cast<int>("
              << field_start_offset << "));\n";
  cc_contents << "  CHECK_EQ(" << length << ", static_cast<int>(" << length
              << "));\n";
  field_start_offset = "static_cast<int>(" + field_start_offset + ")";
  length = "static_cast<int>(" + length + ")";

  cc_contents << "  for (int i = 0; i < " << length << "; ++i) {\n";
} else {
  // Non-indexed fields have known offsets.
  field_start_offset = std::to_string(*f.offset);
  cc_contents << "  {\n";
}

if (auto struct_type = field_type->StructSupertype()) {
  for (const Field& struct_field : (*struct_type)->fields()) {
    if (struct_field.name_and_type.type->IsSubtypeOf(
            TypeOracle::GetTaggedType())) {
      GenerateFieldValueVerifier(
          class_name, f.index.has_value(),
          field_start_offset + " + " + std::to_string(*struct_field.offset),
          struct_field, std::to_string((*struct_type)->PackedSize()),
          cc_contents, f.name_and_type.name == "map");
    }
  }
} else {
  GenerateFieldValueVerifier(class_name, f.index.has_value(),
                             field_start_offset, f, "kTaggedSize",
                             cc_contents, f.name_and_type.name == "map");
}

cc_contents << "  }\n";
5236}

5238}  // namespace

5240void ImplementationVisitor::GenerateClassVerifiers(
  const std::string& output_directory) {
std::string file_name = "class-verifiers";
std::stringstream h_contents;
std::stringstream cc_contents;
{
  IncludeGuardScope include_guard(h_contents, file_name + ".h");
  IfDefScope verify_heap_h(h_contents, "VERIFY_HEAP");
  IfDefScope verify_heap_cc(cc_contents, "VERIFY_HEAP");

  h_contents << "#include \"src/base/macros.h\"\n\n";

  cc_contents << "#include \"torque-generated/" << file_name << ".h\"\n\n";
  cc_contents << "#include \"src/objects/all-objects-inl.h\"\n";

  IncludeObjectMacrosScope object_macros(cc_contents);

  NamespaceScope h_namespaces(h_contents, {"v8", "internal"});
  NamespaceScope cc_namespaces(cc_contents, {"v8", "internal"});

  cc_contents
      << "#include \"torque-generated/test/torque/test-torque-tq-inl.inc\"\n";

  // Generate forward declarations to avoid including any headers.
  h_contents << "class Isolate;\n";
  for (const ClassType* type : TypeOracle::GetClasses()) {
    if (!type->ShouldGenerateVerify()) continue;
    h_contents << "class " << type->name() << ";\n";
  }

  const char* verifier_class = "TorqueGeneratedClassVerifiers";

  h_contents << "class V8_EXPORT_PRIVATE " << verifier_class << "{\n";
  h_contents << " public:\n";

  for (const ClassType* type : TypeOracle::GetClasses()) {
    std::string name = type->name();
    if (!type->ShouldGenerateVerify()) continue;

    std::string method_name = name + "Verify";

    h_contents << "  static void " << method_name << "(" << name
               << " o, Isolate* isolate);\n";

    cc_contents << "void " << verifier_class << "::" << method_name << "("
                << name << " o, Isolate* isolate) {\n";

    // First, do any verification for the super class. Not all classes have
    // verifiers, so skip to the nearest super class that has one.
    const ClassType* super_type = type->GetSuperClass();
    while (super_type && !super_type->ShouldGenerateVerify()) {
      super_type = super_type->GetSuperClass();
    }
    if (super_type) {
      std::string super_name = super_type->name();
      cc_contents << "  o." << super_name << "Verify(isolate);\n";
    }

    // Second, verify that this object is what it claims to be.
    cc_contents << "  CHECK(o.Is" << name << "(isolate));\n";

    // Third, verify its properties.
    for (auto f : type->fields()) {
      GenerateClassFieldVerifier(name, *type, f, h_contents, cc_contents);
    }

    cc_contents << "}\n";
  }

  h_contents << "};\n";
}
WriteFile(output_directory + "/" + file_name + ".h", h_contents.str());
WriteFile(output_directory + "/" + file_name + ".cc", cc_contents.str());
5313}

5315void ImplementationVisitor::GenerateEnumVerifiers(
  const std::string& output_directory) {
std::string file_name = "enum-verifiers";
std::stringstream cc_contents;
{
  cc_contents << "#include \"src/compiler/code-assembler.h\"\n";
  for (const std::string& include_path : GlobalContext::CppIncludes()) {
    cc_contents << "#include " << StringLiteralQuote(include_path) << "\n";
  }
  cc_contents << "\n";

  NamespaceScope cc_namespaces(cc_contents, {"v8", "internal", ""});

  cc_contents << "class EnumVerifier {\n";
  for (const auto& desc : GlobalContext::Get().ast()->EnumDescriptions()) {
    cc_contents << "  // " << desc.name << " (" << desc.pos << ")\n";
    cc_contents << "  void VerifyEnum_" << desc.name << "("
                << desc.constexpr_generates
                << " x) {\n"
                   "    switch(x) {\n";
    for (const auto& entry : desc.entries) {
      cc_contents << "      case " << entry << ": break;\n";
    }
    if (desc.is_open) cc_contents << "      default: break;\n";
    cc_contents << "    }\n  }\n\n";
  }
  cc_contents << "};\n";
}

WriteFile(output_directory + "/" + file_name + ".cc", cc_contents.str());
5345}

5347void ImplementationVisitor::GenerateExportedMacrosAssembler(
  const std::string& output_directory) {
std::string file_name = "exported-macros-assembler";
std::stringstream h_contents;
std::stringstream cc_contents;
{
  IncludeGuardScope include_guard(h_contents, file_name + ".h");

  h_contents << "#include \"src/compiler/code-assembler.h\"\n";
  h_contents << "#include \"src/execution/frames.h\"\n";
  h_contents << "#include \"torque-generated/csa-types.h\"\n";

  for (const std::string& include_path : GlobalContext::CppIncludes()) {
    cc_contents << "#include " << StringLiteralQuote(include_path) << "\n";
  }
  cc_contents << "#include \"torque-generated/" << file_name << ".h\"\n";

  for (SourceId file : SourceFileMap::AllSources()) {
    cc_contents << "#include \"torque-generated/" +
                       SourceFileMap::PathFromV8RootWithoutExtension(file) +
                       "-tq-csa.h\"\n";
  }

  NamespaceScope h_namespaces(h_contents, {"v8", "internal"});
  NamespaceScope cc_namespaces(cc_contents, {"v8", "internal"});

  h_contents << "class V8_EXPORT_PRIVATE "
                "TorqueGeneratedExportedMacrosAssembler {\n"
             << " public:\n"
             << "  explicit TorqueGeneratedExportedMacrosAssembler"
                "(compiler::CodeAssemblerState* state) : state_(state) {\n"
             << "    USE(state_);\n"
             << "  }\n";

  for (auto& declarable : GlobalContext::AllDeclarables()) {
    TorqueMacro* macro = TorqueMacro::DynamicCast(declarable.get());
    if (!(macro && macro->IsExportedToCSA())) continue;
    CurrentSourcePosition::Scope position_activator(macro->Position());

    cpp::Class assembler("TorqueGeneratedExportedMacrosAssembler");
    std::vector<std::string> generated_parameter_names;
    cpp::Function f = GenerateFunction(
        &assembler, macro->ReadableName(), macro->signature(),
        macro->parameter_names(), false, &generated_parameter_names);

    f.PrintDeclaration(h_contents);
    f.PrintDefinition(cc_contents, [&](std::ostream& stream) {
      stream << "return " << macro->ExternalName() << "(state_";
      for (const auto& name : generated_parameter_names) {
        stream << ", " << name;
      }
      stream << ");";
    });
  }

  h_contents << " private:\n"
             << "  compiler::CodeAssemblerState* state_;\n"
             << "};\n";
}
WriteFile(output_directory + "/" + file_name + ".h", h_contents.str());
WriteFile(output_directory + "/" + file_name + ".cc", cc_contents.str());
5408}

5410namespace {

5412void CollectAllFields(const std::string& path, const Field& field,
                    std::vector<std::string>& result) {
if (field.name_and_type.type->StructSupertype()) {
  std::string next_path = path + field.name_and_type.name + ".";
  const StructType* struct_type =
      StructType::DynamicCast(field.name_and_type.type);
  for (const auto& inner_field : struct_type->fields()) {
    CollectAllFields(next_path, inner_field, result);
  }
} else {
  result.push_back(path + field.name_and_type.name);
}
5424}

5426}  // namespace

5428void ImplementationVisitor::GenerateCSATypes(
  const std::string& output_directory) {
std::string file_name = "csa-types";
std::stringstream h_contents;
{
  IncludeGuardScope include_guard(h_contents, file_name + ".h");
  h_contents << "#include \"src/compiler/code-assembler.h\"\n\n";

  NamespaceScope h_namespaces(h_contents, {"v8", "internal"});

  // Generates headers for all structs in a topologically-sorted order, since
  // TypeOracle keeps them in the order of their resolution
  for (const auto& type : TypeOracle::GetAggregateTypes()) {
    const StructType* struct_type = StructType::DynamicCast(type.get());
    if (!struct_type) continue;
    h_contents << "struct " << struct_type->GetGeneratedTypeNameImpl()
               << " {\n";
    for (auto& field : struct_type->fields()) {
      h_contents << "  " << field.name_and_type.type->GetGeneratedTypeName();
      h_contents << " " << field.name_and_type.name << ";\n";
    }
    h_contents << "\n  std::tuple<";
    bool first = true;
    for (const Type* lowered_type : LowerType(struct_type)) {
      if (!first) {
        h_contents << ", ";
      }
      first = false;
      h_contents << lowered_type->GetGeneratedTypeName();
    }
    std::vector<std::string> all_fields;
    for (auto& field : struct_type->fields()) {
      CollectAllFields("", field, all_fields);
    }
    h_contents << "> Flatten() const {\n"
                  "    return std::make_tuple(";
    PrintCommaSeparatedList(h_contents, all_fields);
    h_contents << ");\n";
    h_contents << "  }\n";
    h_contents << "};\n";
  }
}
WriteFile(output_directory + "/" + file_name + ".h", h_contents.str());
5471}

5473void ReportAllUnusedMacros() {
for (const auto& declarable : GlobalContext::AllDeclarables()) {
  if (!declarable->IsMacro() || declarable->IsExternMacro()) continue;

  Macro* macro = Macro::cast(declarable.get());
  if (macro->IsUsed()) continue;

  if (macro->IsTorqueMacro() && TorqueMacro::cast(macro)->IsExportedToCSA()) {
    continue;
  }
  // TODO(gsps): Mark methods of generic structs used if they are used in any
  // instantiation
  if (Method* method = Method::DynamicCast(macro)) {
    if (StructType* struct_type =
            StructType::DynamicCast(method->aggregate_type())) {
      if (struct_type->GetSpecializedFrom().has_value()) {
        continue;
      }
    }
  }

  std::vector<std::string> ignored_prefixes = {"Convert<", "Cast<",
                                               "FromConstexpr<"};
  const std::string name = macro->ReadableName();
  const bool ignore =
      StartsWithSingleUnderscore(name) ||
      std::any_of(ignored_prefixes.begin(), ignored_prefixes.end(),
                  [&name](const std::string& prefix) {
                    return StringStartsWith(name, prefix);
                  });

  if (!ignore) {
    Lint("Macro '", macro->ReadableName(), "' is never used.")
        .Position(macro->IdentifierPosition());
  }
}
5509}

5511}  // namespace torque
5512}  // namespace internal
5513}  // namespace v8