summaryrefslogtreecommitdiffstats
path: root/js/src/frontend/Parser.h
diff options
context:
space:
mode:
authorMatt A. Tobin <mattatobin@localhost.localdomain>2018-02-02 04:16:08 -0500
committerMatt A. Tobin <mattatobin@localhost.localdomain>2018-02-02 04:16:08 -0500
commit5f8de423f190bbb79a62f804151bc24824fa32d8 (patch)
tree10027f336435511475e392454359edea8e25895d /js/src/frontend/Parser.h
parent49ee0794b5d912db1f95dce6eb52d781dc210db5 (diff)
downloadUXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.gz
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.lz
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.xz
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.zip
Add m-esr52 at 52.6.0
Diffstat (limited to 'js/src/frontend/Parser.h')
-rw-r--r--js/src/frontend/Parser.h1430
1 files changed, 1430 insertions, 0 deletions
diff --git a/js/src/frontend/Parser.h b/js/src/frontend/Parser.h
new file mode 100644
index 000000000..0ad4d56a0
--- /dev/null
+++ b/js/src/frontend/Parser.h
@@ -0,0 +1,1430 @@
+/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+ * vim: set ts=8 sts=4 et sw=4 tw=99:
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+/* JS parser. */
+
+#ifndef frontend_Parser_h
+#define frontend_Parser_h
+
+#include "mozilla/Array.h"
+#include "mozilla/Maybe.h"
+
+#include "jspubtd.h"
+
+#include "frontend/BytecodeCompiler.h"
+#include "frontend/FullParseHandler.h"
+#include "frontend/NameAnalysisTypes.h"
+#include "frontend/NameCollections.h"
+#include "frontend/SharedContext.h"
+#include "frontend/SyntaxParseHandler.h"
+
+namespace js {
+
+class ModuleObject;
+
+namespace frontend {
+
+/*
+ * The struct ParseContext stores information about the current parsing context,
+ * which is part of the parser state (see the field Parser::pc). The current
+ * parsing context is either the global context, or the function currently being
+ * parsed. When the parser encounters a function definition, it creates a new
+ * ParseContext, makes it the new current context.
+ */
+class ParseContext : public Nestable<ParseContext>
+{
+ public:
+ // The intra-function statement stack.
+ //
+ // Used for early error checking that depend on the nesting structure of
+ // statements, such as continue/break targets, labels, and unbraced
+ // lexical declarations.
+ class Statement : public Nestable<Statement>
+ {
+ StatementKind kind_;
+
+ public:
+ using Nestable<Statement>::enclosing;
+ using Nestable<Statement>::findNearest;
+
+ Statement(ParseContext* pc, StatementKind kind)
+ : Nestable<Statement>(&pc->innermostStatement_),
+ kind_(kind)
+ { }
+
+ template <typename T> inline bool is() const;
+ template <typename T> inline T& as();
+
+ StatementKind kind() const {
+ return kind_;
+ }
+
+ void refineForKind(StatementKind newForKind) {
+ MOZ_ASSERT(kind_ == StatementKind::ForLoop);
+ MOZ_ASSERT(newForKind == StatementKind::ForInLoop ||
+ newForKind == StatementKind::ForOfLoop);
+ kind_ = newForKind;
+ }
+ };
+
+ class LabelStatement : public Statement
+ {
+ RootedAtom label_;
+
+ public:
+ LabelStatement(ParseContext* pc, JSAtom* label)
+ : Statement(pc, StatementKind::Label),
+ label_(pc->sc_->context, label)
+ { }
+
+ HandleAtom label() const {
+ return label_;
+ }
+ };
+
+ // The intra-function scope stack.
+ //
+ // Tracks declared and used names within a scope.
+ class Scope : public Nestable<Scope>
+ {
+ // Names declared in this scope. Corresponds to the union of
+ // VarDeclaredNames and LexicallyDeclaredNames in the ES spec.
+ //
+ // A 'var' declared name is a member of the declared name set of every
+ // scope in its scope contour.
+ //
+ // A lexically declared name is a member only of the declared name set of
+ // the scope in which it is declared.
+ PooledMapPtr<DeclaredNameMap> declared_;
+
+ // Monotonically increasing id.
+ uint32_t id_;
+
+ bool maybeReportOOM(ParseContext* pc, bool result) {
+ if (!result)
+ ReportOutOfMemory(pc->sc()->context);
+ return result;
+ }
+
+ public:
+ using DeclaredNamePtr = DeclaredNameMap::Ptr;
+ using AddDeclaredNamePtr = DeclaredNameMap::AddPtr;
+
+ using Nestable<Scope>::enclosing;
+
+ template <typename ParseHandler>
+ explicit Scope(Parser<ParseHandler>* parser)
+ : Nestable<Scope>(&parser->pc->innermostScope_),
+ declared_(parser->context->frontendCollectionPool()),
+ id_(parser->usedNames.nextScopeId())
+ { }
+
+ void dump(ParseContext* pc);
+
+ uint32_t id() const {
+ return id_;
+ }
+
+ MOZ_MUST_USE bool init(ParseContext* pc) {
+ if (id_ == UINT32_MAX) {
+ pc->tokenStream_.reportError(JSMSG_NEED_DIET, js_script_str);
+ return false;
+ }
+
+ return declared_.acquire(pc->sc()->context);
+ }
+
+ DeclaredNamePtr lookupDeclaredName(JSAtom* name) {
+ return declared_->lookup(name);
+ }
+
+ AddDeclaredNamePtr lookupDeclaredNameForAdd(JSAtom* name) {
+ return declared_->lookupForAdd(name);
+ }
+
+ MOZ_MUST_USE bool addDeclaredName(ParseContext* pc, AddDeclaredNamePtr& p, JSAtom* name,
+ DeclarationKind kind)
+ {
+ return maybeReportOOM(pc, declared_->add(p, name, DeclaredNameInfo(kind)));
+ }
+
+ // Remove all VarForAnnexBLexicalFunction declarations of a certain
+ // name from all scopes in pc's scope stack.
+ static void removeVarForAnnexBLexicalFunction(ParseContext* pc, JSAtom* name);
+
+ // Add and remove catch parameter names. Used to implement the odd
+ // semantics of catch bodies.
+ bool addCatchParameters(ParseContext* pc, Scope& catchParamScope);
+ void removeCatchParameters(ParseContext* pc, Scope& catchParamScope);
+
+ void useAsVarScope(ParseContext* pc) {
+ MOZ_ASSERT(!pc->varScope_);
+ pc->varScope_ = this;
+ }
+
+ // An iterator for the set of names a scope binds: the set of all
+ // declared names for 'var' scopes, and the set of lexically declared
+ // names for non-'var' scopes.
+ class BindingIter
+ {
+ friend class Scope;
+
+ DeclaredNameMap::Range declaredRange_;
+ mozilla::DebugOnly<uint32_t> count_;
+ bool isVarScope_;
+
+ BindingIter(Scope& scope, bool isVarScope)
+ : declaredRange_(scope.declared_->all()),
+ count_(0),
+ isVarScope_(isVarScope)
+ {
+ settle();
+ }
+
+ void settle() {
+ // Both var and lexically declared names are binding in a var
+ // scope.
+ if (isVarScope_)
+ return;
+
+ // Otherwise, pop only lexically declared names are
+ // binding. Pop the range until we find such a name.
+ while (!declaredRange_.empty()) {
+ if (BindingKindIsLexical(kind()))
+ break;
+ declaredRange_.popFront();
+ }
+ }
+
+ public:
+ bool done() const {
+ return declaredRange_.empty();
+ }
+
+ explicit operator bool() const {
+ return !done();
+ }
+
+ JSAtom* name() {
+ MOZ_ASSERT(!done());
+ return declaredRange_.front().key();
+ }
+
+ DeclarationKind declarationKind() {
+ MOZ_ASSERT(!done());
+ return declaredRange_.front().value()->kind();
+ }
+
+ BindingKind kind() {
+ return DeclarationKindToBindingKind(declarationKind());
+ }
+
+ bool closedOver() {
+ MOZ_ASSERT(!done());
+ return declaredRange_.front().value()->closedOver();
+ }
+
+ void setClosedOver() {
+ MOZ_ASSERT(!done());
+ return declaredRange_.front().value()->setClosedOver();
+ }
+
+ void operator++(int) {
+ MOZ_ASSERT(!done());
+ MOZ_ASSERT(count_ != UINT32_MAX);
+ declaredRange_.popFront();
+ settle();
+ }
+ };
+
+ inline BindingIter bindings(ParseContext* pc);
+ };
+
+ class VarScope : public Scope
+ {
+ public:
+ template <typename ParseHandler>
+ explicit VarScope(Parser<ParseHandler>* parser)
+ : Scope(parser)
+ {
+ useAsVarScope(parser->pc);
+ }
+ };
+
+ private:
+ // Context shared between parsing and bytecode generation.
+ SharedContext* sc_;
+
+ // TokenStream used for error reporting.
+ TokenStream& tokenStream_;
+
+ // The innermost statement, i.e., top of the statement stack.
+ Statement* innermostStatement_;
+
+ // The innermost scope, i.e., top of the scope stack.
+ //
+ // The outermost scope in the stack is usually varScope_. In the case of
+ // functions, the outermost scope is functionScope_, which may be
+ // varScope_. See comment above functionScope_.
+ Scope* innermostScope_;
+
+ // If isFunctionBox() and the function is a named lambda, the DeclEnv
+ // scope for named lambdas.
+ mozilla::Maybe<Scope> namedLambdaScope_;
+
+ // If isFunctionBox(), the scope for the function. If there are no
+ // parameter expressions, this is scope for the entire function. If there
+ // are parameter expressions, this holds the special function names
+ // ('.this', 'arguments') and the formal parameers.
+ mozilla::Maybe<Scope> functionScope_;
+
+ // The body-level scope. This always exists, but not necessarily at the
+ // beginning of parsing the script in the case of functions with parameter
+ // expressions.
+ Scope* varScope_;
+
+ // Inner function boxes in this context to try Annex B.3.3 semantics
+ // on. Only used when full parsing.
+ PooledVectorPtr<FunctionBoxVector> innerFunctionBoxesForAnnexB_;
+
+ // Simple formal parameter names, in order of appearance. Only used when
+ // isFunctionBox().
+ PooledVectorPtr<AtomVector> positionalFormalParameterNames_;
+
+ // Closed over binding names, in order of appearance. Null-delimited
+ // between scopes. Only used when syntax parsing.
+ PooledVectorPtr<AtomVector> closedOverBindingsForLazy_;
+
+ // Monotonically increasing id.
+ uint32_t scriptId_;
+
+ // Set when compiling a function using Parser::standaloneFunctionBody via
+ // the Function or Generator constructor.
+ bool isStandaloneFunctionBody_;
+
+ // Set when encountering a super.property inside a method. We need to mark
+ // the nearest super scope as needing a home object.
+ bool superScopeNeedsHomeObject_;
+
+ public:
+ // lastYieldOffset stores the offset of the last yield that was parsed.
+ // NoYieldOffset is its initial value.
+ static const uint32_t NoYieldOffset = UINT32_MAX;
+ uint32_t lastYieldOffset;
+
+ // lastAwaitOffset stores the offset of the last await that was parsed.
+ // NoAwaitOffset is its initial value.
+ static const uint32_t NoAwaitOffset = UINT32_MAX;
+ uint32_t lastAwaitOffset;
+
+ // All inner functions in this context. Only used when syntax parsing.
+ Rooted<GCVector<JSFunction*, 8>> innerFunctionsForLazy;
+
+ // In a function context, points to a Directive struct that can be updated
+ // to reflect new directives encountered in the Directive Prologue that
+ // require reparsing the function. In global/module/generator-tail contexts,
+ // we don't need to reparse when encountering a DirectivePrologue so this
+ // pointer may be nullptr.
+ Directives* newDirectives;
+
+ // Set when parsing a declaration-like destructuring pattern. This flag
+ // causes PrimaryExpr to create PN_NAME parse nodes for variable references
+ // which are not hooked into any definition's use chain, added to any tree
+ // context's AtomList, etc. etc. checkDestructuring will do that work
+ // later.
+ //
+ // The comments atop checkDestructuring explain the distinction between
+ // assignment-like and declaration-like destructuring patterns, and why
+ // they need to be treated differently.
+ mozilla::Maybe<DeclarationKind> inDestructuringDecl;
+
+ // Set when parsing a function and it has 'return <expr>;'
+ bool funHasReturnExpr;
+
+ // Set when parsing a function and it has 'return;'
+ bool funHasReturnVoid;
+
+ public:
+ template <typename ParseHandler>
+ ParseContext(Parser<ParseHandler>* prs, SharedContext* sc, Directives* newDirectives)
+ : Nestable<ParseContext>(&prs->pc),
+ sc_(sc),
+ tokenStream_(prs->tokenStream),
+ innermostStatement_(nullptr),
+ innermostScope_(nullptr),
+ varScope_(nullptr),
+ innerFunctionBoxesForAnnexB_(prs->context->frontendCollectionPool()),
+ positionalFormalParameterNames_(prs->context->frontendCollectionPool()),
+ closedOverBindingsForLazy_(prs->context->frontendCollectionPool()),
+ scriptId_(prs->usedNames.nextScriptId()),
+ isStandaloneFunctionBody_(false),
+ superScopeNeedsHomeObject_(false),
+ lastYieldOffset(NoYieldOffset),
+ lastAwaitOffset(NoAwaitOffset),
+ innerFunctionsForLazy(prs->context, GCVector<JSFunction*, 8>(prs->context)),
+ newDirectives(newDirectives),
+ funHasReturnExpr(false),
+ funHasReturnVoid(false)
+ {
+ if (isFunctionBox()) {
+ if (functionBox()->function()->isNamedLambda())
+ namedLambdaScope_.emplace(prs);
+ functionScope_.emplace(prs);
+ }
+ }
+
+ ~ParseContext();
+
+ MOZ_MUST_USE bool init();
+
+ SharedContext* sc() {
+ return sc_;
+ }
+
+ bool isFunctionBox() const {
+ return sc_->isFunctionBox();
+ }
+
+ FunctionBox* functionBox() {
+ return sc_->asFunctionBox();
+ }
+
+ Statement* innermostStatement() {
+ return innermostStatement_;
+ }
+
+ Scope* innermostScope() {
+ // There is always at least one scope: the 'var' scope.
+ MOZ_ASSERT(innermostScope_);
+ return innermostScope_;
+ }
+
+ Scope& namedLambdaScope() {
+ MOZ_ASSERT(functionBox()->function()->isNamedLambda());
+ return *namedLambdaScope_;
+ }
+
+ Scope& functionScope() {
+ MOZ_ASSERT(isFunctionBox());
+ return *functionScope_;
+ }
+
+ Scope& varScope() {
+ MOZ_ASSERT(varScope_);
+ return *varScope_;
+ }
+
+ bool isFunctionExtraBodyVarScopeInnermost() {
+ return isFunctionBox() && functionBox()->hasParameterExprs &&
+ innermostScope() == varScope_;
+ }
+
+ template <typename Predicate /* (Statement*) -> bool */>
+ Statement* findInnermostStatement(Predicate predicate) {
+ return Statement::findNearest(innermostStatement_, predicate);
+ }
+
+ template <typename T, typename Predicate /* (Statement*) -> bool */>
+ T* findInnermostStatement(Predicate predicate) {
+ return Statement::findNearest<T>(innermostStatement_, predicate);
+ }
+
+ AtomVector& positionalFormalParameterNames() {
+ return *positionalFormalParameterNames_;
+ }
+
+ AtomVector& closedOverBindingsForLazy() {
+ return *closedOverBindingsForLazy_;
+ }
+
+ MOZ_MUST_USE bool addInnerFunctionBoxForAnnexB(FunctionBox* funbox);
+ void removeInnerFunctionBoxesForAnnexB(JSAtom* name);
+ void finishInnerFunctionBoxesForAnnexB();
+
+ // True if we are at the topmost level of a entire script or function body.
+ // For example, while parsing this code we would encounter f1 and f2 at
+ // body level, but we would not encounter f3 or f4 at body level:
+ //
+ // function f1() { function f2() { } }
+ // if (cond) { function f3() { if (cond) { function f4() { } } } }
+ //
+ bool atBodyLevel() {
+ return !innermostStatement_;
+ }
+
+ bool atGlobalLevel() {
+ return atBodyLevel() && sc_->isGlobalContext();
+ }
+
+ // True if we are at the topmost level of a module only.
+ bool atModuleLevel() {
+ return atBodyLevel() && sc_->isModuleContext();
+ }
+
+ void setIsStandaloneFunctionBody() {
+ isStandaloneFunctionBody_ = true;
+ }
+
+ bool isStandaloneFunctionBody() const {
+ return isStandaloneFunctionBody_;
+ }
+
+ void setSuperScopeNeedsHomeObject() {
+ MOZ_ASSERT(sc_->allowSuperProperty());
+ superScopeNeedsHomeObject_ = true;
+ }
+
+ bool superScopeNeedsHomeObject() const {
+ return superScopeNeedsHomeObject_;
+ }
+
+ bool useAsmOrInsideUseAsm() const {
+ return sc_->isFunctionBox() && sc_->asFunctionBox()->useAsmOrInsideUseAsm();
+ }
+
+ // Most functions start off being parsed as non-generators.
+ // Non-generators transition to LegacyGenerator on parsing "yield" in JS 1.7.
+ // An ES6 generator is marked as a "star generator" before its body is parsed.
+ GeneratorKind generatorKind() const {
+ return sc_->isFunctionBox() ? sc_->asFunctionBox()->generatorKind() : NotGenerator;
+ }
+
+ bool isGenerator() const {
+ return generatorKind() != NotGenerator;
+ }
+
+ bool isLegacyGenerator() const {
+ return generatorKind() == LegacyGenerator;
+ }
+
+ bool isStarGenerator() const {
+ return generatorKind() == StarGenerator;
+ }
+
+ bool isAsync() const {
+ return sc_->isFunctionBox() && sc_->asFunctionBox()->isAsync();
+ }
+
+ FunctionAsyncKind asyncKind() const {
+ return isAsync() ? AsyncFunction : SyncFunction;
+ }
+
+ bool isArrowFunction() const {
+ return sc_->isFunctionBox() && sc_->asFunctionBox()->function()->isArrow();
+ }
+
+ bool isMethod() const {
+ return sc_->isFunctionBox() && sc_->asFunctionBox()->function()->isMethod();
+ }
+
+ uint32_t scriptId() const {
+ return scriptId_;
+ }
+};
+
+template <>
+inline bool
+ParseContext::Statement::is<ParseContext::LabelStatement>() const
+{
+ return kind_ == StatementKind::Label;
+}
+
+template <typename T>
+inline T&
+ParseContext::Statement::as()
+{
+ MOZ_ASSERT(is<T>());
+ return static_cast<T&>(*this);
+}
+
+inline ParseContext::Scope::BindingIter
+ParseContext::Scope::bindings(ParseContext* pc)
+{
+ // In function scopes with parameter expressions, function special names
+ // (like '.this') are declared as vars in the function scope, despite its
+ // not being the var scope.
+ return BindingIter(*this, pc->varScope_ == this || pc->functionScope_.ptrOr(nullptr) == this);
+}
+
+inline
+Directives::Directives(ParseContext* parent)
+ : strict_(parent->sc()->strict()),
+ asmJS_(parent->useAsmOrInsideUseAsm())
+{}
+
+enum VarContext { HoistVars, DontHoistVars };
+enum PropListType { ObjectLiteral, ClassBody, DerivedClassBody };
+enum class PropertyType {
+ Normal,
+ Shorthand,
+ CoverInitializedName,
+ Getter,
+ GetterNoExpressionClosure,
+ Setter,
+ SetterNoExpressionClosure,
+ Method,
+ GeneratorMethod,
+ AsyncMethod,
+ Constructor,
+ DerivedConstructor
+};
+
+// Specify a value for an ES6 grammar parametrization. We have no enum for
+// [Return] because its behavior is exactly equivalent to checking whether
+// we're in a function box -- easier and simpler than passing an extra
+// parameter everywhere.
+enum YieldHandling { YieldIsName, YieldIsKeyword };
+enum InHandling { InAllowed, InProhibited };
+enum DefaultHandling { NameRequired, AllowDefaultName };
+enum TripledotHandling { TripledotAllowed, TripledotProhibited };
+
+// A data structure for tracking used names per parsing session in order to
+// compute which bindings are closed over. Scripts and scopes are numbered
+// monotonically in textual order and name uses are tracked by lists of
+// (script id, scope id) pairs of their use sites.
+//
+// Intuitively, in a pair (P,S), P tracks the most nested function that has a
+// use of u, and S tracks the most nested scope that is still being parsed.
+//
+// P is used to answer the question "is u used by a nested function?"
+// S is used to answer the question "is u used in any scopes currently being
+// parsed?"
+//
+// The algorithm:
+//
+// Let Used by a map of names to lists.
+//
+// 1. Number all scopes in monotonic increasing order in textual order.
+// 2. Number all scripts in monotonic increasing order in textual order.
+// 3. When an identifier u is used in scope numbered S in script numbered P,
+// and u is found in Used,
+// a. Append (P,S) to Used[u].
+// b. Otherwise, assign the the list [(P,S)] to Used[u].
+// 4. When we finish parsing a scope S in script P, for each declared name d in
+// Declared(S):
+// a. If d is found in Used, mark d as closed over if there is a value
+// (P_d, S_d) in Used[d] such that P_d > P and S_d > S.
+// b. Remove all values (P_d, S_d) in Used[d] such that S_d are >= S.
+//
+// Steps 1 and 2 are implemented by UsedNameTracker::next{Script,Scope}Id.
+// Step 3 is implemented by UsedNameTracker::noteUsedInScope.
+// Step 4 is implemented by UsedNameTracker::noteBoundInScope and
+// Parser::propagateFreeNamesAndMarkClosedOverBindings.
+class UsedNameTracker
+{
+ public:
+ struct Use
+ {
+ uint32_t scriptId;
+ uint32_t scopeId;
+ };
+
+ class UsedNameInfo
+ {
+ friend class UsedNameTracker;
+
+ Vector<Use, 6> uses_;
+
+ void resetToScope(uint32_t scriptId, uint32_t scopeId);
+
+ public:
+ explicit UsedNameInfo(ExclusiveContext* cx)
+ : uses_(cx)
+ { }
+
+ UsedNameInfo(UsedNameInfo&& other)
+ : uses_(mozilla::Move(other.uses_))
+ { }
+
+ bool noteUsedInScope(uint32_t scriptId, uint32_t scopeId) {
+ if (uses_.empty() || uses_.back().scopeId < scopeId)
+ return uses_.append(Use { scriptId, scopeId });
+ return true;
+ }
+
+ void noteBoundInScope(uint32_t scriptId, uint32_t scopeId, bool* closedOver) {
+ *closedOver = false;
+ while (!uses_.empty()) {
+ Use& innermost = uses_.back();
+ if (innermost.scopeId < scopeId)
+ break;
+ if (innermost.scriptId > scriptId)
+ *closedOver = true;
+ uses_.popBack();
+ }
+ }
+
+ bool isUsedInScript(uint32_t scriptId) const {
+ return !uses_.empty() && uses_.back().scriptId >= scriptId;
+ }
+ };
+
+ using UsedNameMap = HashMap<JSAtom*,
+ UsedNameInfo,
+ DefaultHasher<JSAtom*>>;
+
+ private:
+ // The map of names to chains of uses.
+ UsedNameMap map_;
+
+ // Monotonically increasing id for all nested scripts.
+ uint32_t scriptCounter_;
+
+ // Monotonically increasing id for all nested scopes.
+ uint32_t scopeCounter_;
+
+ public:
+ explicit UsedNameTracker(ExclusiveContext* cx)
+ : map_(cx),
+ scriptCounter_(0),
+ scopeCounter_(0)
+ { }
+
+ MOZ_MUST_USE bool init() {
+ return map_.init();
+ }
+
+ uint32_t nextScriptId() {
+ MOZ_ASSERT(scriptCounter_ != UINT32_MAX,
+ "ParseContext::Scope::init should have prevented wraparound");
+ return scriptCounter_++;
+ }
+
+ uint32_t nextScopeId() {
+ MOZ_ASSERT(scopeCounter_ != UINT32_MAX);
+ return scopeCounter_++;
+ }
+
+ UsedNameMap::Ptr lookup(JSAtom* name) const {
+ return map_.lookup(name);
+ }
+
+ MOZ_MUST_USE bool noteUse(ExclusiveContext* cx, JSAtom* name,
+ uint32_t scriptId, uint32_t scopeId);
+
+ struct RewindToken
+ {
+ private:
+ friend class UsedNameTracker;
+ uint32_t scriptId;
+ uint32_t scopeId;
+ };
+
+ RewindToken getRewindToken() const {
+ RewindToken token;
+ token.scriptId = scriptCounter_;
+ token.scopeId = scopeCounter_;
+ return token;
+ }
+
+ // Resets state so that scriptId and scopeId are the innermost script and
+ // scope, respectively. Used for rewinding state on syntax parse failure.
+ void rewind(RewindToken token);
+
+ // Resets state to beginning of compilation.
+ void reset() {
+ map_.clear();
+ RewindToken token;
+ token.scriptId = 0;
+ token.scopeId = 0;
+ rewind(token);
+ }
+};
+
+template <typename ParseHandler>
+class Parser final : private JS::AutoGCRooter, public StrictModeGetter
+{
+ private:
+ using Node = typename ParseHandler::Node;
+
+ /*
+ * A class for temporarily stashing errors while parsing continues.
+ *
+ * The ability to stash an error is useful for handling situations where we
+ * aren't able to verify that an error has occurred until later in the parse.
+ * For instance | ({x=1}) | is always parsed as an object literal with
+ * a SyntaxError, however, in the case where it is followed by '=>' we rewind
+ * and reparse it as a valid arrow function. Here a PossibleError would be
+ * set to 'pending' when the initial SyntaxError was encountered then 'resolved'
+ * just before rewinding the parser.
+ *
+ * There are currently two kinds of PossibleErrors: Expression and
+ * Destructuring errors. Expression errors are used to mark a possible
+ * syntax error when a grammar production is used in an expression context.
+ * For example in |{x = 1}|, we mark the CoverInitializedName |x = 1| as a
+ * possible expression error, because CoverInitializedName productions
+ * are disallowed when an actual ObjectLiteral is expected.
+ * Destructuring errors are used to record possible syntax errors in
+ * destructuring contexts. For example in |[...rest, ] = []|, we initially
+ * mark the trailing comma after the spread expression as a possible
+ * destructuring error, because the ArrayAssignmentPattern grammar
+ * production doesn't allow a trailing comma after the rest element.
+ *
+ * When using PossibleError one should set a pending error at the location
+ * where an error occurs. From that point, the error may be resolved
+ * (invalidated) or left until the PossibleError is checked.
+ *
+ * Ex:
+ * PossibleError possibleError(*this);
+ * possibleError.setPendingExpressionError(pn, JSMSG_BAD_PROP_ID);
+ * // A JSMSG_BAD_PROP_ID ParseError is reported, returns false.
+ * if (!possibleError.checkForExpressionError())
+ * return false; // we reach this point with a pending exception
+ *
+ * PossibleError possibleError(*this);
+ * possibleError.setPendingExpressionError(pn, JSMSG_BAD_PROP_ID);
+ * // Returns true, no error is reported.
+ * if (!possibleError.checkForDestructuringError())
+ * return false; // not reached, no pending exception
+ *
+ * PossibleError possibleError(*this);
+ * // Returns true, no error is reported.
+ * if (!possibleError.checkForExpressionError())
+ * return false; // not reached, no pending exception
+ */
+ class MOZ_STACK_CLASS PossibleError
+ {
+ private:
+ enum class ErrorKind { Expression, Destructuring };
+
+ enum class ErrorState { None, Pending };
+
+ struct Error {
+ ErrorState state_ = ErrorState::None;
+
+ // Error reporting fields.
+ uint32_t offset_;
+ unsigned errorNumber_;
+ };
+
+ Parser<ParseHandler>& parser_;
+ Error exprError_;
+ Error destructuringError_;
+
+ // Returns the error report.
+ Error& error(ErrorKind kind);
+
+ // Return true if an error is pending without reporting
+ bool hasError(ErrorKind kind);
+
+ // Resolve any pending error.
+ void setResolved(ErrorKind kind);
+
+ // Set a pending error. Only a single error may be set per instance and
+ // error kind.
+ void setPending(ErrorKind kind, Node pn, unsigned errorNumber);
+
+ // If there is a pending error, report it and return false, otherwise
+ // return true.
+ bool checkForError(ErrorKind kind);
+
+ // Transfer an existing error to another instance.
+ void transferErrorTo(ErrorKind kind, PossibleError* other);
+
+ public:
+ explicit PossibleError(Parser<ParseHandler>& parser);
+
+ // Set a pending destructuring error. Only a single error may be set
+ // per instance, i.e. subsequent calls to this method are ignored and
+ // won't overwrite the existing pending error.
+ void setPendingDestructuringError(Node pn, unsigned errorNumber);
+
+ // Set a pending expression error. Only a single error may be set per
+ // instance, i.e. subsequent calls to this method are ignored and won't
+ // overwrite the existing pending error.
+ void setPendingExpressionError(Node pn, unsigned errorNumber);
+
+ // If there is a pending destructuring error, report it and return
+ // false, otherwise return true. Clears any pending expression error.
+ bool checkForDestructuringError();
+
+ // If there is a pending expression error, report it and return false,
+ // otherwise return true. Clears any pending destructuring error.
+ bool checkForExpressionError();
+
+ // Pass pending errors between possible error instances. This is useful
+ // for extending the lifetime of a pending error beyond the scope of
+ // the PossibleError where it was initially set (keeping in mind that
+ // PossibleError is a MOZ_STACK_CLASS).
+ void transferErrorsTo(PossibleError* other);
+ };
+
+ public:
+ ExclusiveContext* const context;
+
+ LifoAlloc& alloc;
+
+ TokenStream tokenStream;
+ LifoAlloc::Mark tempPoolMark;
+
+ /* list of parsed objects for GC tracing */
+ ObjectBox* traceListHead;
+
+ /* innermost parse context (stack-allocated) */
+ ParseContext* pc;
+
+ // For tracking used names in this parsing session.
+ UsedNameTracker& usedNames;
+
+ /* Compression token for aborting. */
+ SourceCompressionTask* sct;
+
+ ScriptSource* ss;
+
+ /* Root atoms and objects allocated for the parsed tree. */
+ AutoKeepAtoms keepAtoms;
+
+ /* Perform constant-folding; must be true when interfacing with the emitter. */
+ const bool foldConstants:1;
+
+ private:
+#if DEBUG
+ /* Our fallible 'checkOptions' member function has been called. */
+ bool checkOptionsCalled:1;
+#endif
+
+ /*
+ * Not all language constructs can be handled during syntax parsing. If it
+ * is not known whether the parse succeeds or fails, this bit is set and
+ * the parse will return false.
+ */
+ bool abortedSyntaxParse:1;
+
+ /* Unexpected end of input, i.e. TOK_EOF not at top-level. */
+ bool isUnexpectedEOF_:1;
+
+ public:
+ /* State specific to the kind of parse being performed. */
+ ParseHandler handler;
+
+ void prepareNodeForMutation(Node node) { handler.prepareNodeForMutation(node); }
+ void freeTree(Node node) { handler.freeTree(node); }
+
+ private:
+ bool reportHelper(ParseReportKind kind, bool strict, uint32_t offset,
+ unsigned errorNumber, va_list args);
+ public:
+ bool report(ParseReportKind kind, bool strict, Node pn, unsigned errorNumber, ...);
+ bool reportNoOffset(ParseReportKind kind, bool strict, unsigned errorNumber, ...);
+ bool reportWithOffset(ParseReportKind kind, bool strict, uint32_t offset, unsigned errorNumber,
+ ...);
+
+ Parser(ExclusiveContext* cx, LifoAlloc& alloc, const ReadOnlyCompileOptions& options,
+ const char16_t* chars, size_t length, bool foldConstants, UsedNameTracker& usedNames,
+ Parser<SyntaxParseHandler>* syntaxParser, LazyScript* lazyOuterFunction);
+ ~Parser();
+
+ bool checkOptions();
+
+ // A Parser::Mark is the extension of the LifoAlloc::Mark to the entire
+ // Parser's state. Note: clients must still take care that any ParseContext
+ // that points into released ParseNodes is destroyed.
+ class Mark
+ {
+ friend class Parser;
+ LifoAlloc::Mark mark;
+ ObjectBox* traceListHead;
+ };
+ Mark mark() const {
+ Mark m;
+ m.mark = alloc.mark();
+ m.traceListHead = traceListHead;
+ return m;
+ }
+ void release(Mark m) {
+ alloc.release(m.mark);
+ traceListHead = m.traceListHead;
+ }
+
+ friend void js::frontend::MarkParser(JSTracer* trc, JS::AutoGCRooter* parser);
+
+ const char* getFilename() const { return tokenStream.getFilename(); }
+ JSVersion versionNumber() const { return tokenStream.versionNumber(); }
+
+ /*
+ * Parse a top-level JS script.
+ */
+ Node parse();
+
+ /*
+ * Allocate a new parsed object or function container from
+ * cx->tempLifoAlloc.
+ */
+ ObjectBox* newObjectBox(JSObject* obj);
+ FunctionBox* newFunctionBox(Node fn, JSFunction* fun, Directives directives,
+ GeneratorKind generatorKind, FunctionAsyncKind asyncKind,
+ bool tryAnnexB);
+
+ /*
+ * Create a new function object given a name (which is optional if this is
+ * a function expression).
+ */
+ JSFunction* newFunction(HandleAtom atom, FunctionSyntaxKind kind,
+ GeneratorKind generatorKind, FunctionAsyncKind asyncKind,
+ HandleObject proto);
+
+ void trace(JSTracer* trc);
+
+ bool hadAbortedSyntaxParse() {
+ return abortedSyntaxParse;
+ }
+ void clearAbortedSyntaxParse() {
+ abortedSyntaxParse = false;
+ }
+
+ bool isUnexpectedEOF() const { return isUnexpectedEOF_; }
+
+ bool checkUnescapedName();
+
+ private:
+ Parser* thisForCtor() { return this; }
+
+ JSAtom* stopStringCompression();
+
+ Node stringLiteral();
+ Node noSubstitutionTemplate();
+ Node templateLiteral(YieldHandling yieldHandling);
+ bool taggedTemplate(YieldHandling yieldHandling, Node nodeList, TokenKind tt);
+ bool appendToCallSiteObj(Node callSiteObj);
+ bool addExprAndGetNextTemplStrToken(YieldHandling yieldHandling, Node nodeList,
+ TokenKind* ttp);
+ bool checkStatementsEOF();
+
+ inline Node newName(PropertyName* name);
+ inline Node newName(PropertyName* name, TokenPos pos);
+ inline Node newYieldExpression(uint32_t begin, Node expr, bool isYieldStar = false);
+ inline Node newAwaitExpression(uint32_t begin, Node expr);
+
+ inline bool abortIfSyntaxParser();
+
+ public:
+ /* Public entry points for parsing. */
+ Node statement(YieldHandling yieldHandling);
+ Node statementListItem(YieldHandling yieldHandling, bool canHaveDirectives = false);
+
+ bool maybeParseDirective(Node list, Node pn, bool* cont);
+
+ // Parse the body of an eval.
+ //
+ // Eval scripts are distinguished from global scripts in that in ES6, per
+ // 18.2.1.1 steps 9 and 10, all eval scripts are executed under a fresh
+ // lexical scope.
+ Node evalBody(EvalSharedContext* evalsc);
+
+ // Parse the body of a global script.
+ Node globalBody(GlobalSharedContext* globalsc);
+
+ // Parse a module.
+ Node moduleBody(ModuleSharedContext* modulesc);
+
+ // Parse a function, given only its body. Used for the Function and
+ // Generator constructors.
+ Node standaloneFunctionBody(HandleFunction fun, HandleScope enclosingScope,
+ Handle<PropertyNameVector> formals,
+ GeneratorKind generatorKind, FunctionAsyncKind asyncKind,
+ Directives inheritedDirectives, Directives* newDirectives);
+
+ // Parse a function, given only its arguments and body. Used for lazily
+ // parsed functions.
+ Node standaloneLazyFunction(HandleFunction fun, bool strict,
+ GeneratorKind generatorKind, FunctionAsyncKind asyncKind);
+
+ // Parse an inner function given an enclosing ParseContext and a
+ // FunctionBox for the inner function.
+ bool innerFunction(Node pn, ParseContext* outerpc, FunctionBox* funbox, InHandling inHandling,
+ YieldHandling yieldHandling, FunctionSyntaxKind kind,
+ Directives inheritedDirectives, Directives* newDirectives);
+
+ // Parse a function's formal parameters and its body assuming its function
+ // ParseContext is already on the stack.
+ bool functionFormalParametersAndBody(InHandling inHandling, YieldHandling yieldHandling,
+ Node pn, FunctionSyntaxKind kind);
+
+ // Determine whether |yield| is a valid name in the current context, or
+ // whether it's prohibited due to strictness, JS version, or occurrence
+ // inside a star generator.
+ bool yieldExpressionsSupported() {
+ return (versionNumber() >= JSVERSION_1_7 || pc->isGenerator()) && !pc->isAsync();
+ }
+
+ // Match the current token against the BindingIdentifier production with
+ // the given Yield parameter. If there is no match, report a syntax
+ // error.
+ PropertyName* bindingIdentifier(YieldHandling yieldHandling);
+
+ virtual bool strictMode() { return pc->sc()->strict(); }
+ bool setLocalStrictMode(bool strict) {
+ MOZ_ASSERT(tokenStream.debugHasNoLookahead());
+ return pc->sc()->setLocalStrictMode(strict);
+ }
+
+ const ReadOnlyCompileOptions& options() const {
+ return tokenStream.options();
+ }
+
+ private:
+ enum InvokedPrediction { PredictUninvoked = false, PredictInvoked = true };
+ enum ForInitLocation { InForInit, NotInForInit };
+
+ private:
+ /*
+ * JS parsers, from lowest to highest precedence.
+ *
+ * Each parser must be called during the dynamic scope of a ParseContext
+ * object, pointed to by this->pc.
+ *
+ * Each returns a parse node tree or null on error.
+ *
+ * Parsers whose name has a '1' suffix leave the TokenStream state
+ * pointing to the token one past the end of the parsed fragment. For a
+ * number of the parsers this is convenient and avoids a lot of
+ * unnecessary ungetting and regetting of tokens.
+ *
+ * Some parsers have two versions: an always-inlined version (with an 'i'
+ * suffix) and a never-inlined version (with an 'n' suffix).
+ */
+ Node functionStmt(YieldHandling yieldHandling, DefaultHandling defaultHandling,
+ FunctionAsyncKind asyncKind = SyncFunction);
+ Node functionExpr(InvokedPrediction invoked = PredictUninvoked,
+ FunctionAsyncKind asyncKind = SyncFunction);
+
+ Node statementList(YieldHandling yieldHandling);
+
+ Node blockStatement(YieldHandling yieldHandling,
+ unsigned errorNumber = JSMSG_CURLY_IN_COMPOUND);
+ Node doWhileStatement(YieldHandling yieldHandling);
+ Node whileStatement(YieldHandling yieldHandling);
+
+ Node forStatement(YieldHandling yieldHandling);
+ bool forHeadStart(YieldHandling yieldHandling,
+ ParseNodeKind* forHeadKind,
+ Node* forInitialPart,
+ mozilla::Maybe<ParseContext::Scope>& forLetImpliedScope,
+ Node* forInOrOfExpression);
+ bool validateForInOrOfLHSExpression(Node target, PossibleError* possibleError);
+ Node expressionAfterForInOrOf(ParseNodeKind forHeadKind, YieldHandling yieldHandling);
+
+ Node switchStatement(YieldHandling yieldHandling);
+ Node continueStatement(YieldHandling yieldHandling);
+ Node breakStatement(YieldHandling yieldHandling);
+ Node returnStatement(YieldHandling yieldHandling);
+ Node withStatement(YieldHandling yieldHandling);
+ Node throwStatement(YieldHandling yieldHandling);
+ Node tryStatement(YieldHandling yieldHandling);
+ Node catchBlockStatement(YieldHandling yieldHandling, ParseContext::Scope& catchParamScope);
+ Node debuggerStatement();
+
+ Node variableStatement(YieldHandling yieldHandling);
+
+ Node labeledStatement(YieldHandling yieldHandling);
+ Node labeledItem(YieldHandling yieldHandling);
+
+ Node ifStatement(YieldHandling yieldHandling);
+ Node consequentOrAlternative(YieldHandling yieldHandling);
+
+ // While on a |let| TOK_NAME token, examine |next|. Indicate whether
+ // |next|, the next token already gotten with modifier TokenStream::None,
+ // continues a LexicalDeclaration.
+ bool nextTokenContinuesLetDeclaration(TokenKind next, YieldHandling yieldHandling);
+
+ Node lexicalDeclaration(YieldHandling yieldHandling, bool isConst);
+
+ Node importDeclaration();
+ Node exportDeclaration();
+ Node expressionStatement(YieldHandling yieldHandling,
+ InvokedPrediction invoked = PredictUninvoked);
+
+ // Declaration parsing. The main entrypoint is Parser::declarationList,
+ // with sub-functionality split out into the remaining methods.
+
+ // |blockScope| may be non-null only when |kind| corresponds to a lexical
+ // declaration (that is, PNK_LET or PNK_CONST).
+ //
+ // The for* parameters, for normal declarations, should be null/ignored.
+ // They should be non-null only when Parser::forHeadStart parses a
+ // declaration at the start of a for-loop head.
+ //
+ // In this case, on success |*forHeadKind| is PNK_FORHEAD, PNK_FORIN, or
+ // PNK_FOROF, corresponding to the three for-loop kinds. The precise value
+ // indicates what was parsed.
+ //
+ // If parsing recognized a for(;;) loop, the next token is the ';' within
+ // the loop-head that separates the init/test parts.
+ //
+ // Otherwise, for for-in/of loops, the next token is the ')' ending the
+ // loop-head. Additionally, the expression that the loop iterates over was
+ // parsed into |*forInOrOfExpression|.
+ Node declarationList(YieldHandling yieldHandling,
+ ParseNodeKind kind,
+ ParseNodeKind* forHeadKind = nullptr,
+ Node* forInOrOfExpression = nullptr);
+
+ // The items in a declaration list are either patterns or names, with or
+ // without initializers. These two methods parse a single pattern/name and
+ // any associated initializer -- and if parsing an |initialDeclaration|
+ // will, if parsing in a for-loop head (as specified by |forHeadKind| being
+ // non-null), consume additional tokens up to the closing ')' in a
+ // for-in/of loop head, returning the iterated expression in
+ // |*forInOrOfExpression|. (An "initial declaration" is the first
+ // declaration in a declaration list: |a| but not |b| in |var a, b|, |{c}|
+ // but not |d| in |let {c} = 3, d|.)
+ Node declarationPattern(Node decl, DeclarationKind declKind, TokenKind tt,
+ bool initialDeclaration, YieldHandling yieldHandling,
+ ParseNodeKind* forHeadKind, Node* forInOrOfExpression);
+ Node declarationName(Node decl, DeclarationKind declKind, TokenKind tt,
+ bool initialDeclaration, YieldHandling yieldHandling,
+ ParseNodeKind* forHeadKind, Node* forInOrOfExpression);
+
+ // Having parsed a name (not found in a destructuring pattern) declared by
+ // a declaration, with the current token being the '=' separating the name
+ // from its initializer, parse and bind that initializer -- and possibly
+ // consume trailing in/of and subsequent expression, if so directed by
+ // |forHeadKind|.
+ bool initializerInNameDeclaration(Node decl, Node binding, Handle<PropertyName*> name,
+ DeclarationKind declKind, bool initialDeclaration,
+ YieldHandling yieldHandling, ParseNodeKind* forHeadKind,
+ Node* forInOrOfExpression);
+
+ Node expr(InHandling inHandling, YieldHandling yieldHandling,
+ TripledotHandling tripledotHandling, PossibleError* possibleError = nullptr,
+ InvokedPrediction invoked = PredictUninvoked);
+ Node assignExpr(InHandling inHandling, YieldHandling yieldHandling,
+ TripledotHandling tripledotHandling, PossibleError* possibleError = nullptr,
+ InvokedPrediction invoked = PredictUninvoked);
+ Node assignExprWithoutYieldOrAwait(YieldHandling yieldHandling);
+ Node yieldExpression(InHandling inHandling);
+ Node condExpr1(InHandling inHandling, YieldHandling yieldHandling,
+ TripledotHandling tripledotHandling,
+ PossibleError* possibleError,
+ InvokedPrediction invoked = PredictUninvoked);
+ Node orExpr1(InHandling inHandling, YieldHandling yieldHandling,
+ TripledotHandling tripledotHandling,
+ PossibleError* possibleError,
+ InvokedPrediction invoked = PredictUninvoked);
+ Node unaryExpr(YieldHandling yieldHandling, TripledotHandling tripledotHandling,
+ PossibleError* possibleError = nullptr,
+ InvokedPrediction invoked = PredictUninvoked);
+ Node memberExpr(YieldHandling yieldHandling, TripledotHandling tripledotHandling,
+ TokenKind tt, bool allowCallSyntax = true,
+ PossibleError* possibleError = nullptr,
+ InvokedPrediction invoked = PredictUninvoked);
+ Node primaryExpr(YieldHandling yieldHandling, TripledotHandling tripledotHandling,
+ TokenKind tt, PossibleError* possibleError,
+ InvokedPrediction invoked = PredictUninvoked);
+ Node exprInParens(InHandling inHandling, YieldHandling yieldHandling,
+ TripledotHandling tripledotHandling, PossibleError* possibleError = nullptr);
+
+ bool tryNewTarget(Node& newTarget);
+ bool checkAndMarkSuperScope();
+
+ Node methodDefinition(PropertyType propType, HandleAtom funName);
+
+ /*
+ * Additional JS parsers.
+ */
+ bool functionArguments(YieldHandling yieldHandling, FunctionSyntaxKind kind,
+ Node funcpn);
+
+ Node functionDefinition(InHandling inHandling, YieldHandling yieldHandling, HandleAtom name,
+ FunctionSyntaxKind kind,
+ GeneratorKind generatorKind, FunctionAsyncKind asyncKind,
+ InvokedPrediction invoked = PredictUninvoked);
+
+ // Parse a function body. Pass StatementListBody if the body is a list of
+ // statements; pass ExpressionBody if the body is a single expression.
+ enum FunctionBodyType { StatementListBody, ExpressionBody };
+ Node functionBody(InHandling inHandling, YieldHandling yieldHandling, FunctionSyntaxKind kind,
+ FunctionBodyType type);
+
+ Node unaryOpExpr(YieldHandling yieldHandling, ParseNodeKind kind, JSOp op, uint32_t begin);
+
+ Node condition(InHandling inHandling, YieldHandling yieldHandling);
+
+ /* comprehensions */
+ Node generatorComprehensionLambda(unsigned begin);
+ Node comprehensionFor(GeneratorKind comprehensionKind);
+ Node comprehensionIf(GeneratorKind comprehensionKind);
+ Node comprehensionTail(GeneratorKind comprehensionKind);
+ Node comprehension(GeneratorKind comprehensionKind);
+ Node arrayComprehension(uint32_t begin);
+ Node generatorComprehension(uint32_t begin);
+
+ bool argumentList(YieldHandling yieldHandling, Node listNode, bool* isSpread,
+ PossibleError* possibleError = nullptr);
+ Node destructuringDeclaration(DeclarationKind kind, YieldHandling yieldHandling,
+ TokenKind tt);
+ Node destructuringDeclarationWithoutYieldOrAwait(DeclarationKind kind, YieldHandling yieldHandling,
+ TokenKind tt);
+
+ bool namedImportsOrNamespaceImport(TokenKind tt, Node importSpecSet);
+ bool checkExportedName(JSAtom* exportName);
+ bool checkExportedNamesForDeclaration(Node node);
+
+ enum ClassContext { ClassStatement, ClassExpression };
+ Node classDefinition(YieldHandling yieldHandling, ClassContext classContext,
+ DefaultHandling defaultHandling);
+
+ PropertyName* labelOrIdentifierReference(YieldHandling yieldHandling,
+ bool yieldTokenizedAsName);
+
+ PropertyName* labelIdentifier(YieldHandling yieldHandling) {
+ return labelOrIdentifierReference(yieldHandling, false);
+ }
+
+ PropertyName* identifierReference(YieldHandling yieldHandling,
+ bool yieldTokenizedAsName = false)
+ {
+ return labelOrIdentifierReference(yieldHandling, yieldTokenizedAsName);
+ }
+
+ PropertyName* importedBinding() {
+ return bindingIdentifier(YieldIsName);
+ }
+
+ Node identifierReference(Handle<PropertyName*> name);
+
+ bool matchLabel(YieldHandling yieldHandling, MutableHandle<PropertyName*> label);
+
+ bool allowsForEachIn() {
+#if !JS_HAS_FOR_EACH_IN
+ return false;
+#else
+ return versionNumber() >= JSVERSION_1_6;
+#endif
+ }
+
+ enum AssignmentFlavor {
+ PlainAssignment,
+ CompoundAssignment,
+ KeyedDestructuringAssignment,
+ IncrementAssignment,
+ DecrementAssignment,
+ ForInOrOfTarget
+ };
+
+ bool checkAndMarkAsAssignmentLhs(Node pn, AssignmentFlavor flavor,
+ PossibleError* possibleError=nullptr);
+ bool matchInOrOf(bool* isForInp, bool* isForOfp);
+
+ bool hasUsedFunctionSpecialName(HandlePropertyName name);
+ bool declareFunctionArgumentsObject();
+ bool declareFunctionThis();
+ Node newInternalDotName(HandlePropertyName name);
+ Node newThisName();
+ Node newDotGeneratorName();
+ bool declareDotGeneratorName();
+
+ bool checkFunctionDefinition(HandleAtom funAtom, Node pn, FunctionSyntaxKind kind,
+ GeneratorKind generatorKind, bool* tryAnnexB);
+ bool skipLazyInnerFunction(Node pn, FunctionSyntaxKind kind, bool tryAnnexB);
+ bool innerFunction(Node pn, ParseContext* outerpc, HandleFunction fun,
+ InHandling inHandling, YieldHandling yieldHandling,
+ FunctionSyntaxKind kind,
+ GeneratorKind generatorKind, FunctionAsyncKind asyncKind, bool tryAnnexB,
+ Directives inheritedDirectives, Directives* newDirectives);
+ bool trySyntaxParseInnerFunction(Node pn, HandleFunction fun, InHandling inHandling,
+ YieldHandling yieldHandling, FunctionSyntaxKind kind,
+ GeneratorKind generatorKind, FunctionAsyncKind asyncKind,
+ bool tryAnnexB,
+ Directives inheritedDirectives, Directives* newDirectives);
+ bool finishFunctionScopes();
+ bool finishFunction();
+ bool leaveInnerFunction(ParseContext* outerpc);
+
+ public:
+ enum FunctionCallBehavior {
+ PermitAssignmentToFunctionCalls,
+ ForbidAssignmentToFunctionCalls
+ };
+
+ bool isValidSimpleAssignmentTarget(Node node,
+ FunctionCallBehavior behavior = ForbidAssignmentToFunctionCalls);
+
+ private:
+ bool reportIfArgumentsEvalTarget(Node nameNode);
+ bool reportIfNotValidSimpleAssignmentTarget(Node target, AssignmentFlavor flavor);
+
+ bool checkAndMarkAsIncOperand(Node kid, AssignmentFlavor flavor);
+ bool checkStrictAssignment(Node lhs);
+ bool checkStrictBinding(PropertyName* name, TokenPos pos);
+
+ bool hasValidSimpleStrictParameterNames();
+
+ bool isValidStrictBinding(PropertyName* name);
+
+ void reportRedeclaration(HandlePropertyName name, DeclarationKind kind, TokenPos pos);
+ bool notePositionalFormalParameter(Node fn, HandlePropertyName name,
+ bool disallowDuplicateParams, bool* duplicatedParam);
+ bool noteDestructuredPositionalFormalParameter(Node fn, Node destruct);
+ mozilla::Maybe<DeclarationKind> isVarRedeclaredInEval(HandlePropertyName name,
+ DeclarationKind kind);
+ bool tryDeclareVar(HandlePropertyName name, DeclarationKind kind,
+ mozilla::Maybe<DeclarationKind>* redeclaredKind);
+ bool tryDeclareVarForAnnexBLexicalFunction(HandlePropertyName name, bool* tryAnnexB);
+ bool checkLexicalDeclarationDirectlyWithinBlock(ParseContext::Statement& stmt,
+ DeclarationKind kind, TokenPos pos);
+ bool noteDeclaredName(HandlePropertyName name, DeclarationKind kind, TokenPos pos);
+ bool noteUsedName(HandlePropertyName name);
+ bool hasUsedName(HandlePropertyName name);
+
+ // Required on Scope exit.
+ bool propagateFreeNamesAndMarkClosedOverBindings(ParseContext::Scope& scope);
+
+ mozilla::Maybe<GlobalScope::Data*> newGlobalScopeData(ParseContext::Scope& scope);
+ mozilla::Maybe<ModuleScope::Data*> newModuleScopeData(ParseContext::Scope& scope);
+ mozilla::Maybe<EvalScope::Data*> newEvalScopeData(ParseContext::Scope& scope);
+ mozilla::Maybe<FunctionScope::Data*> newFunctionScopeData(ParseContext::Scope& scope,
+ bool hasParameterExprs);
+ mozilla::Maybe<VarScope::Data*> newVarScopeData(ParseContext::Scope& scope);
+ mozilla::Maybe<LexicalScope::Data*> newLexicalScopeData(ParseContext::Scope& scope);
+ Node finishLexicalScope(ParseContext::Scope& scope, Node body);
+
+ Node propertyName(YieldHandling yieldHandling, Node propList,
+ PropertyType* propType, MutableHandleAtom propAtom);
+ Node computedPropertyName(YieldHandling yieldHandling, Node literal);
+ Node arrayInitializer(YieldHandling yieldHandling, PossibleError* possibleError);
+ Node newRegExp();
+
+ Node objectLiteral(YieldHandling yieldHandling, PossibleError* possibleError);
+
+ // Top-level entrypoint into destructuring pattern checking/name-analyzing.
+ bool checkDestructuringPattern(Node pattern, mozilla::Maybe<DeclarationKind> maybeDecl,
+ PossibleError* possibleError = nullptr);
+
+ // Recursive methods for checking/name-analyzing subcomponents of a
+ // destructuring pattern. The array/object methods *must* be passed arrays
+ // or objects. The name method may be passed anything but will report an
+ // error if not passed a name.
+ bool checkDestructuringArray(Node arrayPattern, mozilla::Maybe<DeclarationKind> maybeDecl);
+ bool checkDestructuringObject(Node objectPattern, mozilla::Maybe<DeclarationKind> maybeDecl);
+ bool checkDestructuringName(Node expr, mozilla::Maybe<DeclarationKind> maybeDecl);
+
+ bool checkAssignmentToCall(Node node, unsigned errnum);
+
+ Node newNumber(const Token& tok) {
+ return handler.newNumber(tok.number(), tok.decimalPoint(), tok.pos);
+ }
+
+ static Node null() { return ParseHandler::null(); }
+
+ bool reportBadReturn(Node pn, ParseReportKind kind, unsigned errnum, unsigned anonerrnum);
+
+ JSAtom* prefixAccessorName(PropertyType propType, HandleAtom propAtom);
+
+ TokenPos pos() const { return tokenStream.currentToken().pos; }
+
+ bool asmJS(Node list);
+
+ void addTelemetry(JSCompartment::DeprecatedLanguageExtension e);
+
+ bool warnOnceAboutExprClosure();
+ bool warnOnceAboutForEach();
+};
+
+} /* namespace frontend */
+} /* namespace js */
+
+#endif /* frontend_Parser_h */