gecko-dev/dom/bindings/BindingUtils.cpp

3821 строка
118 KiB
C++
Исходник Обычный вид История

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this file,
* You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "BindingUtils.h"
#include <algorithm>
#include <stdarg.h>
#include "mozilla/Assertions.h"
#include "mozilla/DebugOnly.h"
#include "mozilla/FloatingPoint.h"
#include "mozilla/Preferences.h"
#include "mozilla/Unused.h"
#include "mozilla/UseCounter.h"
#include "AccessCheck.h"
#include "jsfriendapi.h"
#include "nsContentCreatorFunctions.h"
#include "nsContentUtils.h"
#include "nsGlobalWindow.h"
#include "nsIDocShell.h"
#include "nsIDOMGlobalPropertyInitializer.h"
#include "nsIParserService.h"
#include "nsIPermissionManager.h"
#include "nsIPrincipal.h"
#include "nsIXPConnect.h"
#include "nsUTF8Utils.h"
#include "WorkerPrivate.h"
#include "WorkerRunnable.h"
#include "WrapperFactory.h"
#include "xpcprivate.h"
#include "XrayWrapper.h"
#include "nsPrintfCString.h"
#include "mozilla/Sprintf.h"
#include "nsGlobalWindow.h"
#include "mozilla/dom/ScriptSettings.h"
#include "mozilla/dom/CustomElementRegistry.h"
#include "mozilla/dom/DOMError.h"
#include "mozilla/dom/DOMErrorBinding.h"
#include "mozilla/dom/DOMException.h"
#include "mozilla/dom/ElementBinding.h"
#include "mozilla/dom/HTMLObjectElement.h"
#include "mozilla/dom/HTMLObjectElementBinding.h"
#include "mozilla/dom/HTMLEmbedElement.h"
#include "mozilla/dom/HTMLElementBinding.h"
#include "mozilla/dom/HTMLEmbedElementBinding.h"
#include "mozilla/dom/Promise.h"
#include "mozilla/dom/ResolveSystemBinding.h"
#include "mozilla/dom/WebIDLGlobalNameHash.h"
#include "mozilla/dom/WorkerPrivate.h"
#include "mozilla/dom/WorkerScope.h"
#include "mozilla/dom/XrayExpandoClass.h"
#include "mozilla/jsipc/CrossProcessObjectWrappers.h"
#include "nsDOMClassInfo.h"
#include "ipc/ErrorIPCUtils.h"
#include "mozilla/UseCounter.h"
#include "mozilla/dom/DocGroup.h"
namespace mozilla {
namespace dom {
using namespace workers;
// Forward declare GetConstructorObject methods.
#define HTML_TAG(_tag, _classname, _interfacename) \
namespace HTML##_interfacename##ElementBinding { \
JSObject* GetConstructorObject(JSContext*); \
}
#define HTML_OTHER(_tag)
#include "nsHTMLTagList.h"
#undef HTML_TAG
#undef HTML_OTHER
typedef JSObject* (*constructorGetterCallback)(JSContext*);
// Mapping of html tag and GetConstructorObject methods.
#define HTML_TAG(_tag, _classname, _interfacename) HTML##_interfacename##ElementBinding::GetConstructorObject,
#define HTML_OTHER(_tag) nullptr,
// We use eHTMLTag_foo (where foo is the tag) which is defined in nsHTMLTags.h
// to index into this array.
static const constructorGetterCallback sConstructorGetterCallback[] = {
HTMLUnknownElementBinding::GetConstructorObject,
#include "nsHTMLTagList.h"
#undef HTML_TAG
#undef HTML_OTHER
};
const JSErrorFormatString ErrorFormatString[] = {
#define MSG_DEF(_name, _argc, _exn, _str) \
{ #_name, _str, _argc, _exn },
#include "mozilla/dom/Errors.msg"
#undef MSG_DEF
};
#define MSG_DEF(_name, _argc, _exn, _str) \
static_assert(_argc < JS::MaxNumErrorArguments, \
#_name " must only have as many error arguments as the JS engine can support");
#include "mozilla/dom/Errors.msg"
#undef MSG_DEF
const JSErrorFormatString*
GetErrorMessage(void* aUserRef, const unsigned aErrorNumber)
{
MOZ_ASSERT(aErrorNumber < ArrayLength(ErrorFormatString));
return &ErrorFormatString[aErrorNumber];
}
uint16_t
GetErrorArgCount(const ErrNum aErrorNumber)
{
return GetErrorMessage(nullptr, aErrorNumber)->argCount;
}
void
binding_detail::ThrowErrorMessage(JSContext* aCx, const unsigned aErrorNumber, ...)
{
va_list ap;
va_start(ap, aErrorNumber);
JS_ReportErrorNumberUTF8VA(aCx, GetErrorMessage, nullptr, aErrorNumber, ap);
va_end(ap);
}
bool
ThrowInvalidThis(JSContext* aCx, const JS::CallArgs& aArgs,
bool aSecurityError, const char* aInterfaceName)
{
NS_ConvertASCIItoUTF16 ifaceName(aInterfaceName);
// This should only be called for DOM methods/getters/setters, which
// are JSNative-backed functions, so we can assume that
// JS_ValueToFunction and JS_GetFunctionDisplayId will both return
// non-null and that JS_GetStringCharsZ returns non-null.
JS::Rooted<JSFunction*> func(aCx, JS_ValueToFunction(aCx, aArgs.calleev()));
MOZ_ASSERT(func);
JS::Rooted<JSString*> funcName(aCx, JS_GetFunctionDisplayId(func));
MOZ_ASSERT(funcName);
nsAutoJSString funcNameStr;
if (!funcNameStr.init(aCx, funcName)) {
return false;
}
const ErrNum errorNumber = aSecurityError ?
MSG_METHOD_THIS_UNWRAPPING_DENIED :
MSG_METHOD_THIS_DOES_NOT_IMPLEMENT_INTERFACE;
MOZ_RELEASE_ASSERT(GetErrorArgCount(errorNumber) <= 2);
JS_ReportErrorNumberUC(aCx, GetErrorMessage, nullptr,
static_cast<const unsigned>(errorNumber),
funcNameStr.get(), ifaceName.get());
return false;
}
bool
ThrowInvalidThis(JSContext* aCx, const JS::CallArgs& aArgs,
bool aSecurityError,
prototypes::ID aProtoId)
{
return ThrowInvalidThis(aCx, aArgs, aSecurityError,
NamesOfInterfacesWithProtos(aProtoId));
}
bool
ThrowNoSetterArg(JSContext* aCx, prototypes::ID aProtoId)
{
nsPrintfCString errorMessage("%s attribute setter",
NamesOfInterfacesWithProtos(aProtoId));
return ThrowErrorMessage(aCx, MSG_MISSING_ARGUMENTS, errorMessage.get());
}
} // namespace dom
namespace binding_danger {
template<typename CleanupPolicy>
struct TErrorResult<CleanupPolicy>::Message {
Message() { MOZ_COUNT_CTOR(TErrorResult::Message); }
~Message() { MOZ_COUNT_DTOR(TErrorResult::Message); }
nsTArray<nsString> mArgs;
dom::ErrNum mErrorNumber;
bool HasCorrectNumberOfArguments()
{
return GetErrorArgCount(mErrorNumber) == mArgs.Length();
}
};
template<typename CleanupPolicy>
nsTArray<nsString>&
TErrorResult<CleanupPolicy>::CreateErrorMessageHelper(const dom::ErrNum errorNumber,
nsresult errorType)
{
AssertInOwningThread();
mResult = errorType;
mMessage = new Message();
mMessage->mErrorNumber = errorNumber;
return mMessage->mArgs;
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SerializeMessage(IPC::Message* aMsg) const
{
using namespace IPC;
AssertInOwningThread();
MOZ_ASSERT(mUnionState == HasMessage);
MOZ_ASSERT(mMessage);
WriteParam(aMsg, mMessage->mArgs);
WriteParam(aMsg, mMessage->mErrorNumber);
}
template<typename CleanupPolicy>
bool
TErrorResult<CleanupPolicy>::DeserializeMessage(const IPC::Message* aMsg,
PickleIterator* aIter)
{
using namespace IPC;
AssertInOwningThread();
nsAutoPtr<Message> readMessage(new Message());
if (!ReadParam(aMsg, aIter, &readMessage->mArgs) ||
!ReadParam(aMsg, aIter, &readMessage->mErrorNumber)) {
return false;
}
if (!readMessage->HasCorrectNumberOfArguments()) {
return false;
}
MOZ_ASSERT(mUnionState == HasNothing);
mMessage = readMessage.forget();
#ifdef DEBUG
mUnionState = HasMessage;
#endif // DEBUG
return true;
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SetPendingExceptionWithMessage(JSContext* aCx)
{
AssertInOwningThread();
MOZ_ASSERT(mMessage, "SetPendingExceptionWithMessage() can be called only once");
MOZ_ASSERT(mUnionState == HasMessage);
Message* message = mMessage;
MOZ_RELEASE_ASSERT(message->HasCorrectNumberOfArguments());
const uint32_t argCount = message->mArgs.Length();
const char16_t* args[JS::MaxNumErrorArguments + 1];
for (uint32_t i = 0; i < argCount; ++i) {
args[i] = message->mArgs.ElementAt(i).get();
}
args[argCount] = nullptr;
JS_ReportErrorNumberUCArray(aCx, dom::GetErrorMessage, nullptr,
static_cast<const unsigned>(message->mErrorNumber),
argCount > 0 ? args : nullptr);
ClearMessage();
mResult = NS_OK;
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::ClearMessage()
{
AssertInOwningThread();
MOZ_ASSERT(IsErrorWithMessage());
delete mMessage;
mMessage = nullptr;
#ifdef DEBUG
mUnionState = HasNothing;
#endif // DEBUG
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::ThrowJSException(JSContext* cx, JS::Handle<JS::Value> exn)
{
AssertInOwningThread();
MOZ_ASSERT(mMightHaveUnreportedJSException,
"Why didn't you tell us you planned to throw a JS exception?");
ClearUnionData();
// Make sure mJSException is initialized _before_ we try to root it. But
// don't set it to exn yet, because we don't want to do that until after we
// root.
mJSException.setUndefined();
if (!js::AddRawValueRoot(cx, &mJSException, "TErrorResult::mJSException")) {
// Don't use NS_ERROR_INTERNAL_ERRORRESULT_JS_EXCEPTION, because that
// indicates we have in fact rooted mJSException.
mResult = NS_ERROR_OUT_OF_MEMORY;
} else {
mJSException = exn;
mResult = NS_ERROR_INTERNAL_ERRORRESULT_JS_EXCEPTION;
#ifdef DEBUG
mUnionState = HasJSException;
#endif // DEBUG
}
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SetPendingJSException(JSContext* cx)
{
AssertInOwningThread();
MOZ_ASSERT(!mMightHaveUnreportedJSException,
"Why didn't you tell us you planned to handle JS exceptions?");
MOZ_ASSERT(mUnionState == HasJSException);
JS::Rooted<JS::Value> exception(cx, mJSException);
if (JS_WrapValue(cx, &exception)) {
JS_SetPendingException(cx, exception);
}
mJSException = exception;
// If JS_WrapValue failed, not much we can do about it... No matter
// what, go ahead and unroot mJSException.
js::RemoveRawValueRoot(cx, &mJSException);
mResult = NS_OK;
#ifdef DEBUG
mUnionState = HasNothing;
#endif // DEBUG
}
template<typename CleanupPolicy>
struct TErrorResult<CleanupPolicy>::DOMExceptionInfo {
DOMExceptionInfo(nsresult rv, const nsACString& message)
: mMessage(message)
, mRv(rv)
{}
nsCString mMessage;
nsresult mRv;
};
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SerializeDOMExceptionInfo(IPC::Message* aMsg) const
{
using namespace IPC;
AssertInOwningThread();
MOZ_ASSERT(mDOMExceptionInfo);
MOZ_ASSERT(mUnionState == HasDOMExceptionInfo);
WriteParam(aMsg, mDOMExceptionInfo->mMessage);
WriteParam(aMsg, mDOMExceptionInfo->mRv);
}
template<typename CleanupPolicy>
bool
TErrorResult<CleanupPolicy>::DeserializeDOMExceptionInfo(const IPC::Message* aMsg,
PickleIterator* aIter)
{
using namespace IPC;
AssertInOwningThread();
nsCString message;
nsresult rv;
if (!ReadParam(aMsg, aIter, &message) ||
!ReadParam(aMsg, aIter, &rv)) {
return false;
}
MOZ_ASSERT(mUnionState == HasNothing);
MOZ_ASSERT(IsDOMException());
mDOMExceptionInfo = new DOMExceptionInfo(rv, message);
#ifdef DEBUG
mUnionState = HasDOMExceptionInfo;
#endif // DEBUG
return true;
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::ThrowDOMException(nsresult rv,
const nsACString& message)
{
AssertInOwningThread();
ClearUnionData();
mResult = NS_ERROR_INTERNAL_ERRORRESULT_DOMEXCEPTION;
mDOMExceptionInfo = new DOMExceptionInfo(rv, message);
#ifdef DEBUG
mUnionState = HasDOMExceptionInfo;
#endif
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SetPendingDOMException(JSContext* cx)
{
AssertInOwningThread();
MOZ_ASSERT(mDOMExceptionInfo,
"SetPendingDOMException() can be called only once");
MOZ_ASSERT(mUnionState == HasDOMExceptionInfo);
dom::Throw(cx, mDOMExceptionInfo->mRv, mDOMExceptionInfo->mMessage);
ClearDOMExceptionInfo();
mResult = NS_OK;
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::ClearDOMExceptionInfo()
{
AssertInOwningThread();
MOZ_ASSERT(IsDOMException());
MOZ_ASSERT(mUnionState == HasDOMExceptionInfo || !mDOMExceptionInfo);
delete mDOMExceptionInfo;
mDOMExceptionInfo = nullptr;
#ifdef DEBUG
mUnionState = HasNothing;
#endif // DEBUG
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::ClearUnionData()
{
AssertInOwningThread();
if (IsJSException()) {
JSContext* cx = dom::danger::GetJSContext();
MOZ_ASSERT(cx);
mJSException.setUndefined();
js::RemoveRawValueRoot(cx, &mJSException);
#ifdef DEBUG
mUnionState = HasNothing;
#endif // DEBUG
} else if (IsErrorWithMessage()) {
ClearMessage();
} else if (IsDOMException()) {
ClearDOMExceptionInfo();
}
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SetPendingGenericErrorException(JSContext* cx)
{
AssertInOwningThread();
MOZ_ASSERT(!IsErrorWithMessage());
MOZ_ASSERT(!IsJSException());
MOZ_ASSERT(!IsDOMException());
dom::Throw(cx, ErrorCode());
mResult = NS_OK;
}
template<typename CleanupPolicy>
TErrorResult<CleanupPolicy>&
TErrorResult<CleanupPolicy>::operator=(TErrorResult<CleanupPolicy>&& aRHS)
{
AssertInOwningThread();
aRHS.AssertInOwningThread();
// Clear out any union members we may have right now, before we
// start writing to it.
ClearUnionData();
#ifdef DEBUG
mMightHaveUnreportedJSException = aRHS.mMightHaveUnreportedJSException;
aRHS.mMightHaveUnreportedJSException = false;
#endif
if (aRHS.IsErrorWithMessage()) {
mMessage = aRHS.mMessage;
aRHS.mMessage = nullptr;
} else if (aRHS.IsJSException()) {
JSContext* cx = dom::danger::GetJSContext();
MOZ_ASSERT(cx);
mJSException.setUndefined();
if (!js::AddRawValueRoot(cx, &mJSException, "TErrorResult::mJSException")) {
MOZ_CRASH("Could not root mJSException, we're about to OOM");
}
mJSException = aRHS.mJSException;
aRHS.mJSException.setUndefined();
js::RemoveRawValueRoot(cx, &aRHS.mJSException);
} else if (aRHS.IsDOMException()) {
mDOMExceptionInfo = aRHS.mDOMExceptionInfo;
aRHS.mDOMExceptionInfo = nullptr;
} else {
// Null out the union on both sides for hygiene purposes.
mMessage = aRHS.mMessage = nullptr;
}
#ifdef DEBUG
mUnionState = aRHS.mUnionState;
aRHS.mUnionState = HasNothing;
#endif // DEBUG
// Note: It's important to do this last, since this affects the condition
// checks above!
mResult = aRHS.mResult;
aRHS.mResult = NS_OK;
return *this;
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::CloneTo(TErrorResult& aRv) const
{
AssertInOwningThread();
aRv.AssertInOwningThread();
aRv.ClearUnionData();
aRv.mResult = mResult;
#ifdef DEBUG
aRv.mMightHaveUnreportedJSException = mMightHaveUnreportedJSException;
#endif
if (IsErrorWithMessage()) {
#ifdef DEBUG
aRv.mUnionState = HasMessage;
#endif
aRv.mMessage = new Message();
aRv.mMessage->mArgs = mMessage->mArgs;
aRv.mMessage->mErrorNumber = mMessage->mErrorNumber;
} else if (IsDOMException()) {
#ifdef DEBUG
aRv.mUnionState = HasDOMExceptionInfo;
#endif
aRv.mDOMExceptionInfo = new DOMExceptionInfo(mDOMExceptionInfo->mRv,
mDOMExceptionInfo->mMessage);
} else if (IsJSException()) {
#ifdef DEBUG
aRv.mUnionState = HasJSException;
#endif
JSContext* cx = dom::danger::GetJSContext();
JS::Rooted<JS::Value> exception(cx, mJSException);
aRv.ThrowJSException(cx, exception);
}
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SuppressException()
{
AssertInOwningThread();
WouldReportJSException();
ClearUnionData();
// We don't use AssignErrorCode, because we want to override existing error
// states, which AssignErrorCode is not allowed to do.
mResult = NS_OK;
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::SetPendingException(JSContext* cx)
{
AssertInOwningThread();
if (IsUncatchableException()) {
// Nuke any existing exception on cx, to make sure we're uncatchable.
JS_ClearPendingException(cx);
// Don't do any reporting. Just return, to create an
// uncatchable exception.
mResult = NS_OK;
return;
}
if (IsJSContextException()) {
// Whatever we need to throw is on the JSContext already.
MOZ_ASSERT(JS_IsExceptionPending(cx));
mResult = NS_OK;
return;
}
if (IsErrorWithMessage()) {
SetPendingExceptionWithMessage(cx);
return;
}
if (IsJSException()) {
SetPendingJSException(cx);
return;
}
if (IsDOMException()) {
SetPendingDOMException(cx);
return;
}
SetPendingGenericErrorException(cx);
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::StealExceptionFromJSContext(JSContext* cx)
{
AssertInOwningThread();
MOZ_ASSERT(mMightHaveUnreportedJSException,
"Why didn't you tell us you planned to throw a JS exception?");
JS::Rooted<JS::Value> exn(cx);
if (!JS_GetPendingException(cx, &exn)) {
ThrowUncatchableException();
return;
}
ThrowJSException(cx, exn);
JS_ClearPendingException(cx);
}
template<typename CleanupPolicy>
void
TErrorResult<CleanupPolicy>::NoteJSContextException(JSContext* aCx)
{
AssertInOwningThread();
if (JS_IsExceptionPending(aCx)) {
mResult = NS_ERROR_INTERNAL_ERRORRESULT_EXCEPTION_ON_JSCONTEXT;
} else {
mResult = NS_ERROR_UNCATCHABLE_EXCEPTION;
}
}
template class TErrorResult<JustAssertCleanupPolicy>;
template class TErrorResult<AssertAndSuppressCleanupPolicy>;
template class TErrorResult<JustSuppressCleanupPolicy>;
} // namespace binding_danger
namespace dom {
bool
DefineConstants(JSContext* cx, JS::Handle<JSObject*> obj,
const ConstantSpec* cs)
{
JS::Rooted<JS::Value> value(cx);
for (; cs->name; ++cs) {
value = cs->value;
bool ok =
JS_DefineProperty(cx, obj, cs->name, value,
JSPROP_ENUMERATE | JSPROP_READONLY | JSPROP_PERMANENT);
if (!ok) {
return false;
}
}
return true;
}
static inline bool
Define(JSContext* cx, JS::Handle<JSObject*> obj, const JSFunctionSpec* spec) {
return JS_DefineFunctions(cx, obj, spec);
}
static inline bool
Define(JSContext* cx, JS::Handle<JSObject*> obj, const JSPropertySpec* spec) {
return JS_DefineProperties(cx, obj, spec);
}
static inline bool
Define(JSContext* cx, JS::Handle<JSObject*> obj, const ConstantSpec* spec) {
return DefineConstants(cx, obj, spec);
}
template<typename T>
bool
DefinePrefable(JSContext* cx, JS::Handle<JSObject*> obj,
const Prefable<T>* props)
{
MOZ_ASSERT(props);
MOZ_ASSERT(props->specs);
do {
// Define if enabled
if (props->isEnabled(cx, obj)) {
if (!Define(cx, obj, props->specs)) {
return false;
}
}
} while ((++props)->specs);
return true;
}
bool
DefineUnforgeableMethods(JSContext* cx, JS::Handle<JSObject*> obj,
const Prefable<const JSFunctionSpec>* props)
{
return DefinePrefable(cx, obj, props);
}
bool
DefineUnforgeableAttributes(JSContext* cx, JS::Handle<JSObject*> obj,
const Prefable<const JSPropertySpec>* props)
{
return DefinePrefable(cx, obj, props);
}
// We should use JSFunction objects for interface objects, but we need a custom
// hasInstance hook because we have new interface objects on prototype chains of
// old (XPConnect-based) bindings. We also need Xrays and arbitrary numbers of
// reserved slots (e.g. for named constructors). So we define a custom
// funToString ObjectOps member for interface objects.
JSString*
InterfaceObjectToString(JSContext* aCx, JS::Handle<JSObject*> aObject,
bool /* isToSource */)
{
const js::Class* clasp = js::GetObjectClass(aObject);
MOZ_ASSERT(IsDOMIfaceAndProtoClass(clasp));
const DOMIfaceAndProtoJSClass* ifaceAndProtoJSClass =
DOMIfaceAndProtoJSClass::FromJSClass(clasp);
return JS_NewStringCopyZ(aCx, ifaceAndProtoJSClass->mToString);
}
bool
Constructor(JSContext* cx, unsigned argc, JS::Value* vp)
{
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
const JS::Value& v =
js::GetFunctionNativeReserved(&args.callee(),
CONSTRUCTOR_NATIVE_HOLDER_RESERVED_SLOT);
const JSNativeHolder* nativeHolder =
static_cast<const JSNativeHolder*>(v.toPrivate());
return (nativeHolder->mNative)(cx, argc, vp);
}
static JSObject*
CreateConstructor(JSContext* cx, JS::Handle<JSObject*> global, const char* name,
const JSNativeHolder* nativeHolder, unsigned ctorNargs)
{
JSFunction* fun = js::NewFunctionWithReserved(cx, Constructor, ctorNargs,
JSFUN_CONSTRUCTOR, name);
if (!fun) {
return nullptr;
}
JSObject* constructor = JS_GetFunctionObject(fun);
js::SetFunctionNativeReserved(constructor,
CONSTRUCTOR_NATIVE_HOLDER_RESERVED_SLOT,
js::PrivateValue(const_cast<JSNativeHolder*>(nativeHolder)));
return constructor;
}
static bool
DefineConstructor(JSContext* cx, JS::Handle<JSObject*> global, const char* name,
JS::Handle<JSObject*> constructor)
{
bool alreadyDefined;
if (!JS_AlreadyHasOwnProperty(cx, global, name, &alreadyDefined)) {
return false;
}
// This is Enumerable: False per spec.
return alreadyDefined ||
JS_DefineProperty(cx, global, name, constructor, JSPROP_RESOLVING);
}
static JSObject*
CreateInterfaceObject(JSContext* cx, JS::Handle<JSObject*> global,
JS::Handle<JSObject*> constructorProto,
const js::Class* constructorClass,
unsigned ctorNargs, const NamedConstructor* namedConstructors,
JS::Handle<JSObject*> proto,
const NativeProperties* properties,
const NativeProperties* chromeOnlyProperties,
const char* name, bool defineOnGlobal)
{
JS::Rooted<JSObject*> constructor(cx);
MOZ_ASSERT(constructorProto);
MOZ_ASSERT(constructorClass);
constructor = JS_NewObjectWithGivenProto(cx, Jsvalify(constructorClass),
constructorProto);
if (!constructor) {
return nullptr;
}
if (!JS_DefineProperty(cx, constructor, "length", ctorNargs,
JSPROP_READONLY)) {
return nullptr;
}
// Might as well intern, since we're going to need an atomized
// version of name anyway when we stick our constructor on the
// global.
JS::Rooted<JSString*> nameStr(cx, JS_AtomizeAndPinString(cx, name));
if (!nameStr) {
return nullptr;
}
if (!JS_DefineProperty(cx, constructor, "name", nameStr, JSPROP_READONLY)) {
return nullptr;
}
if (DOMIfaceAndProtoJSClass::FromJSClass(constructorClass)->wantsInterfaceHasInstance) {
JS::Rooted<jsid> hasInstanceId(cx,
SYMBOL_TO_JSID(JS::GetWellKnownSymbol(cx, JS::SymbolCode::hasInstance)));
if (!JS_DefineFunctionById(cx, constructor, hasInstanceId,
InterfaceHasInstance, 1,
// Flags match those of Function[Symbol.hasInstance]
JSPROP_READONLY | JSPROP_PERMANENT)) {
return nullptr;
}
}
if (properties) {
if (properties->HasStaticMethods() &&
!DefinePrefable(cx, constructor, properties->StaticMethods())) {
return nullptr;
}
if (properties->HasStaticAttributes() &&
!DefinePrefable(cx, constructor, properties->StaticAttributes())) {
return nullptr;
}
if (properties->HasConstants() &&
!DefinePrefable(cx, constructor, properties->Constants())) {
return nullptr;
}
}
if (chromeOnlyProperties) {
if (chromeOnlyProperties->HasStaticMethods() &&
!DefinePrefable(cx, constructor,
chromeOnlyProperties->StaticMethods())) {
return nullptr;
}
if (chromeOnlyProperties->HasStaticAttributes() &&
!DefinePrefable(cx, constructor,
chromeOnlyProperties->StaticAttributes())) {
return nullptr;
}
if (chromeOnlyProperties->HasConstants() &&
!DefinePrefable(cx, constructor, chromeOnlyProperties->Constants())) {
return nullptr;
}
}
if (proto && !JS_LinkConstructorAndPrototype(cx, constructor, proto)) {
return nullptr;
}
if (defineOnGlobal && !DefineConstructor(cx, global, name, constructor)) {
return nullptr;
}
if (namedConstructors) {
int namedConstructorSlot = DOM_INTERFACE_SLOTS_BASE;
while (namedConstructors->mName) {
JS::Rooted<JSObject*> namedConstructor(cx,
CreateConstructor(cx, global, namedConstructors->mName,
&namedConstructors->mHolder,
namedConstructors->mNargs));
if (!namedConstructor ||
!JS_DefineProperty(cx, namedConstructor, "prototype",
proto,
JSPROP_PERMANENT | JSPROP_READONLY,
JS_STUBGETTER, JS_STUBSETTER) ||
(defineOnGlobal &&
!DefineConstructor(cx, global, namedConstructors->mName,
namedConstructor))) {
return nullptr;
}
js::SetReservedSlot(constructor, namedConstructorSlot++,
JS::ObjectValue(*namedConstructor));
++namedConstructors;
}
}
return constructor;
}
static JSObject*
CreateInterfacePrototypeObject(JSContext* cx, JS::Handle<JSObject*> global,
JS::Handle<JSObject*> parentProto,
const js::Class* protoClass,
const NativeProperties* properties,
const NativeProperties* chromeOnlyProperties,
const char* const* unscopableNames,
bool isGlobal)
{
JS::Rooted<JSObject*> ourProto(cx,
JS_NewObjectWithUniqueType(cx, Jsvalify(protoClass), parentProto));
if (!ourProto ||
// We don't try to define properties on the global's prototype; those
// properties go on the global itself.
(!isGlobal &&
!DefineProperties(cx, ourProto, properties, chromeOnlyProperties))) {
return nullptr;
}
if (unscopableNames) {
JS::Rooted<JSObject*> unscopableObj(cx, JS_NewPlainObject(cx));
if (!unscopableObj) {
return nullptr;
}
for (; *unscopableNames; ++unscopableNames) {
if (!JS_DefineProperty(cx, unscopableObj, *unscopableNames,
JS::TrueHandleValue, JSPROP_ENUMERATE)) {
return nullptr;
}
}
JS::Rooted<jsid> unscopableId(cx,
SYMBOL_TO_JSID(JS::GetWellKnownSymbol(cx, JS::SymbolCode::unscopables)));
// Readonly and non-enumerable to match Array.prototype.
if (!JS_DefinePropertyById(cx, ourProto, unscopableId, unscopableObj,
JSPROP_READONLY)) {
return nullptr;
}
}
return ourProto;
}
bool
DefineProperties(JSContext* cx, JS::Handle<JSObject*> obj,
const NativeProperties* properties,
const NativeProperties* chromeOnlyProperties)
{
if (properties) {
if (properties->HasMethods() &&
!DefinePrefable(cx, obj, properties->Methods())) {
return false;
}
if (properties->HasAttributes() &&
!DefinePrefable(cx, obj, properties->Attributes())) {
return false;
}
if (properties->HasConstants() &&
!DefinePrefable(cx, obj, properties->Constants())) {
return false;
}
}
if (chromeOnlyProperties) {
if (chromeOnlyProperties->HasMethods() &&
!DefinePrefable(cx, obj, chromeOnlyProperties->Methods())) {
return false;
}
if (chromeOnlyProperties->HasAttributes() &&
!DefinePrefable(cx, obj, chromeOnlyProperties->Attributes())) {
return false;
}
if (chromeOnlyProperties->HasConstants() &&
!DefinePrefable(cx, obj, chromeOnlyProperties->Constants())) {
return false;
}
}
return true;
}
void
CreateInterfaceObjects(JSContext* cx, JS::Handle<JSObject*> global,
JS::Handle<JSObject*> protoProto,
const js::Class* protoClass, JS::Heap<JSObject*>* protoCache,
JS::Handle<JSObject*> constructorProto,
const js::Class* constructorClass,
unsigned ctorNargs, const NamedConstructor* namedConstructors,
JS::Heap<JSObject*>* constructorCache,
const NativeProperties* properties,
const NativeProperties* chromeOnlyProperties,
const char* name, bool defineOnGlobal,
const char* const* unscopableNames,
bool isGlobal)
{
MOZ_ASSERT(protoClass || constructorClass,
"Need at least one class!");
MOZ_ASSERT(!((properties &&
(properties->HasMethods() || properties->HasAttributes())) ||
(chromeOnlyProperties &&
(chromeOnlyProperties->HasMethods() ||
chromeOnlyProperties->HasAttributes()))) || protoClass,
"Methods or properties but no protoClass!");
MOZ_ASSERT(!((properties &&
(properties->HasStaticMethods() ||
properties->HasStaticAttributes())) ||
(chromeOnlyProperties &&
(chromeOnlyProperties->HasStaticMethods() ||
chromeOnlyProperties->HasStaticAttributes()))) ||
constructorClass,
"Static methods but no constructorClass!");
MOZ_ASSERT(bool(name) == bool(constructorClass),
"Must have name precisely when we have an interface object");
MOZ_ASSERT(!protoClass == !protoCache,
"If, and only if, there is an interface prototype object we need "
"to cache it");
MOZ_ASSERT(bool(constructorClass) == bool(constructorCache),
"If, and only if, there is an interface object we need to cache "
"it");
MOZ_ASSERT(constructorProto || !constructorClass,
"Must have a constructor proto if we plan to create a constructor "
"object");
JS::Rooted<JSObject*> proto(cx);
if (protoClass) {
proto =
CreateInterfacePrototypeObject(cx, global, protoProto, protoClass,
properties, chromeOnlyProperties,
unscopableNames, isGlobal);
if (!proto) {
return;
}
*protoCache = proto;
}
else {
MOZ_ASSERT(!proto);
}
JSObject* interface;
if (constructorClass) {
interface = CreateInterfaceObject(cx, global, constructorProto,
constructorClass, ctorNargs,
namedConstructors, proto, properties,
chromeOnlyProperties, name,
defineOnGlobal);
if (!interface) {
if (protoCache) {
// If we fail we need to make sure to clear the value of protoCache we
// set above.
*protoCache = nullptr;
}
return;
}
*constructorCache = interface;
}
}
bool
NativeInterface2JSObjectAndThrowIfFailed(JSContext* aCx,
JS::Handle<JSObject*> aScope,
JS::MutableHandle<JS::Value> aRetval,
xpcObjectHelper& aHelper,
const nsIID* aIID,
bool aAllowNativeWrapper)
{
js::AssertSameCompartment(aCx, aScope);
nsresult rv;
// Inline some logic from XPCConvert::NativeInterfaceToJSObject that we need
// on all threads.
nsWrapperCache *cache = aHelper.GetWrapperCache();
if (cache && cache->IsDOMBinding()) {
JS::Rooted<JSObject*> obj(aCx, cache->GetWrapper());
if (!obj) {
obj = cache->WrapObject(aCx, nullptr);
}
if (obj && aAllowNativeWrapper && !JS_WrapObject(aCx, &obj)) {
return false;
}
if (obj) {
aRetval.setObject(*obj);
return true;
}
}
MOZ_ASSERT(NS_IsMainThread());
if (!XPCConvert::NativeInterface2JSObject(aRetval, nullptr, aHelper, aIID,
aAllowNativeWrapper, &rv)) {
// I can't tell if NativeInterface2JSObject throws JS exceptions
// or not. This is a sloppy stab at the right semantics; the
// method really ought to be fixed to behave consistently.
if (!JS_IsExceptionPending(aCx)) {
Throw(aCx, NS_FAILED(rv) ? rv : NS_ERROR_UNEXPECTED);
}
return false;
}
return true;
}
bool
TryPreserveWrapper(JSObject* obj)
{
MOZ_ASSERT(IsDOMObject(obj));
if (nsISupports* native = UnwrapDOMObjectToISupports(obj)) {
nsWrapperCache* cache = nullptr;
CallQueryInterface(native, &cache);
if (cache) {
cache->PreserveWrapper(native);
}
return true;
}
// If this DOMClass is not cycle collected, then it isn't wrappercached,
// so it does not need to be preserved. If it is cycle collected, then
// we can't tell if it is wrappercached or not, so we just return false.
const DOMJSClass* domClass = GetDOMClass(obj);
return domClass && !domClass->mParticipant;
}
// Can only be called with a DOM JSClass.
bool
InstanceClassHasProtoAtDepth(const js::Class* clasp,
uint32_t protoID, uint32_t depth)
{
const DOMJSClass* domClass = DOMJSClass::FromJSClass(clasp);
return static_cast<uint32_t>(domClass->mInterfaceChain[depth]) == protoID;
}
// Only set allowNativeWrapper to false if you really know you need it, if in
// doubt use true. Setting it to false disables security wrappers.
bool
XPCOMObjectToJsval(JSContext* cx, JS::Handle<JSObject*> scope,
xpcObjectHelper& helper, const nsIID* iid,
bool allowNativeWrapper, JS::MutableHandle<JS::Value> rval)
{
if (!NativeInterface2JSObjectAndThrowIfFailed(cx, scope, rval, helper, iid,
allowNativeWrapper)) {
return false;
}
#ifdef DEBUG
JSObject* jsobj = rval.toObjectOrNull();
if (jsobj &&
js::GetGlobalForObjectCrossCompartment(jsobj) == jsobj) {
NS_ASSERTION(js::GetObjectClass(jsobj)->flags & JSCLASS_IS_GLOBAL,
"Why did we recreate this wrapper?");
}
#endif
return true;
}
bool
VariantToJsval(JSContext* aCx, nsIVariant* aVariant,
JS::MutableHandle<JS::Value> aRetval)
{
nsresult rv;
if (!XPCVariant::VariantDataToJS(aVariant, &rv, aRetval)) {
// Does it throw? Who knows
if (!JS_IsExceptionPending(aCx)) {
Throw(aCx, NS_FAILED(rv) ? rv : NS_ERROR_UNEXPECTED);
}
return false;
}
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
static int
CompareIdsAtIndices(const void* aElement1, const void* aElement2, void* aClosure)
{
const uint16_t index1 = *static_cast<const uint16_t*>(aElement1);
const uint16_t index2 = *static_cast<const uint16_t*>(aElement2);
const PropertyInfo* infos = static_cast<PropertyInfo*>(aClosure);
MOZ_ASSERT(JSID_BITS(infos[index1].id) != JSID_BITS(infos[index2].id));
return JSID_BITS(infos[index1].id) < JSID_BITS(infos[index2].id) ? -1 : 1;
}
template <typename SpecT>
static bool
InitIdsInternal(JSContext* cx, const Prefable<SpecT>* pref, PropertyInfo* infos,
PropertyType type)
{
MOZ_ASSERT(pref);
MOZ_ASSERT(pref->specs);
// Index of the Prefable that contains the id for the current PropertyInfo.
uint32_t prefIndex = 0;
do {
// We ignore whether the set of ids is enabled and just intern all the IDs,
// because this is only done once per application runtime.
const SpecT* spec = pref->specs;
// Index of the property/function/constant spec for our current PropertyInfo
// in the "specs" array of the relevant Prefable.
uint32_t specIndex = 0;
do {
if (!JS::PropertySpecNameToPermanentId(cx, spec->name, &infos->id)) {
return false;
}
infos->type = type;
infos->prefIndex = prefIndex;
infos->specIndex = specIndex++;
++infos;
} while ((++spec)->name);
++prefIndex;
} while ((++pref)->specs);
return true;
}
#define INIT_IDS_IF_DEFINED(TypeName) { \
if (nativeProperties->Has##TypeName##s() && \
!InitIdsInternal(cx, \
nativeProperties->TypeName##s(), \
nativeProperties->TypeName##PropertyInfos(), \
e##TypeName)) { \
return false; \
} \
}
bool
InitIds(JSContext* cx, const NativeProperties* nativeProperties)
{
INIT_IDS_IF_DEFINED(StaticMethod);
INIT_IDS_IF_DEFINED(StaticAttribute);
INIT_IDS_IF_DEFINED(Method);
INIT_IDS_IF_DEFINED(Attribute);
INIT_IDS_IF_DEFINED(UnforgeableMethod);
INIT_IDS_IF_DEFINED(UnforgeableAttribute);
INIT_IDS_IF_DEFINED(Constant);
// Initialize and sort the index array.
uint16_t* indices = nativeProperties->sortedPropertyIndices;
for (unsigned int i = 0; i < nativeProperties->propertyInfoCount; ++i) {
indices[i] = i;
}
// CompareIdsAtIndices() doesn't actually modify the PropertyInfo array, so
// the const_cast here is OK in spite of the signature of NS_QuickSort().
NS_QuickSort(indices, nativeProperties->propertyInfoCount, sizeof(uint16_t),
CompareIdsAtIndices,
const_cast<PropertyInfo*>(nativeProperties->PropertyInfos()));
return true;
}
#undef INIT_IDS_IF_DEFINED
bool
QueryInterface(JSContext* cx, unsigned argc, JS::Value* vp)
{
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
JS::Rooted<JS::Value> thisv(cx, JS_THIS(cx, vp));
if (thisv.isNull())
return false;
// Get the object. It might be a security wrapper, in which case we do a checked
// unwrap.
JS::Rooted<JSObject*> origObj(cx, &thisv.toObject());
JS::Rooted<JSObject*> obj(cx, js::CheckedUnwrap(origObj,
/* stopAtWindowProxy = */ false));
if (!obj) {
JS_ReportErrorASCII(cx, "Permission denied to access object");
return false;
}
// Switch this to UnwrapDOMObjectToISupports once our global objects are
// using new bindings.
nsCOMPtr<nsISupports> native;
UnwrapArg<nsISupports>(cx, obj, getter_AddRefs(native));
if (!native) {
return Throw(cx, NS_ERROR_FAILURE);
}
if (argc < 1) {
return Throw(cx, NS_ERROR_XPC_NOT_ENOUGH_ARGS);
}
if (!args[0].isObject()) {
return Throw(cx, NS_ERROR_XPC_BAD_CONVERT_JS);
}
nsCOMPtr<nsIJSID> iid;
obj = &args[0].toObject();
if (NS_FAILED(UnwrapArg<nsIJSID>(cx, obj, getter_AddRefs(iid)))) {
return Throw(cx, NS_ERROR_XPC_BAD_CONVERT_JS);
}
MOZ_ASSERT(iid);
if (iid->GetID()->Equals(NS_GET_IID(nsIClassInfo))) {
nsresult rv;
nsCOMPtr<nsIClassInfo> ci = do_QueryInterface(native, &rv);
if (NS_FAILED(rv)) {
return Throw(cx, rv);
}
return WrapObject(cx, ci, &NS_GET_IID(nsIClassInfo), args.rval());
}
nsCOMPtr<nsISupports> unused;
nsresult rv = native->QueryInterface(*iid->GetID(), getter_AddRefs(unused));
if (NS_FAILED(rv)) {
return Throw(cx, rv);
}
*vp = thisv;
return true;
}
void
GetInterfaceImpl(JSContext* aCx, nsIInterfaceRequestor* aRequestor,
nsWrapperCache* aCache, nsIJSID* aIID,
JS::MutableHandle<JS::Value> aRetval, ErrorResult& aError)
{
const nsID* iid = aIID->GetID();
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<nsISupports> result;
aError = aRequestor->GetInterface(*iid, getter_AddRefs(result));
if (aError.Failed()) {
return;
}
if (!WrapObject(aCx, result, iid, aRetval)) {
aError.Throw(NS_ERROR_FAILURE);
}
}
bool
ThrowingConstructor(JSContext* cx, unsigned argc, JS::Value* vp)
{
return ThrowErrorMessage(cx, MSG_ILLEGAL_CONSTRUCTOR);
}
bool
ThrowConstructorWithoutNew(JSContext* cx, const char* name)
{
return ThrowErrorMessage(cx, MSG_CONSTRUCTOR_WITHOUT_NEW, name);
}
inline const NativePropertyHooks*
GetNativePropertyHooksFromConstructorFunction(JS::Handle<JSObject*> obj)
{
MOZ_ASSERT(JS_IsNativeFunction(obj, Constructor));
const JS::Value& v =
js::GetFunctionNativeReserved(obj,
CONSTRUCTOR_NATIVE_HOLDER_RESERVED_SLOT);
const JSNativeHolder* nativeHolder =
static_cast<const JSNativeHolder*>(v.toPrivate());
return nativeHolder->mPropertyHooks;
}
inline const NativePropertyHooks*
GetNativePropertyHooks(JSContext *cx, JS::Handle<JSObject*> obj,
DOMObjectType& type)
{
const js::Class* clasp = js::GetObjectClass(obj);
const DOMJSClass* domClass = GetDOMClass(clasp);
if (domClass) {
bool isGlobal = (clasp->flags & JSCLASS_DOM_GLOBAL) != 0;
type = isGlobal ? eGlobalInstance : eInstance;
return domClass->mNativeHooks;
}
if (JS_ObjectIsFunction(cx, obj)) {
type = eInterface;
return GetNativePropertyHooksFromConstructorFunction(obj);
}
MOZ_ASSERT(IsDOMIfaceAndProtoClass(js::GetObjectClass(obj)));
const DOMIfaceAndProtoJSClass* ifaceAndProtoJSClass =
DOMIfaceAndProtoJSClass::FromJSClass(js::GetObjectClass(obj));
type = ifaceAndProtoJSClass->mType;
return ifaceAndProtoJSClass->mNativeHooks;
}
static JSObject*
XrayCreateFunction(JSContext* cx, JS::Handle<JSObject*> wrapper,
JSNativeWrapper native, unsigned nargs, JS::Handle<jsid> id)
{
JSFunction* fun;
if (JSID_IS_STRING(id)) {
fun = js::NewFunctionByIdWithReserved(cx, native.op, nargs, 0, id);
} else {
// Can't pass this id (probably a symbol) to NewFunctionByIdWithReserved;
// just use an empty name for lack of anything better.
fun = js::NewFunctionWithReserved(cx, native.op, nargs, 0, nullptr);
}
if (!fun) {
return nullptr;
}
SET_JITINFO(fun, native.info);
JSObject* obj = JS_GetFunctionObject(fun);
js::SetFunctionNativeReserved(obj, XRAY_DOM_FUNCTION_PARENT_WRAPPER_SLOT,
JS::ObjectValue(*wrapper));
#ifdef DEBUG
js::SetFunctionNativeReserved(obj, XRAY_DOM_FUNCTION_NATIVE_SLOT_FOR_SELF,
JS::ObjectValue(*obj));
#endif
return obj;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
struct IdToIndexComparator
{
// The id we're searching for.
const jsid& mId;
// The list of ids we're searching in.
const PropertyInfo* mInfos;
explicit IdToIndexComparator(const jsid& aId, const PropertyInfo* aInfos) :
mId(aId), mInfos(aInfos) {}
int operator()(const uint16_t aIndex) const {
if (JSID_BITS(mId) == JSID_BITS(mInfos[aIndex].id)) {
return 0;
}
return JSID_BITS(mId) < JSID_BITS(mInfos[aIndex].id) ? -1 : 1;
}
};
static const PropertyInfo*
XrayFindOwnPropertyInfo(JSContext* cx, JS::Handle<jsid> id,
const NativeProperties* nativeProperties)
{
if (MOZ_UNLIKELY(nativeProperties->iteratorAliasMethodIndex >= 0) &&
id == SYMBOL_TO_JSID(JS::GetWellKnownSymbol(cx, JS::SymbolCode::iterator))) {
return nativeProperties->MethodPropertyInfos() +
nativeProperties->iteratorAliasMethodIndex;
}
size_t idx;
const uint16_t* sortedPropertyIndices = nativeProperties->sortedPropertyIndices;
const PropertyInfo* propertyInfos = nativeProperties->PropertyInfos();
if (BinarySearchIf(sortedPropertyIndices, 0,
nativeProperties->propertyInfoCount,
IdToIndexComparator(id, propertyInfos), &idx)) {
return propertyInfos + sortedPropertyIndices[idx];
}
return nullptr;
}
static bool
XrayResolveAttribute(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj, JS::Handle<jsid> id,
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
const Prefable<const JSPropertySpec>& pref,
const JSPropertySpec& attrSpec,
JS::MutableHandle<JS::PropertyDescriptor> desc,
bool& cacheOnHolder)
{
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
if (!pref.isEnabled(cx, obj)) {
return true;
}
cacheOnHolder = true;
// Because of centralization, we need to make sure we fault in the JitInfos as
// well. At present, until the JSAPI changes, the easiest way to do this is
// wrap them up as functions ourselves.
desc.setAttributes(attrSpec.flags);
// They all have getters, so we can just make it.
JS::Rooted<JSObject*> funobj(cx,
XrayCreateFunction(cx, wrapper, attrSpec.accessors.getter.native, 0, id));
if (!funobj)
return false;
desc.setGetterObject(funobj);
desc.attributesRef() |= JSPROP_GETTER;
if (attrSpec.accessors.setter.native.op) {
// We have a setter! Make it.
funobj =
XrayCreateFunction(cx, wrapper, attrSpec.accessors.setter.native, 1, id);
if (!funobj)
return false;
desc.setSetterObject(funobj);
desc.attributesRef() |= JSPROP_SETTER;
} else {
desc.setSetter(nullptr);
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
desc.object().set(wrapper);
desc.value().setUndefined();
return true;
}
static bool
XrayResolveMethod(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj, JS::Handle<jsid> id,
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
const Prefable<const JSFunctionSpec>& pref,
const JSFunctionSpec& methodSpec,
JS::MutableHandle<JS::PropertyDescriptor> desc,
bool& cacheOnHolder)
{
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
if (!pref.isEnabled(cx, obj)) {
return true;
}
cacheOnHolder = true;
JSObject *funobj;
if (methodSpec.selfHostedName) {
JSFunction* fun =
JS::GetSelfHostedFunction(cx, methodSpec.selfHostedName, id,
methodSpec.nargs);
if (!fun) {
return false;
}
MOZ_ASSERT(!methodSpec.call.op, "Bad FunctionSpec declaration: non-null native");
MOZ_ASSERT(!methodSpec.call.info, "Bad FunctionSpec declaration: non-null jitinfo");
funobj = JS_GetFunctionObject(fun);
} else {
funobj = XrayCreateFunction(cx, wrapper, methodSpec.call,
methodSpec.nargs, id);
if (!funobj) {
return false;
}
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
desc.value().setObject(*funobj);
desc.setAttributes(methodSpec.flags);
desc.object().set(wrapper);
desc.setSetter(nullptr);
desc.setGetter(nullptr);
return true;
}
static bool
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
XrayResolveConstant(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj, JS::Handle<jsid>,
const Prefable<const ConstantSpec>& pref,
const ConstantSpec& constantSpec,
JS::MutableHandle<JS::PropertyDescriptor> desc,
bool& cacheOnHolder)
{
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
if (!pref.isEnabled(cx, obj)) {
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
cacheOnHolder = true;
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
desc.setAttributes(JSPROP_ENUMERATE | JSPROP_READONLY | JSPROP_PERMANENT);
desc.object().set(wrapper);
desc.value().set(constantSpec.value);
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
#define RESOLVE_CASE(PropType, SpecType, Resolver) \
case e##PropType: { \
MOZ_ASSERT(nativeProperties->Has##PropType##s()); \
const Prefable<const SpecType>& pref = \
nativeProperties->PropType##s()[propertyInfo.prefIndex]; \
return Resolver(cx, wrapper, obj, id, pref, \
pref.specs[propertyInfo.specIndex], desc, cacheOnHolder); \
}
static bool
XrayResolveProperty(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj, JS::Handle<jsid> id,
JS::MutableHandle<JS::PropertyDescriptor> desc,
bool& cacheOnHolder, DOMObjectType type,
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
const NativeProperties* nativeProperties,
const PropertyInfo& propertyInfo)
{
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
MOZ_ASSERT(type != eGlobalInterfacePrototype);
// Make sure we resolve for matched object type.
switch (propertyInfo.type) {
case eStaticMethod:
case eStaticAttribute:
if (type != eInterface) {
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
break;
case eMethod:
case eAttribute:
if (type != eGlobalInstance && type != eInterfacePrototype) {
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
break;
case eUnforgeableMethod:
case eUnforgeableAttribute:
if (!IsInstance(type)) {
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
break;
case eConstant:
if (IsInstance(type)) {
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
break;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
switch (propertyInfo.type) {
RESOLVE_CASE(StaticMethod, JSFunctionSpec, XrayResolveMethod)
RESOLVE_CASE(StaticAttribute, JSPropertySpec, XrayResolveAttribute)
RESOLVE_CASE(Method, JSFunctionSpec, XrayResolveMethod)
RESOLVE_CASE(Attribute, JSPropertySpec, XrayResolveAttribute)
RESOLVE_CASE(UnforgeableMethod, JSFunctionSpec, XrayResolveMethod)
RESOLVE_CASE(UnforgeableAttribute, JSPropertySpec, XrayResolveAttribute)
RESOLVE_CASE(Constant, ConstantSpec, XrayResolveConstant)
}
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
#undef RESOLVE_CASE
static bool
ResolvePrototypeOrConstructor(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj,
size_t protoAndIfaceCacheIndex, unsigned attrs,
JS::MutableHandle<JS::PropertyDescriptor> desc,
bool& cacheOnHolder)
{
JS::Rooted<JSObject*> global(cx, js::GetGlobalForObjectCrossCompartment(obj));
{
JSAutoCompartment ac(cx, global);
ProtoAndIfaceCache& protoAndIfaceCache = *GetProtoAndIfaceCache(global);
// This function is called when resolving the "constructor" and "prototype"
// properties of Xrays for DOM prototypes and constructors respectively.
// This means the relevant Xray exists, which means its _target_ exists.
// And that means we managed to successfullly create the prototype or
// constructor, respectively, and hence must have managed to create the
// thing it's pointing to as well. So our entry slot must exist.
JSObject* protoOrIface =
protoAndIfaceCache.EntrySlotMustExist(protoAndIfaceCacheIndex);
MOZ_RELEASE_ASSERT(protoOrIface, "How can this object not exist?");
cacheOnHolder = true;
desc.object().set(wrapper);
desc.setAttributes(attrs);
desc.setGetter(nullptr);
desc.setSetter(nullptr);
desc.value().set(JS::ObjectValue(*protoOrIface));
}
return JS_WrapPropertyDescriptor(cx, desc);
}
#ifdef DEBUG
static void
DEBUG_CheckXBLCallable(JSContext *cx, JSObject *obj)
{
// In general, we shouldn't have cross-compartment wrappers here, because
// we should be running in an XBL scope, and the content prototype should
// contain wrappers to functions defined in the XBL scope. But if the node
// has been adopted into another compartment, those prototypes will now point
// to a different XBL scope (which is ok).
MOZ_ASSERT_IF(js::IsCrossCompartmentWrapper(obj),
xpc::IsContentXBLScope(js::GetObjectCompartment(js::UncheckedUnwrap(obj))));
MOZ_ASSERT(JS::IsCallable(obj));
}
static void
DEBUG_CheckXBLLookup(JSContext *cx, JS::PropertyDescriptor *desc)
{
if (!desc->obj)
return;
if (!desc->value.isUndefined()) {
MOZ_ASSERT(desc->value.isObject());
DEBUG_CheckXBLCallable(cx, &desc->value.toObject());
}
if (desc->getter) {
MOZ_ASSERT(desc->attrs & JSPROP_GETTER);
DEBUG_CheckXBLCallable(cx, JS_FUNC_TO_DATA_PTR(JSObject *, desc->getter));
}
if (desc->setter) {
MOZ_ASSERT(desc->attrs & JSPROP_SETTER);
DEBUG_CheckXBLCallable(cx, JS_FUNC_TO_DATA_PTR(JSObject *, desc->setter));
}
}
#else
#define DEBUG_CheckXBLLookup(a, b) {}
#endif
/* static */ bool
XrayResolveOwnProperty(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj, JS::Handle<jsid> id,
JS::MutableHandle<JS::PropertyDescriptor> desc,
bool& cacheOnHolder)
{
cacheOnHolder = false;
DOMObjectType type;
const NativePropertyHooks *nativePropertyHooks =
GetNativePropertyHooks(cx, obj, type);
ResolveOwnProperty resolveOwnProperty =
nativePropertyHooks->mResolveOwnProperty;
if (type == eNamedPropertiesObject) {
// None of these should be cached on the holder, since they're dynamic.
return resolveOwnProperty(cx, wrapper, obj, id, desc);
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
const NativePropertiesHolder& nativePropertiesHolder =
nativePropertyHooks->mNativeProperties;
const NativeProperties* nativeProperties = nullptr;
const PropertyInfo* found = nullptr;
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
if ((nativeProperties = nativePropertiesHolder.regular)) {
found = XrayFindOwnPropertyInfo(cx, id, nativeProperties);
}
if (!found &&
(nativeProperties = nativePropertiesHolder.chromeOnly) &&
xpc::AccessCheck::isChrome(js::GetObjectCompartment(wrapper))) {
found = XrayFindOwnPropertyInfo(cx, id, nativeProperties);
}
if (IsInstance(type)) {
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
// Check for unforgeable properties first to prevent names provided by
// resolveOwnProperty callback from shadowing them.
if (found && (found->type == eUnforgeableMethod ||
found->type == eUnforgeableAttribute)) {
if (!XrayResolveProperty(cx, wrapper, obj, id, desc, cacheOnHolder, type,
nativeProperties, *found)) {
return false;
}
if (desc.object()) {
return true;
}
}
if (resolveOwnProperty) {
if (!resolveOwnProperty(cx, wrapper, obj, id, desc)) {
return false;
}
if (desc.object()) {
// None of these should be cached on the holder, since they're dynamic.
return true;
}
}
// If we're a special scope for in-content XBL, our script expects to see
// the bound XBL methods and attributes when accessing content. However,
// these members are implemented in content via custom-spliced prototypes,
// and thus aren't visible through Xray wrappers unless we handle them
// explicitly. So we check if we're running in such a scope, and if so,
// whether the wrappee is a bound element. If it is, we do a lookup via
// specialized XBL machinery.
//
// While we have to do some sketchy walking through content land, we should
// be protected by read-only/non-configurable properties, and any functions
// we end up with should _always_ be living in our own scope (the XBL scope).
// Make sure to assert that.
JS::Rooted<JSObject*> maybeElement(cx, obj);
Element* element;
if (xpc::ObjectScope(wrapper)->IsContentXBLScope() &&
NS_SUCCEEDED(UNWRAP_OBJECT(Element, &maybeElement, element))) {
if (!nsContentUtils::LookupBindingMember(cx, element, id, desc)) {
return false;
}
DEBUG_CheckXBLLookup(cx, desc.address());
if (desc.object()) {
// XBL properties shouldn't be cached on the holder, as they might be
// shadowed by own properties returned from mResolveOwnProperty.
desc.object().set(wrapper);
return true;
}
}
// For non-global instance Xrays there are no other properties, so return
// here for them.
if (type != eGlobalInstance) {
return true;
}
} else if (type == eInterface) {
if (id == GetJSIDByIndex(cx, XPCJSContext::IDX_PROTOTYPE)) {
return nativePropertyHooks->mPrototypeID == prototypes::id::_ID_Count ||
ResolvePrototypeOrConstructor(cx, wrapper, obj,
nativePropertyHooks->mPrototypeID,
JSPROP_PERMANENT | JSPROP_READONLY,
desc, cacheOnHolder);
}
if (id == SYMBOL_TO_JSID(JS::GetWellKnownSymbol(cx, JS::SymbolCode::hasInstance)) &&
DOMIfaceAndProtoJSClass::FromJSClass(js::GetObjectClass(obj))->
wantsInterfaceHasInstance) {
cacheOnHolder = true;
JSNativeWrapper interfaceHasInstanceWrapper = { InterfaceHasInstance,
nullptr };
JSObject* funObj = XrayCreateFunction(cx, wrapper,
interfaceHasInstanceWrapper, 1, id);
if (!funObj) {
return false;
}
desc.value().setObject(*funObj);
desc.setAttributes(JSPROP_READONLY | JSPROP_PERMANENT);
desc.object().set(wrapper);
desc.setSetter(nullptr);
desc.setGetter(nullptr);
return true;
}
} else {
MOZ_ASSERT(IsInterfacePrototype(type));
if (id == GetJSIDByIndex(cx, XPCJSContext::IDX_CONSTRUCTOR)) {
return nativePropertyHooks->mConstructorID == constructors::id::_ID_Count ||
ResolvePrototypeOrConstructor(cx, wrapper, obj,
nativePropertyHooks->mConstructorID,
0, desc, cacheOnHolder);
}
// The properties for globals live on the instance, so return here as there
// are no properties on their interface prototype object.
if (type == eGlobalInterfacePrototype) {
return true;
}
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
if (found &&
!XrayResolveProperty(cx, wrapper, obj, id, desc, cacheOnHolder, type,
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
nativeProperties, *found)) {
return false;
}
return true;
}
bool
XrayDefineProperty(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj, JS::Handle<jsid> id,
JS::Handle<JS::PropertyDescriptor> desc,
JS::ObjectOpResult &result, bool *defined)
{
if (!js::IsProxy(obj))
return true;
const DOMProxyHandler* handler = GetDOMProxyHandler(obj);
return handler->defineProperty(cx, wrapper, id, desc, result, defined);
}
template<typename SpecType>
bool
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
XrayAppendPropertyKeys(JSContext* cx, JS::Handle<JSObject*> obj,
const Prefable<const SpecType>* pref,
const PropertyInfo* infos, unsigned flags,
JS::AutoIdVector& props)
{
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
do {
bool prefIsEnabled = pref->isEnabled(cx, obj);
if (prefIsEnabled) {
const SpecType* spec = pref->specs;
do {
const jsid& id = infos++->id;
if (((flags & JSITER_HIDDEN) ||
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
(spec->flags & JSPROP_ENUMERATE)) &&
((flags & JSITER_SYMBOLS) || !JSID_IS_SYMBOL(id)) &&
!props.append(id)) {
return false;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
} while ((++spec)->name);
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
// Break if we have reached the end of pref.
if (!(++pref)->specs) {
break;
}
// Advance infos if the previous pref is disabled. The -1 is required
// because there is an end-of-list terminator between pref->specs and
// (pref - 1)->specs.
if (!prefIsEnabled) {
infos += pref->specs - (pref - 1)->specs - 1;
}
} while (1);
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
template<>
bool
XrayAppendPropertyKeys<ConstantSpec>(JSContext* cx, JS::Handle<JSObject*> obj,
const Prefable<const ConstantSpec>* pref,
const PropertyInfo* infos, unsigned flags,
JS::AutoIdVector& props)
{
do {
bool prefIsEnabled = pref->isEnabled(cx, obj);
if (prefIsEnabled) {
const ConstantSpec* spec = pref->specs;
do {
if (!props.append(infos++->id)) {
return false;
}
} while ((++spec)->name);
}
// Break if we have reached the end of pref.
if (!(++pref)->specs) {
break;
}
// Advance infos if the previous pref is disabled. The -1 is required
// because there is an end-of-list terminator between pref->specs and
// (pref - 1)->specs.
if (!prefIsEnabled) {
infos += pref->specs - (pref - 1)->specs - 1;
}
} while (1);
return true;
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
#define ADD_KEYS_IF_DEFINED(FieldName) { \
if (nativeProperties->Has##FieldName##s() && \
!XrayAppendPropertyKeys(cx, obj, \
nativeProperties->FieldName##s(), \
nativeProperties->FieldName##PropertyInfos(), \
flags, props)) { \
return false; \
} \
}
bool
XrayOwnPropertyKeys(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj,
unsigned flags, JS::AutoIdVector& props,
DOMObjectType type,
const NativeProperties* nativeProperties)
{
MOZ_ASSERT(type != eNamedPropertiesObject);
if (IsInstance(type)) {
ADD_KEYS_IF_DEFINED(UnforgeableMethod);
ADD_KEYS_IF_DEFINED(UnforgeableAttribute);
if (type == eGlobalInstance) {
ADD_KEYS_IF_DEFINED(Method);
ADD_KEYS_IF_DEFINED(Attribute);
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
} else {
MOZ_ASSERT(type != eGlobalInterfacePrototype);
if (type == eInterface) {
ADD_KEYS_IF_DEFINED(StaticMethod);
ADD_KEYS_IF_DEFINED(StaticAttribute);
} else {
MOZ_ASSERT(type == eInterfacePrototype);
ADD_KEYS_IF_DEFINED(Method);
ADD_KEYS_IF_DEFINED(Attribute);
}
Bug 1348099 part 1 - Binary search property id when resolve DOM Xrays own property. r=bz Currently we resolve a property by iterating every prefable and check whether it is enabled. If it is, we linear search the ids that it manages. This patch changes that to binary searching to find whether the id being resolved is present first, and checking whether its prefable is enabled only when we find it. This improves the performance of property resolution, especially when the property is not present. The patch stores all the property ids a NativePropertiesN owns in a single array of PropertyInfo structs. Each struct contains an id and the information needed to find the corresponding Prefable for the enabled check, as well as the information needed to find the correct property descriptor in the Prefable. We also store an array of indices into the PropertyInfo array, sorted by bits of the corresponding jsid. Given a jsid, this allows us to binary search for the index of the corresponding PropertyInfo, if any. The index array requires 2 bytes for each property, which is ~20k across all our bindings. The extra information stored in each PropertyInfo requires 4 bytes for each property, which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit builds due to alignment requirements on PropertyInfo. However we save a bit of memory from changing NativePropertiesN's trios to duos. The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only properties that are enabled. Without it, we will need to check every single property to know whether its prefable is enabled or not, which is inefficient. With this patch, initializing property ids takes longer because of the sorting. I measured also insertion sort because I thought the ids should be nearly sorted as they are generated sequentially at run time, but that's not the case and NS_QuickSort() runs faster. MozReview-Commit-ID: Lc4Z1ui3t0o --HG-- extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
2017-06-12 07:13:38 +03:00
ADD_KEYS_IF_DEFINED(Constant);
}
return true;
}
#undef ADD_KEYS_IF_DEFINED
bool
XrayOwnNativePropertyKeys(JSContext* cx, JS::Handle<JSObject*> wrapper,
const NativePropertyHooks* nativePropertyHooks,
DOMObjectType type, JS::Handle<JSObject*> obj,
unsigned flags, JS::AutoIdVector& props)
{
MOZ_ASSERT(type != eNamedPropertiesObject);
if (type == eInterface &&
nativePropertyHooks->mPrototypeID != prototypes::id::_ID_Count &&
!AddStringToIDVector(cx, props, "prototype")) {
return false;
}
if (IsInterfacePrototype(type) &&
nativePropertyHooks->mConstructorID != constructors::id::_ID_Count &&
(flags & JSITER_HIDDEN) &&
!AddStringToIDVector(cx, props, "constructor")) {
return false;
}
const NativePropertiesHolder& nativeProperties =
nativePropertyHooks->mNativeProperties;
if (nativeProperties.regular &&
!XrayOwnPropertyKeys(cx, wrapper, obj, flags, props, type,
nativeProperties.regular)) {
return false;
}
if (nativeProperties.chromeOnly &&
xpc::AccessCheck::isChrome(js::GetObjectCompartment(wrapper)) &&
!XrayOwnPropertyKeys(cx, wrapper, obj, flags, props, type,
nativeProperties.chromeOnly)) {
return false;
}
return true;
}
bool
XrayOwnPropertyKeys(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj,
unsigned flags, JS::AutoIdVector& props)
{
DOMObjectType type;
const NativePropertyHooks* nativePropertyHooks =
GetNativePropertyHooks(cx, obj, type);
EnumerateOwnProperties enumerateOwnProperties =
nativePropertyHooks->mEnumerateOwnProperties;
if (type == eNamedPropertiesObject) {
return enumerateOwnProperties(cx, wrapper, obj, props);
}
if (IsInstance(type)) {
// FIXME https://bugzilla.mozilla.org/show_bug.cgi?id=1071189
// Should do something about XBL properties too.
if (enumerateOwnProperties &&
!enumerateOwnProperties(cx, wrapper, obj, props)) {
return false;
}
}
return type == eGlobalInterfacePrototype ||
XrayOwnNativePropertyKeys(cx, wrapper, nativePropertyHooks, type,
obj, flags, props);
}
const JSClass*
XrayGetExpandoClass(JSContext* cx, JS::Handle<JSObject*> obj)
{
DOMObjectType type;
const NativePropertyHooks* nativePropertyHooks =
GetNativePropertyHooks(cx, obj, type);
if (!IsInstance(type)) {
// Non-instances don't need any special expando classes.
return &DefaultXrayExpandoObjectClass;
}
return nativePropertyHooks->mXrayExpandoClass;
}
bool
XrayDeleteNamedProperty(JSContext* cx, JS::Handle<JSObject*> wrapper,
JS::Handle<JSObject*> obj, JS::Handle<jsid> id,
JS::ObjectOpResult& opresult)
{
DOMObjectType type;
const NativePropertyHooks* nativePropertyHooks =
GetNativePropertyHooks(cx, obj, type);
if (!IsInstance(type) || !nativePropertyHooks->mDeleteNamedProperty) {
return opresult.succeed();
}
return nativePropertyHooks->mDeleteNamedProperty(cx, wrapper, obj, id,
opresult);
}
JSObject*
GetCachedSlotStorageObjectSlow(JSContext* cx, JS::Handle<JSObject*> obj,
bool* isXray)
{
if (!xpc::WrapperFactory::IsXrayWrapper(obj)) {
JSObject* retval = js::UncheckedUnwrap(obj, /* stopAtWindowProxy = */ false);
MOZ_ASSERT(IsDOMObject(retval));
*isXray = false;
return retval;
}
*isXray = true;
return xpc::EnsureXrayExpandoObject(cx, obj);;
}
DEFINE_XRAY_EXPANDO_CLASS(, DefaultXrayExpandoObjectClass, 0);
NativePropertyHooks sEmptyNativePropertyHooks = {
nullptr,
nullptr,
nullptr,
{
nullptr,
nullptr
},
prototypes::id::_ID_Count,
constructors::id::_ID_Count,
nullptr
};
const js::ClassOps sBoringInterfaceObjectClassClassOps = {
nullptr, /* addProperty */
nullptr, /* delProperty */
nullptr, /* getProperty */
nullptr, /* setProperty */
nullptr, /* enumerate */
nullptr, /* newEnumerate */
nullptr, /* resolve */
nullptr, /* mayResolve */
nullptr, /* finalize */
ThrowingConstructor, /* call */
nullptr, /* hasInstance */
ThrowingConstructor, /* construct */
nullptr, /* trace */
};
const js::ObjectOps sInterfaceObjectClassObjectOps = {
nullptr, /* lookupProperty */
nullptr, /* defineProperty */
nullptr, /* hasProperty */
nullptr, /* getProperty */
nullptr, /* setProperty */
nullptr, /* getOwnPropertyDescriptor */
nullptr, /* deleteProperty */
nullptr, /* watch */
nullptr, /* unwatch */
nullptr, /* getElements */
InterfaceObjectToString, /* funToString */
};
bool
GetPropertyOnPrototype(JSContext* cx, JS::Handle<JSObject*> proxy,
JS::Handle<JS::Value> receiver, JS::Handle<jsid> id,
bool* found, JS::MutableHandle<JS::Value> vp)
{
JS::Rooted<JSObject*> proto(cx);
if (!js::GetObjectProto(cx, proxy, &proto)) {
return false;
}
if (!proto) {
*found = false;
return true;
}
if (!JS_HasPropertyById(cx, proto, id, found)) {
return false;
}
if (!*found) {
return true;
}
return JS_ForwardGetPropertyTo(cx, proto, id, receiver, vp);
}
bool
HasPropertyOnPrototype(JSContext* cx, JS::Handle<JSObject*> proxy,
JS::Handle<jsid> id, bool* has)
{
JS::Rooted<JSObject*> proto(cx);
if (!js::GetObjectProto(cx, proxy, &proto)) {
return false;
}
if (!proto) {
*has = false;
return true;
}
return JS_HasPropertyById(cx, proto, id, has);
}
bool
AppendNamedPropertyIds(JSContext* cx, JS::Handle<JSObject*> proxy,
nsTArray<nsString>& names,
bool shadowPrototypeProperties,
JS::AutoIdVector& props)
{
for (uint32_t i = 0; i < names.Length(); ++i) {
JS::Rooted<JS::Value> v(cx);
if (!xpc::NonVoidStringToJsval(cx, names[i], &v)) {
return false;
}
JS::Rooted<jsid> id(cx);
if (!JS_ValueToId(cx, v, &id)) {
return false;
}
bool shouldAppend = shadowPrototypeProperties;
if (!shouldAppend) {
bool has;
if (!HasPropertyOnPrototype(cx, proxy, id, &has)) {
return false;
}
shouldAppend = !has;
}
if (shouldAppend) {
if (!props.append(id)) {
return false;
}
}
}
return true;
}
bool
DictionaryBase::ParseJSON(JSContext* aCx,
const nsAString& aJSON,
JS::MutableHandle<JS::Value> aVal)
{
if (aJSON.IsEmpty()) {
return true;
}
return JS_ParseJSON(aCx, PromiseFlatString(aJSON).get(), aJSON.Length(), aVal);
}
bool
DictionaryBase::StringifyToJSON(JSContext* aCx,
JS::Handle<JSObject*> aObj,
nsAString& aJSON) const
{
return JS::ToJSONMaybeSafely(aCx, aObj, AppendJSONToString, &aJSON);
}
/* static */
bool
DictionaryBase::AppendJSONToString(const char16_t* aJSONData,
uint32_t aDataLength,
void* aString)
{
nsAString* string = static_cast<nsAString*>(aString);
string->Append(aJSONData, aDataLength);
return true;
}
nsresult
ReparentWrapper(JSContext* aCx, JS::Handle<JSObject*> aObjArg)
{
js::AssertSameCompartment(aCx, aObjArg);
// Check if we're anywhere near the stack limit before we reach the
// transplanting code, since it has no good way to handle errors. This uses
// the untrusted script limit, which is not strictly necessary since no
// actual script should run.
if (!js::CheckRecursionLimitConservative(aCx)) {
return NS_ERROR_FAILURE;
}
JS::Rooted<JSObject*> aObj(aCx, aObjArg);
const DOMJSClass* domClass = GetDOMClass(aObj);
// DOM things are always parented to globals.
JS::Rooted<JSObject*> oldParent(aCx,
js::GetGlobalForObjectCrossCompartment(aObj));
MOZ_ASSERT(js::GetGlobalForObjectCrossCompartment(oldParent) == oldParent);
JS::Rooted<JSObject*> newParent(aCx,
domClass->mGetAssociatedGlobal(aCx, aObj));
MOZ_ASSERT(JS_IsGlobalObject(newParent));
JSAutoCompartment oldAc(aCx, oldParent);
JSCompartment* oldCompartment = js::GetObjectCompartment(oldParent);
JSCompartment* newCompartment = js::GetObjectCompartment(newParent);
if (oldCompartment == newCompartment) {
MOZ_ASSERT(oldParent == newParent);
return NS_OK;
}
nsISupports* native = UnwrapDOMObjectToISupports(aObj);
if (!native) {
return NS_OK;
}
bool isProxy = js::IsProxy(aObj);
JS::Rooted<JSObject*> expandoObject(aCx);
if (isProxy) {
expandoObject = DOMProxyHandler::GetAndClearExpandoObject(aObj);
}
JSAutoCompartment newAc(aCx, newParent);
// First we clone the reflector. We get a copy of its properties and clone its
// expando chain.
JS::Handle<JSObject*> proto = (domClass->mGetProto)(aCx);
if (!proto) {
return NS_ERROR_FAILURE;
}
JS::Rooted<JSObject*> newobj(aCx, JS_CloneObject(aCx, aObj, proto));
if (!newobj) {
return NS_ERROR_FAILURE;
}
JS::Rooted<JSObject*> propertyHolder(aCx);
JS::Rooted<JSObject*> copyFrom(aCx, isProxy ? expandoObject : aObj);
if (copyFrom) {
propertyHolder = JS_NewObjectWithGivenProto(aCx, nullptr, nullptr);
if (!propertyHolder) {
return NS_ERROR_OUT_OF_MEMORY;
}
if (!JS_CopyPropertiesFrom(aCx, propertyHolder, copyFrom)) {
return NS_ERROR_FAILURE;
}
} else {
propertyHolder = nullptr;
}
// Expandos from other compartments are attached to the target JS object.
// Copy them over, and let the old ones die a natural death.
// Note that at this point the DOM_OBJECT_SLOT for |newobj| has not been set.
// CloneExpandoChain() will use this property of |newobj| when it calls
// preserveWrapper() via attachExpandoObject() if |aObj| has expandos set, and
// preserveWrapper() will not do anything in this case. This is safe because
// if expandos are present then the wrapper will already have been preserved
// for this native.
if (!xpc::XrayUtils::CloneExpandoChain(aCx, newobj, aObj)) {
return NS_ERROR_FAILURE;
}
// We've set up |newobj|, so we make it own the native by setting its reserved
// slot and nulling out the reserved slot of |obj|.
//
// NB: It's important to do this _after_ copying the properties to
// propertyHolder. Otherwise, an object with |foo.x === foo| will
// crash when JS_CopyPropertiesFrom tries to call wrap() on foo.x.
js::SetReservedSlot(newobj, DOM_OBJECT_SLOT,
js::GetReservedSlot(aObj, DOM_OBJECT_SLOT));
js::SetReservedSlot(aObj, DOM_OBJECT_SLOT, JS::PrivateValue(nullptr));
2014-03-19 20:35:45 +04:00
aObj = xpc::TransplantObject(aCx, aObj, newobj);
if (!aObj) {
MOZ_CRASH();
}
2014-03-19 20:35:45 +04:00
nsWrapperCache* cache = nullptr;
CallQueryInterface(native, &cache);
bool preserving = cache->PreservingWrapper();
cache->SetPreservingWrapper(false);
cache->SetWrapper(aObj);
cache->SetPreservingWrapper(preserving);
if (propertyHolder) {
JS::Rooted<JSObject*> copyTo(aCx);
if (isProxy) {
copyTo = DOMProxyHandler::EnsureExpandoObject(aCx, aObj);
} else {
copyTo = aObj;
}
if (!copyTo || !JS_CopyPropertiesFrom(aCx, copyTo, propertyHolder)) {
MOZ_CRASH();
}
}
JS::Rooted<JSObject*> maybeObjLC(aCx, aObj);
nsObjectLoadingContent* htmlobject;
nsresult rv = UNWRAP_OBJECT(HTMLObjectElement, &maybeObjLC, htmlobject);
if (NS_FAILED(rv)) {
rv = UNWRAP_OBJECT(HTMLEmbedElement, &maybeObjLC, htmlobject);
if (NS_FAILED(rv)) {
htmlobject = nullptr;
}
}
if (htmlobject) {
htmlobject->SetupProtoChain(aCx, aObj);
}
// Now we can just return the wrapper
return NS_OK;
}
GlobalObject::GlobalObject(JSContext* aCx, JSObject* aObject)
: mGlobalJSObject(aCx),
mCx(aCx),
mGlobalObject(nullptr)
{
MOZ_ASSERT(mCx);
JS::Rooted<JSObject*> obj(aCx, aObject);
if (js::IsWrapper(obj)) {
obj = js::CheckedUnwrap(obj, /* stopAtWindowProxy = */ false);
if (!obj) {
// We should never end up here on a worker thread, since there shouldn't
// be any security wrappers to worry about.
if (!MOZ_LIKELY(NS_IsMainThread())) {
MOZ_CRASH();
}
Throw(aCx, NS_ERROR_XPC_SECURITY_MANAGER_VETO);
return;
}
}
mGlobalJSObject = js::GetGlobalForObjectCrossCompartment(obj);
}
nsISupports*
GlobalObject::GetAsSupports() const
{
if (mGlobalObject) {
return mGlobalObject;
}
MOZ_ASSERT(!js::IsWrapper(mGlobalJSObject));
// Most of our globals are DOM objects. Try that first. Note that this
// assumes that either the first nsISupports in the object is the canonical
// one or that we don't care about the canonical nsISupports here.
mGlobalObject = UnwrapDOMObjectToISupports(mGlobalJSObject);
if (mGlobalObject) {
return mGlobalObject;
}
MOZ_ASSERT(NS_IsMainThread(), "All our worker globals are DOM objects");
// Remove everything below here once all our global objects are using new
// bindings. If that ever happens; it would need to include Sandbox and
// BackstagePass.
// See whether mGlobalJSObject is an XPCWrappedNative. This will redo the
// IsWrapper bit above and the UnwrapDOMObjectToISupports in the case when
// we're not actually an XPCWrappedNative, but this should be a rare-ish case
// anyway.
Bug 1371259 part 9. Make UnwrapReflectorToISupports return already_AddRefed<nsISupports>. r=peterv The main reason to not do this would be performance (avoiding the addref/release), but there are two main mitigating factors: 1) All calls to UnwrapReflectorToISupports that pass in a Web IDL object already do the addref (and in fact QI). So this only affects the XPCWrappedNative case. 2) The vast majority of the callers proceed to QI on the pointer anyway, and a second addref is cheap; it's the first addref after a CC that can be expensive on a cycle-collected object. Going through the changes one by one: * In GlobalObject::GetAsSupports, we do have a change that slightly slows down precisely in the XPCWrappedNative global case. That's the message managers and the backstagepass. And this really only affects calls to Web IDL statics from those globals. * In UnwrapArgImpl we're talking about a Web IDL method taking an "external interface" type, and the UnwrapReflectorToISupports call is immediately followed by QI anyway. * In UnwrapXPConnectImpl we're talking about the case when we have a non-WebIDL-object implementation of a Web IDL interface. Again, this is the message manager globals, for EventTarget. And we have a QI call immediately after the UnwrapReflectorToISupports. * In the generated HasInstance hook for EventTarget we will be slightly slower when the LHS of the instanceof is an XPCWrappedNative. And not much slower, because again there's an immediate QI. * In InstallXBLField we're never going to have an XPCWrappedNative as thisObj; it's always an Element in practice. So this is no more expensive than before. * In sandbox's GetPrincipalOrSOP we now have an extra addref. But it was followed by various QIs anyway. * In XPCConvert::JSValToXPCException we have an extra addref if someone throws an XPCWrappedNative, which is fairly unlikely; our actual Exception objects are on Web IDL bindings. Plus we have an immediate QI. * In xpc::HasInstance we have an extra addred if the LHS of instanceof is an XPCWrappedNative. But, again, there's an immediated QI after the UnwrapReflectorToISupports. * In xpcJSWeakReference::Init we are likely doing an extra addref, but again immediately followed by QI. I think it's worth making this change just to remove the footgun and that the perf impact, if any, is pretty minimal.
2017-07-10 23:05:26 +03:00
nsCOMPtr<nsISupports> supp = xpc::UnwrapReflectorToISupports(mGlobalJSObject);
if (supp) {
// See documentation for mGlobalJSObject for why this assignment is OK.
mGlobalObject = supp;
return mGlobalObject;
}
// And now a final hack. Sandbox is not a reflector, but it does have an
// nsIGlobalObject hanging out in its private slot. Handle that case here,
// (though again, this will do the useless UnwrapDOMObjectToISupports if we
// got here for something that is somehow not a DOM object, not an
// XPCWrappedNative _and_ not a Sandbox).
if (XPCConvert::GetISupportsFromJSObject(mGlobalJSObject, &mGlobalObject)) {
return mGlobalObject;
}
MOZ_ASSERT(!mGlobalObject);
Throw(mCx, NS_ERROR_XPC_BAD_CONVERT_JS);
return nullptr;
}
nsIPrincipal*
GlobalObject::GetSubjectPrincipal() const
{
if (!NS_IsMainThread()) {
return nullptr;
}
JSCompartment* compartment = js::GetContextCompartment(mCx);
MOZ_ASSERT(compartment);
JSPrincipals* principals = JS_GetCompartmentPrincipals(compartment);
return nsJSPrincipals::get(principals);
}
CallerType
GlobalObject::CallerType() const
{
return nsContentUtils::ThreadsafeIsSystemCaller(mCx) ?
dom::CallerType::System : dom::CallerType::NonSystem;
}
static bool
CallOrdinaryHasInstance(JSContext* cx, JS::CallArgs& args)
{
JS::Rooted<JSObject*> thisObj(cx, &args.thisv().toObject());
bool isInstance;
if (!JS::OrdinaryHasInstance(cx, thisObj, args.get(0), &isInstance)) {
return false;
}
args.rval().setBoolean(isInstance);
return true;
}
bool
InterfaceHasInstance(JSContext* cx, unsigned argc, JS::Value* vp)
{
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
// If the thing we were passed is not an object, return false like
// OrdinaryHasInstance does.
if (!args.get(0).isObject()) {
args.rval().setBoolean(false);
return true;
}
// If "this" is not an object, likewise return false (again, like
// OrdinaryHasInstance).
if (!args.thisv().isObject()) {
args.rval().setBoolean(false);
return true;
}
// If "this" doesn't have a DOMIfaceAndProtoJSClass, it's not a DOM
// constructor, so just fall back to OrdinaryHasInstance. But note that we
// should CheckedUnwrap here, because otherwise we won't get the right
// answers.
JS::Rooted<JSObject*> thisObj(cx, js::CheckedUnwrap(&args.thisv().toObject()));
if (!thisObj) {
// Just fall back on the normal thing, in case it still happens to work.
return CallOrdinaryHasInstance(cx, args);
}
const js::Class* thisClass = js::GetObjectClass(thisObj);
if (!IsDOMIfaceAndProtoClass(thisClass)) {
return CallOrdinaryHasInstance(cx, args);
}
const DOMIfaceAndProtoJSClass* clasp =
DOMIfaceAndProtoJSClass::FromJSClass(thisClass);
// If "this" isn't a DOM constructor or is a constructor for an interface
// without a prototype, just fall back to OrdinaryHasInstance.
if (clasp->mType != eInterface ||
clasp->mPrototypeID == prototypes::id::_ID_Count) {
return CallOrdinaryHasInstance(cx, args);
}
JS::Rooted<JSObject*> instance(cx, &args[0].toObject());
const DOMJSClass* domClass =
GetDOMClass(js::UncheckedUnwrap(instance, /* stopAtWindowProxy = */ false));
if (domClass &&
domClass->mInterfaceChain[clasp->mDepth] == clasp->mPrototypeID) {
args.rval().setBoolean(true);
return true;
}
if (jsipc::IsWrappedCPOW(instance)) {
bool boolp = false;
if (!jsipc::DOMInstanceOf(cx, js::UncheckedUnwrap(instance), clasp->mPrototypeID,
clasp->mDepth, &boolp)) {
return false;
}
args.rval().setBoolean(boolp);
return true;
}
return CallOrdinaryHasInstance(cx, args);
}
bool
InterfaceHasInstance(JSContext* cx, int prototypeID, int depth,
JS::Handle<JSObject*> instance,
bool* bp)
{
const DOMJSClass* domClass = GetDOMClass(js::UncheckedUnwrap(instance));
MOZ_ASSERT(!domClass || prototypeID != prototypes::id::_ID_Count,
"Why do we have a hasInstance hook if we don't have a prototype "
"ID?");
*bp = (domClass && domClass->mInterfaceChain[depth] == prototypeID);
return true;
}
bool
ReportLenientThisUnwrappingFailure(JSContext* cx, JSObject* obj)
{
JS::Rooted<JSObject*> rootedObj(cx, obj);
GlobalObject global(cx, rootedObj);
if (global.Failed()) {
return false;
}
nsCOMPtr<nsPIDOMWindowInner> window = do_QueryInterface(global.GetAsSupports());
if (window && window->GetDoc()) {
window->GetDoc()->WarnOnceAbout(nsIDocument::eLenientThis);
}
return true;
}
bool
GetContentGlobalForJSImplementedObject(JSContext* cx, JS::Handle<JSObject*> obj,
nsIGlobalObject** globalObj)
{
// Be very careful to not get tricked here.
MOZ_ASSERT(NS_IsMainThread());
if (!xpc::AccessCheck::isChrome(js::GetObjectCompartment(obj))) {
MOZ_CRASH("Should have a chrome object here");
}
// Look up the content-side object.
JS::Rooted<JS::Value> domImplVal(cx);
if (!JS_GetProperty(cx, obj, "__DOM_IMPL__", &domImplVal)) {
return false;
}
if (!domImplVal.isObject()) {
ThrowErrorMessage(cx, MSG_NOT_OBJECT, "Value");
return false;
}
// Go ahead and get the global from it. GlobalObject will handle
// doing unwrapping as needed.
GlobalObject global(cx, &domImplVal.toObject());
if (global.Failed()) {
return false;
}
DebugOnly<nsresult> rv = CallQueryInterface(global.GetAsSupports(), globalObj);
MOZ_ASSERT(NS_SUCCEEDED(rv));
MOZ_ASSERT(*globalObj);
return true;
}
already_AddRefed<nsIGlobalObject>
ConstructJSImplementation(const char* aContractId,
const GlobalObject& aGlobal,
JS::MutableHandle<JSObject*> aObject,
ErrorResult& aRv)
{
// Get the global object to use as a parent and for initialization.
nsCOMPtr<nsIGlobalObject> global = do_QueryInterface(aGlobal.GetAsSupports());
if (!global) {
aRv.Throw(NS_ERROR_FAILURE);
return nullptr;
}
ConstructJSImplementation(aContractId, global, aObject, aRv);
if (aRv.Failed()) {
return nullptr;
}
return global.forget();
}
void
ConstructJSImplementation(const char* aContractId,
nsIGlobalObject* aGlobal,
JS::MutableHandle<JSObject*> aObject,
ErrorResult& aRv)
{
MOZ_ASSERT(NS_IsMainThread());
// Make sure to divorce ourselves from the calling JS while creating and
// initializing the object, so exceptions from that will get reported
// properly, since those are never exceptions that a spec wants to be thrown.
{
AutoNoJSAPI nojsapi;
// Get the XPCOM component containing the JS implementation.
nsresult rv;
nsCOMPtr<nsISupports> implISupports = do_CreateInstance(aContractId, &rv);
if (!implISupports) {
nsPrintfCString msg("Failed to get JS implementation for contract \"%s\"",
aContractId);
NS_WARNING(msg.get());
aRv.Throw(rv);
return;
}
// Initialize the object, if it implements nsIDOMGlobalPropertyInitializer
// and our global is a window.
nsCOMPtr<nsIDOMGlobalPropertyInitializer> gpi =
do_QueryInterface(implISupports);
nsCOMPtr<nsPIDOMWindowInner> window = do_QueryInterface(aGlobal);
if (gpi) {
JS::Rooted<JS::Value> initReturn(RootingCx());
rv = gpi->Init(window, &initReturn);
if (NS_FAILED(rv)) {
aRv.Throw(rv);
return;
}
// With JS-implemented WebIDL, the return value of init() is not used to determine
// if init() failed, so init() should only return undefined. Any kind of permission
// or pref checking must happen by adding an attribute to the WebIDL interface.
if (!initReturn.isUndefined()) {
MOZ_ASSERT(false, "The init() method for JS-implemented WebIDL should not return anything");
MOZ_CRASH();
}
}
// Extract the JS implementation from the XPCOM object.
nsCOMPtr<nsIXPConnectWrappedJS> implWrapped =
do_QueryInterface(implISupports, &rv);
MOZ_ASSERT(implWrapped, "Failed to get wrapped JS from XPCOM component.");
if (!implWrapped) {
aRv.Throw(rv);
return;
}
aObject.set(implWrapped->GetJSObject());
if (!aObject) {
aRv.Throw(NS_ERROR_FAILURE);
}
}
}
bool
NonVoidByteStringToJsval(JSContext *cx, const nsACString &str,
JS::MutableHandle<JS::Value> rval)
{
// ByteStrings are not UTF-8 encoded.
JSString* jsStr = JS_NewStringCopyN(cx, str.Data(), str.Length());
if (!jsStr)
return false;
rval.setString(jsStr);
return true;
}
template<typename T> static void
NormalizeUSVStringInternal(T& aString)
{
char16_t* start = aString.BeginWriting();
// Must use const here because we can't pass char** to UTF16CharEnumerator as
// it expects const char**. Unclear why this is illegal...
const char16_t* nextChar = start;
const char16_t* end = aString.Data() + aString.Length();
while (nextChar < end) {
uint32_t enumerated = UTF16CharEnumerator::NextChar(&nextChar, end);
if (enumerated == UCS2_REPLACEMENT_CHAR) {
int32_t lastCharIndex = (nextChar - start) - 1;
start[lastCharIndex] = static_cast<char16_t>(enumerated);
}
}
}
void
NormalizeUSVString(nsAString& aString)
{
NormalizeUSVStringInternal(aString);
}
void
NormalizeUSVString(binding_detail::FakeString& aString)
{
NormalizeUSVStringInternal(aString);
}
bool
ConvertJSValueToByteString(JSContext* cx, JS::Handle<JS::Value> v,
bool nullable, nsACString& result)
{
JS::Rooted<JSString*> s(cx);
if (v.isString()) {
s = v.toString();
} else {
if (nullable && v.isNullOrUndefined()) {
result.SetIsVoid(true);
return true;
}
s = JS::ToString(cx, v);
if (!s) {
return false;
}
}
// Conversion from Javascript string to ByteString is only valid if all
// characters < 256. This is always the case for Latin1 strings.
size_t length;
if (!js::StringHasLatin1Chars(s)) {
// ThrowErrorMessage can GC, so we first scan the string for bad chars
// and report the error outside the AutoCheckCannotGC scope.
bool foundBadChar = false;
size_t badCharIndex;
char16_t badChar;
{
JS::AutoCheckCannotGC nogc;
const char16_t* chars = JS_GetTwoByteStringCharsAndLength(cx, nogc, s, &length);
if (!chars) {
return false;
}
for (size_t i = 0; i < length; i++) {
if (chars[i] > 255) {
badCharIndex = i;
badChar = chars[i];
foundBadChar = true;
break;
}
}
}
if (foundBadChar) {
MOZ_ASSERT(badCharIndex < length);
MOZ_ASSERT(badChar > 255);
// The largest unsigned 64 bit number (18,446,744,073,709,551,615) has
// 20 digits, plus one more for the null terminator.
char index[21];
Bug 895322 - Part 1: Replace the usages of MOZ_STATIC_ASSERT with C++11 static_assert; r=Waldo This patch was mostly generated by running the following scripts on the codebase, with some manual changes made afterwards: # static_assert.sh #!/bin/bash # Command to convert an NSPR integer type to the equivalent standard integer type function convert() { echo "Converting $1 to $2..." find . ! -wholename "*nsprpub*" \ ! -wholename "*security/nss*" \ ! -wholename "*/.hg*" \ ! -wholename "obj-ff-dbg*" \ ! -name nsXPCOMCID.h \ ! -name prtypes.h \ -type f \ \( -iname "*.cpp" \ -o -iname "*.h" \ -o -iname "*.cc" \ -o -iname "*.mm" \) | \ xargs -n 1 `dirname $0`/assert_replacer.py #sed -i -e "s/\b$1\b/$2/g" } convert MOZ_STATIC_ASSERT static_assert hg rev --no-backup mfbt/Assertions.h \ media/webrtc/signaling/src/sipcc/core/includes/ccapi.h \ modules/libmar/src/mar_private.h \ modules/libmar/src/mar.h # assert_replacer.py #!/usr/bin/python import sys import re pattern = re.compile(r"\bMOZ_STATIC_ASSERT\b") def replaceInPlace(fname): print fname f = open(fname, "rw+") lines = f.readlines() for i in range(0, len(lines)): while True: index = re.search(pattern, lines[i]) if index != None: index = index.start() lines[i] = lines[i][0:index] + "static_assert" + lines[i][index+len("MOZ_STATIC_ASSERT"):] for j in range(i + 1, len(lines)): if lines[j].find(" ", index) == index: lines[j] = lines[j][0:index] + lines[j][index+4:] else: break else: break f.seek(0, 0) f.truncate() f.write("".join(lines)) f.close() argc = len(sys.argv) for i in range(1, argc): replaceInPlace(sys.argv[i]) --HG-- extra : rebase_source : 4b4a4047d82f2c205b9fad8d56dfc3f1afc0b045
2013-07-18 21:59:53 +04:00
static_assert(sizeof(size_t) <= 8, "index array too small");
SprintfLiteral(index, "%zu", badCharIndex);
// A char16_t is 16 bits long. The biggest unsigned 16 bit
// number (65,535) has 5 digits, plus one more for the null
// terminator.
char badCharArray[6];
static_assert(sizeof(char16_t) <= 2, "badCharArray too small");
SprintfLiteral(badCharArray, "%d", badChar);
ThrowErrorMessage(cx, MSG_INVALID_BYTESTRING, index, badCharArray);
return false;
}
} else {
length = js::GetStringLength(s);
}
static_assert(js::MaxStringLength < UINT32_MAX,
"length+1 shouldn't overflow");
if (!result.SetLength(length, fallible)) {
return false;
}
JS_EncodeStringToBuffer(cx, s, result.BeginWriting(), length);
return true;
}
void
FinalizeGlobal(JSFreeOp* aFreeOp, JSObject* aObj)
{
MOZ_ASSERT(js::GetObjectClass(aObj)->flags & JSCLASS_DOM_GLOBAL);
mozilla::dom::DestroyProtoAndIfaceCache(aObj);
}
bool
ResolveGlobal(JSContext* aCx, JS::Handle<JSObject*> aObj,
JS::Handle<jsid> aId, bool* aResolvedp)
{
MOZ_ASSERT(JS_IsGlobalObject(aObj),
"Should have a global here, since we plan to resolve standard "
"classes!");
return JS_ResolveStandardClass(aCx, aObj, aId, aResolvedp);
}
bool
MayResolveGlobal(const JSAtomState& aNames, jsid aId, JSObject* aMaybeObj)
{
return JS_MayResolveStandardClass(aNames, aId, aMaybeObj);
}
bool
EnumerateGlobal(JSContext* aCx, JS::HandleObject aObj,
JS::AutoIdVector& aProperties, bool aEnumerableOnly)
{
MOZ_ASSERT(JS_IsGlobalObject(aObj),
"Should have a global here, since we plan to enumerate standard "
"classes!");
return JS_NewEnumerateStandardClasses(aCx, aObj, aProperties,
aEnumerableOnly);
}
bool
IsNonExposedGlobal(JSContext* aCx, JSObject* aGlobal,
uint32_t aNonExposedGlobals)
{
MOZ_ASSERT(aNonExposedGlobals, "Why did we get called?");
MOZ_ASSERT((aNonExposedGlobals &
~(GlobalNames::Window |
GlobalNames::BackstagePass |
GlobalNames::DedicatedWorkerGlobalScope |
GlobalNames::SharedWorkerGlobalScope |
GlobalNames::ServiceWorkerGlobalScope |
GlobalNames::WorkerDebuggerGlobalScope |
GlobalNames::WorkletGlobalScope)) == 0,
"Unknown non-exposed global type");
const char* name = js::GetObjectClass(aGlobal)->name;
if ((aNonExposedGlobals & GlobalNames::Window) &&
!strcmp(name, "Window")) {
return true;
}
if ((aNonExposedGlobals & GlobalNames::BackstagePass) &&
!strcmp(name, "BackstagePass")) {
return true;
}
if ((aNonExposedGlobals & GlobalNames::DedicatedWorkerGlobalScope) &&
!strcmp(name, "DedicatedWorkerGlobalScope")) {
return true;
}
if ((aNonExposedGlobals & GlobalNames::SharedWorkerGlobalScope) &&
!strcmp(name, "SharedWorkerGlobalScope")) {
return true;
}
if ((aNonExposedGlobals & GlobalNames::ServiceWorkerGlobalScope) &&
!strcmp(name, "ServiceWorkerGlobalScope")) {
return true;
}
if ((aNonExposedGlobals & GlobalNames::WorkerDebuggerGlobalScope) &&
!strcmp(name, "WorkerDebuggerGlobalScopex")) {
return true;
}
if ((aNonExposedGlobals & GlobalNames::WorkletGlobalScope) &&
!strcmp(name, "WorkletGlobalScope")) {
return true;
}
return false;
}
void
HandlePrerenderingViolation(nsPIDOMWindowInner* aWindow)
{
// Freeze the window and its workers, and its children too.
aWindow->Freeze();
// Suspend event handling on the document
nsCOMPtr<nsIDocument> doc = aWindow->GetExtantDoc();
if (doc) {
doc->SuppressEventHandling();
}
}
bool
EnforceNotInPrerendering(JSContext* aCx, JSObject* aObj)
{
JS::Rooted<JSObject*> thisObj(aCx, js::CheckedUnwrap(aObj));
if (!thisObj) {
// Without a this object, we cannot check the safety.
return true;
}
nsGlobalWindow* window = xpc::WindowGlobalOrNull(thisObj);
if (!window) {
// Without a window, we cannot check the safety.
return true;
}
if (window->GetIsPrerendered()) {
HandlePrerenderingViolation(window->AsInner());
// When the bindings layer sees a false return value, it returns false form
// the JSNative in order to trigger an uncatchable exception.
return false;
}
return true;
}
bool
GenericBindingGetter(JSContext* cx, unsigned argc, JS::Value* vp)
{
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
const JSJitInfo *info = FUNCTION_VALUE_TO_JITINFO(args.calleev());
prototypes::ID protoID = static_cast<prototypes::ID>(info->protoID);
if (!args.thisv().isObject()) {
return ThrowInvalidThis(cx, args, false, protoID);
}
JS::Rooted<JSObject*> obj(cx, &args.thisv().toObject());
// NOTE: we want to leave obj in its initial compartment, so don't want to
// pass it to UnwrapObject.
JS::Rooted<JSObject*> rootSelf(cx, obj);
void* self;
{
binding_detail::MutableObjectHandleWrapper wrapper(&rootSelf);
nsresult rv = binding_detail::UnwrapObjectInternal<void, true>(wrapper,
self,
protoID,
info->depth);
if (NS_FAILED(rv)) {
return ThrowInvalidThis(cx, args,
rv == NS_ERROR_XPC_SECURITY_MANAGER_VETO,
protoID);
}
}
MOZ_ASSERT(info->type() == JSJitInfo::Getter);
JSJitGetterOp getter = info->getter;
bool ok = getter(cx, obj, self, JSJitGetterCallArgs(args));
#ifdef DEBUG
if (ok) {
AssertReturnTypeMatchesJitinfo(info, args.rval());
}
#endif
return ok;
}
bool
GenericPromiseReturningBindingGetter(JSContext* cx, unsigned argc, JS::Value* vp)
{
// Make sure to save the callee before someone maybe messes with rval().
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
JS::Rooted<JSObject*> callee(cx, &args.callee());
// We could invoke GenericBindingGetter here, but that involves an
// extra call. Manually inline it instead.
const JSJitInfo *info = FUNCTION_VALUE_TO_JITINFO(args.calleev());
prototypes::ID protoID = static_cast<prototypes::ID>(info->protoID);
if (!args.thisv().isObject()) {
ThrowInvalidThis(cx, args, false, protoID);
return ConvertExceptionToPromise(cx, xpc::XrayAwareCalleeGlobal(callee),
args.rval());
}
JS::Rooted<JSObject*> obj(cx, &args.thisv().toObject());
// NOTE: we want to leave obj in its initial compartment, so don't want to
// pass it to UnwrapObject.
JS::Rooted<JSObject*> rootSelf(cx, obj);
void* self;
{
binding_detail::MutableObjectHandleWrapper wrapper(&rootSelf);
nsresult rv = binding_detail::UnwrapObjectInternal<void, true>(wrapper,
self,
protoID,
info->depth);
if (NS_FAILED(rv)) {
ThrowInvalidThis(cx, args, rv == NS_ERROR_XPC_SECURITY_MANAGER_VETO,
protoID);
return ConvertExceptionToPromise(cx, xpc::XrayAwareCalleeGlobal(callee),
args.rval());
}
}
MOZ_ASSERT(info->type() == JSJitInfo::Getter);
JSJitGetterOp getter = info->getter;
bool ok = getter(cx, obj, self, JSJitGetterCallArgs(args));
if (ok) {
#ifdef DEBUG
AssertReturnTypeMatchesJitinfo(info, args.rval());
#endif
return true;
}
// Promise-returning getters always return objects
MOZ_ASSERT(info->returnType() == JSVAL_TYPE_OBJECT);
return ConvertExceptionToPromise(cx, xpc::XrayAwareCalleeGlobal(callee),
args.rval());
}
bool
GenericBindingSetter(JSContext* cx, unsigned argc, JS::Value* vp)
{
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
const JSJitInfo *info = FUNCTION_VALUE_TO_JITINFO(args.calleev());
prototypes::ID protoID = static_cast<prototypes::ID>(info->protoID);
if (!args.thisv().isObject()) {
return ThrowInvalidThis(cx, args, false, protoID);
}
JS::Rooted<JSObject*> obj(cx, &args.thisv().toObject());
// NOTE: we want to leave obj in its initial compartment, so don't want to
// pass it to UnwrapObject.
JS::Rooted<JSObject*> rootSelf(cx, obj);
void* self;
{
binding_detail::MutableObjectHandleWrapper wrapper(&rootSelf);
nsresult rv = binding_detail::UnwrapObjectInternal<void, true>(wrapper,
self,
protoID,
info->depth);
if (NS_FAILED(rv)) {
return ThrowInvalidThis(cx, args,
rv == NS_ERROR_XPC_SECURITY_MANAGER_VETO,
protoID);
}
}
if (args.length() == 0) {
return ThrowNoSetterArg(cx, protoID);
}
MOZ_ASSERT(info->type() == JSJitInfo::Setter);
JSJitSetterOp setter = info->setter;
if (!setter(cx, obj, self, JSJitSetterCallArgs(args))) {
return false;
}
args.rval().setUndefined();
#ifdef DEBUG
AssertReturnTypeMatchesJitinfo(info, args.rval());
#endif
return true;
}
bool
GenericBindingMethod(JSContext* cx, unsigned argc, JS::Value* vp)
{
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
const JSJitInfo *info = FUNCTION_VALUE_TO_JITINFO(args.calleev());
prototypes::ID protoID = static_cast<prototypes::ID>(info->protoID);
if (!args.thisv().isObject()) {
return ThrowInvalidThis(cx, args, false, protoID);
}
JS::Rooted<JSObject*> obj(cx, &args.thisv().toObject());
// NOTE: we want to leave obj in its initial compartment, so don't want to
// pass it to UnwrapObject.
JS::Rooted<JSObject*> rootSelf(cx, obj);
void* self;
{
binding_detail::MutableObjectHandleWrapper wrapper(&rootSelf);
nsresult rv = binding_detail::UnwrapObjectInternal<void, true>(wrapper,
self,
protoID,
info->depth);
if (NS_FAILED(rv)) {
return ThrowInvalidThis(cx, args,
rv == NS_ERROR_XPC_SECURITY_MANAGER_VETO,
protoID);
}
}
MOZ_ASSERT(info->type() == JSJitInfo::Method);
JSJitMethodOp method = info->method;
bool ok = method(cx, obj, self, JSJitMethodCallArgs(args));
#ifdef DEBUG
if (ok) {
AssertReturnTypeMatchesJitinfo(info, args.rval());
}
#endif
return ok;
}
bool
GenericPromiseReturningBindingMethod(JSContext* cx, unsigned argc, JS::Value* vp)
{
// Make sure to save the callee before someone maybe messes with rval().
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
JS::Rooted<JSObject*> callee(cx, &args.callee());
// We could invoke GenericBindingMethod here, but that involves an
// extra call. Manually inline it instead.
const JSJitInfo *info = FUNCTION_VALUE_TO_JITINFO(args.calleev());
prototypes::ID protoID = static_cast<prototypes::ID>(info->protoID);
if (!args.thisv().isObject()) {
ThrowInvalidThis(cx, args, false, protoID);
return ConvertExceptionToPromise(cx, xpc::XrayAwareCalleeGlobal(callee),
args.rval());
}
JS::Rooted<JSObject*> obj(cx, &args.thisv().toObject());
// NOTE: we want to leave obj in its initial compartment, so don't want to
// pass it to UnwrapObject.
JS::Rooted<JSObject*> rootSelf(cx, obj);
void* self;
{
binding_detail::MutableObjectHandleWrapper wrapper(&rootSelf);
nsresult rv = binding_detail::UnwrapObjectInternal<void, true>(wrapper,
self,
protoID,
info->depth);
if (NS_FAILED(rv)) {
ThrowInvalidThis(cx, args, rv == NS_ERROR_XPC_SECURITY_MANAGER_VETO,
protoID);
return ConvertExceptionToPromise(cx, xpc::XrayAwareCalleeGlobal(callee),
args.rval());
}
}
MOZ_ASSERT(info->type() == JSJitInfo::Method);
JSJitMethodOp method = info->method;
bool ok = method(cx, obj, self, JSJitMethodCallArgs(args));
if (ok) {
#ifdef DEBUG
AssertReturnTypeMatchesJitinfo(info, args.rval());
#endif
return true;
}
// Promise-returning methods always return objects
MOZ_ASSERT(info->returnType() == JSVAL_TYPE_OBJECT);
return ConvertExceptionToPromise(cx, xpc::XrayAwareCalleeGlobal(callee),
args.rval());
}
bool
StaticMethodPromiseWrapper(JSContext* cx, unsigned argc, JS::Value* vp)
{
// Make sure to save the callee before someone maybe messes with rval().
JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
JS::Rooted<JSObject*> callee(cx, &args.callee());
const JSJitInfo *info = FUNCTION_VALUE_TO_JITINFO(args.calleev());
MOZ_ASSERT(info);
MOZ_ASSERT(info->type() == JSJitInfo::StaticMethod);
bool ok = info->staticMethod(cx, argc, vp);
if (ok) {
return true;
}
return ConvertExceptionToPromise(cx, xpc::XrayAwareCalleeGlobal(callee),
args.rval());
}
bool
ConvertExceptionToPromise(JSContext* cx,
JSObject* promiseScope,
JS::MutableHandle<JS::Value> rval)
{
{
JSAutoCompartment ac(cx, promiseScope);
JS::Rooted<JS::Value> exn(cx);
if (!JS_GetPendingException(cx, &exn)) {
// This is very important: if there is no pending exception here but we're
// ending up in this code, that means the callee threw an uncatchable
// exception. Just propagate that out as-is.
return false;
}
JS_ClearPendingException(cx);
JSObject* promise = JS::CallOriginalPromiseReject(cx, exn);
if (!promise) {
// We just give up. Put the exception back.
JS_SetPendingException(cx, exn);
return false;
}
rval.setObject(*promise);
}
// Now make sure we rewrap promise back into the compartment we want
return JS_WrapValue(cx, rval);
}
/* static */
void
CreateGlobalOptions<nsGlobalWindow>::TraceGlobal(JSTracer* aTrc, JSObject* aObj)
{
xpc::TraceXPCGlobal(aTrc, aObj);
}
static bool sRegisteredDOMNames = false;
nsresult
RegisterDOMNames()
{
if (sRegisteredDOMNames) {
return NS_OK;
}
// Register new DOM bindings
WebIDLGlobalNameHash::Init();
nsresult rv = nsDOMClassInfo::Init();
if (NS_FAILED(rv)) {
NS_ERROR("Could not initialize nsDOMClassInfo");
return rv;
}
sRegisteredDOMNames = true;
return NS_OK;
}
/* static */
bool
CreateGlobalOptions<nsGlobalWindow>::PostCreateGlobal(JSContext* aCx,
JS::Handle<JSObject*> aGlobal)
{
nsresult rv = RegisterDOMNames();
if (NS_FAILED(rv)) {
return Throw(aCx, rv);
}
// Invoking the XPCWrappedNativeScope constructor automatically hooks it
// up to the compartment of aGlobal.
(void) new XPCWrappedNativeScope(aCx, aGlobal);
return true;
}
#ifdef DEBUG
void
AssertReturnTypeMatchesJitinfo(const JSJitInfo* aJitInfo,
JS::Handle<JS::Value> aValue)
{
switch (aJitInfo->returnType()) {
case JSVAL_TYPE_UNKNOWN:
// Any value is good.
break;
case JSVAL_TYPE_DOUBLE:
// The value could actually be an int32 value as well.
MOZ_ASSERT(aValue.isNumber());
break;
case JSVAL_TYPE_INT32:
MOZ_ASSERT(aValue.isInt32());
break;
case JSVAL_TYPE_UNDEFINED:
MOZ_ASSERT(aValue.isUndefined());
break;
case JSVAL_TYPE_BOOLEAN:
MOZ_ASSERT(aValue.isBoolean());
break;
case JSVAL_TYPE_STRING:
MOZ_ASSERT(aValue.isString());
break;
case JSVAL_TYPE_NULL:
MOZ_ASSERT(aValue.isNull());
break;
case JSVAL_TYPE_OBJECT:
MOZ_ASSERT(aValue.isObject());
break;
default:
// Someone messed up their jitinfo type.
MOZ_ASSERT(false, "Unexpected JSValueType stored in jitinfo");
break;
}
}
#endif
bool
CallerSubsumes(JSObject *aObject)
{
nsIPrincipal* objPrin = nsContentUtils::ObjectPrincipal(js::UncheckedUnwrap(aObject));
return nsContentUtils::SubjectPrincipal()->Subsumes(objPrin);
}
nsresult
UnwrapArgImpl(JSContext* cx,
JS::Handle<JSObject*> src,
const nsIID &iid,
void **ppArg)
{
if (!NS_IsMainThread()) {
return NS_ERROR_NOT_AVAILABLE;
}
Bug 1371259 part 9. Make UnwrapReflectorToISupports return already_AddRefed<nsISupports>. r=peterv The main reason to not do this would be performance (avoiding the addref/release), but there are two main mitigating factors: 1) All calls to UnwrapReflectorToISupports that pass in a Web IDL object already do the addref (and in fact QI). So this only affects the XPCWrappedNative case. 2) The vast majority of the callers proceed to QI on the pointer anyway, and a second addref is cheap; it's the first addref after a CC that can be expensive on a cycle-collected object. Going through the changes one by one: * In GlobalObject::GetAsSupports, we do have a change that slightly slows down precisely in the XPCWrappedNative global case. That's the message managers and the backstagepass. And this really only affects calls to Web IDL statics from those globals. * In UnwrapArgImpl we're talking about a Web IDL method taking an "external interface" type, and the UnwrapReflectorToISupports call is immediately followed by QI anyway. * In UnwrapXPConnectImpl we're talking about the case when we have a non-WebIDL-object implementation of a Web IDL interface. Again, this is the message manager globals, for EventTarget. And we have a QI call immediately after the UnwrapReflectorToISupports. * In the generated HasInstance hook for EventTarget we will be slightly slower when the LHS of the instanceof is an XPCWrappedNative. And not much slower, because again there's an immediate QI. * In InstallXBLField we're never going to have an XPCWrappedNative as thisObj; it's always an Element in practice. So this is no more expensive than before. * In sandbox's GetPrincipalOrSOP we now have an extra addref. But it was followed by various QIs anyway. * In XPCConvert::JSValToXPCException we have an extra addref if someone throws an XPCWrappedNative, which is fairly unlikely; our actual Exception objects are on Web IDL bindings. Plus we have an immediate QI. * In xpc::HasInstance we have an extra addred if the LHS of instanceof is an XPCWrappedNative. But, again, there's an immediated QI after the UnwrapReflectorToISupports. * In xpcJSWeakReference::Init we are likely doing an extra addref, but again immediately followed by QI. I think it's worth making this change just to remove the footgun and that the perf impact, if any, is pretty minimal.
2017-07-10 23:05:26 +03:00
nsCOMPtr<nsISupports> iface = xpc::UnwrapReflectorToISupports(src);
if (iface) {
if (NS_FAILED(iface->QueryInterface(iid, ppArg))) {
return NS_ERROR_XPC_BAD_CONVERT_JS;
}
return NS_OK;
}
// Only allow XPCWrappedJS stuff in system code. Ideally we would remove this
// even there, but that involves converting some things to WebIDL callback
// interfaces and making some other things builtinclass...
if (!nsContentUtils::IsSystemCaller(cx)) {
return NS_ERROR_XPC_BAD_CONVERT_JS;
}
RefPtr<nsXPCWrappedJS> wrappedJS;
nsresult rv = nsXPCWrappedJS::GetNewOrUsed(src, iid, getter_AddRefs(wrappedJS));
if (NS_FAILED(rv) || !wrappedJS) {
return rv;
}
// We need to go through the QueryInterface logic to make this return
// the right thing for the various 'special' interfaces; e.g.
// nsIPropertyBag. We must use AggregatedQueryInterface in cases where
// there is an outer to avoid nasty recursion.
return wrappedJS->QueryInterface(iid, ppArg);
}
nsresult
UnwrapXPConnectImpl(JSContext* cx,
JS::MutableHandle<JS::Value> src,
const nsIID &iid,
void **ppArg)
{
if (!NS_IsMainThread()) {
return NS_ERROR_NOT_AVAILABLE;
}
MOZ_ASSERT(src.isObject());
// Unwrap ourselves, because we're going to want access to the unwrapped
// object.
JS::Rooted<JSObject*> obj(cx,
js::CheckedUnwrap(&src.toObject(),
/* stopAtWindowProxy = */ false));
if (!obj) {
return NS_ERROR_NOT_AVAILABLE;
}
Bug 1371259 part 9. Make UnwrapReflectorToISupports return already_AddRefed<nsISupports>. r=peterv The main reason to not do this would be performance (avoiding the addref/release), but there are two main mitigating factors: 1) All calls to UnwrapReflectorToISupports that pass in a Web IDL object already do the addref (and in fact QI). So this only affects the XPCWrappedNative case. 2) The vast majority of the callers proceed to QI on the pointer anyway, and a second addref is cheap; it's the first addref after a CC that can be expensive on a cycle-collected object. Going through the changes one by one: * In GlobalObject::GetAsSupports, we do have a change that slightly slows down precisely in the XPCWrappedNative global case. That's the message managers and the backstagepass. And this really only affects calls to Web IDL statics from those globals. * In UnwrapArgImpl we're talking about a Web IDL method taking an "external interface" type, and the UnwrapReflectorToISupports call is immediately followed by QI anyway. * In UnwrapXPConnectImpl we're talking about the case when we have a non-WebIDL-object implementation of a Web IDL interface. Again, this is the message manager globals, for EventTarget. And we have a QI call immediately after the UnwrapReflectorToISupports. * In the generated HasInstance hook for EventTarget we will be slightly slower when the LHS of the instanceof is an XPCWrappedNative. And not much slower, because again there's an immediate QI. * In InstallXBLField we're never going to have an XPCWrappedNative as thisObj; it's always an Element in practice. So this is no more expensive than before. * In sandbox's GetPrincipalOrSOP we now have an extra addref. But it was followed by various QIs anyway. * In XPCConvert::JSValToXPCException we have an extra addref if someone throws an XPCWrappedNative, which is fairly unlikely; our actual Exception objects are on Web IDL bindings. Plus we have an immediate QI. * In xpc::HasInstance we have an extra addred if the LHS of instanceof is an XPCWrappedNative. But, again, there's an immediated QI after the UnwrapReflectorToISupports. * In xpcJSWeakReference::Init we are likely doing an extra addref, but again immediately followed by QI. I think it's worth making this change just to remove the footgun and that the perf impact, if any, is pretty minimal.
2017-07-10 23:05:26 +03:00
nsCOMPtr<nsISupports> iface = xpc::UnwrapReflectorToISupports(obj);
if (!iface) {
return NS_ERROR_XPC_BAD_CONVERT_JS;
}
if (NS_FAILED(iface->QueryInterface(iid, ppArg))) {
return NS_ERROR_XPC_BAD_CONVERT_JS;
}
// Now update our source to keep rooting our object.
src.setObject(*obj);
return NS_OK;
}
nsresult
UnwrapWindowProxyImpl(JSContext* cx,
JS::Handle<JSObject*> src,
nsPIDOMWindowOuter** ppArg)
{
nsCOMPtr<nsPIDOMWindowInner> inner;
nsresult rv = UnwrapArg<nsPIDOMWindowInner>(cx, src, getter_AddRefs(inner));
NS_ENSURE_SUCCESS(rv, rv);
nsCOMPtr<nsPIDOMWindowOuter> outer = inner->GetOuterWindow();
outer.forget(ppArg);
return NS_OK;
}
bool
SystemGlobalResolve(JSContext* cx, JS::Handle<JSObject*> obj,
JS::Handle<jsid> id, bool* resolvedp)
{
if (!ResolveGlobal(cx, obj, id, resolvedp)) {
return false;
}
if (*resolvedp) {
return true;
}
return ResolveSystemBinding(cx, obj, id, resolvedp);
}
bool
SystemGlobalEnumerate(JSContext* cx, JS::Handle<JSObject*> obj)
{
bool ignored = false;
return JS_EnumerateStandardClasses(cx, obj) &&
ResolveSystemBinding(cx, obj, JSID_VOIDHANDLE, &ignored);
}
template<decltype(JS::NewMapObject) Method>
bool
GetMaplikeSetlikeBackingObject(JSContext* aCx, JS::Handle<JSObject*> aObj,
size_t aSlotIndex,
JS::MutableHandle<JSObject*> aBackingObj,
bool* aBackingObjCreated)
{
JS::Rooted<JSObject*> reflector(aCx);
reflector = IsDOMObject(aObj) ? aObj : js::UncheckedUnwrap(aObj,
/* stopAtWindowProxy = */ false);
// Retrieve the backing object from the reserved slot on the maplike/setlike
// object. If it doesn't exist yet, create it.
JS::Rooted<JS::Value> slotValue(aCx);
slotValue = js::GetReservedSlot(reflector, aSlotIndex);
if (slotValue.isUndefined()) {
// Since backing object access can happen in non-originating compartments,
// make sure to create the backing object in reflector compartment.
{
JSAutoCompartment ac(aCx, reflector);
JS::Rooted<JSObject*> newBackingObj(aCx);
newBackingObj.set(Method(aCx));
if (NS_WARN_IF(!newBackingObj)) {
return false;
}
js::SetReservedSlot(reflector, aSlotIndex, JS::ObjectValue(*newBackingObj));
}
slotValue = js::GetReservedSlot(reflector, aSlotIndex);
*aBackingObjCreated = true;
} else {
*aBackingObjCreated = false;
}
if (!MaybeWrapNonDOMObjectValue(aCx, &slotValue)) {
return false;
}
aBackingObj.set(&slotValue.toObject());
return true;
}
bool
GetMaplikeBackingObject(JSContext* aCx, JS::Handle<JSObject*> aObj,
size_t aSlotIndex,
JS::MutableHandle<JSObject*> aBackingObj,
bool* aBackingObjCreated)
{
return GetMaplikeSetlikeBackingObject<JS::NewMapObject>(aCx, aObj, aSlotIndex,
aBackingObj,
aBackingObjCreated);
}
bool
GetSetlikeBackingObject(JSContext* aCx, JS::Handle<JSObject*> aObj,
size_t aSlotIndex,
JS::MutableHandle<JSObject*> aBackingObj,
bool* aBackingObjCreated)
{
return GetMaplikeSetlikeBackingObject<JS::NewSetObject>(aCx, aObj, aSlotIndex,
aBackingObj,
aBackingObjCreated);
}
bool
ForEachHandler(JSContext* aCx, unsigned aArgc, JS::Value* aVp)
{
JS::CallArgs args = CallArgsFromVp(aArgc, aVp);
// Unpack callback and object from slots
JS::Rooted<JS::Value>
callbackFn(aCx, js::GetFunctionNativeReserved(&args.callee(),
FOREACH_CALLBACK_SLOT));
JS::Rooted<JS::Value>
maplikeOrSetlikeObj(aCx,
js::GetFunctionNativeReserved(&args.callee(),
FOREACH_MAPLIKEORSETLIKEOBJ_SLOT));
MOZ_ASSERT(aArgc == 3);
JS::AutoValueVector newArgs(aCx);
// Arguments are passed in as value, key, object. Keep value and key, replace
// object with the maplike/setlike object.
if (!newArgs.append(args.get(0))) {
return false;
}
if (!newArgs.append(args.get(1))) {
return false;
}
if (!newArgs.append(maplikeOrSetlikeObj)) {
return false;
}
JS::Rooted<JS::Value> rval(aCx, JS::UndefinedValue());
// Now actually call the user specified callback
return JS::Call(aCx, args.thisv(), callbackFn, newArgs, &rval);
}
static inline prototypes::ID
GetProtoIdForNewtarget(JS::Handle<JSObject*> aNewTarget)
{
const js::Class* newTargetClass = js::GetObjectClass(aNewTarget);
if (IsDOMIfaceAndProtoClass(newTargetClass)) {
const DOMIfaceAndProtoJSClass* newTargetIfaceClass =
DOMIfaceAndProtoJSClass::FromJSClass(newTargetClass);
if (newTargetIfaceClass->mType == eInterface) {
return newTargetIfaceClass->mPrototypeID;
}
} else if (JS_IsNativeFunction(aNewTarget, Constructor)) {
return GetNativePropertyHooksFromConstructorFunction(aNewTarget)->mPrototypeID;
}
return prototypes::id::_ID_Count;
}
bool
GetDesiredProto(JSContext* aCx, const JS::CallArgs& aCallArgs,
JS::MutableHandle<JSObject*> aDesiredProto)
{
if (!aCallArgs.isConstructing()) {
aDesiredProto.set(nullptr);
return true;
}
// The desired prototype depends on the actual constructor that was invoked,
// which is passed to us as the newTarget in the callargs. We want to do
// something akin to the ES6 specification's GetProtototypeFromConstructor (so
// get .prototype on the newTarget, with a fallback to some sort of default).
// First, a fast path for the case when the the constructor is in fact one of
// our DOM constructors. This is safe because on those the "constructor"
// property is non-configurable and non-writable, so we don't have to do the
// slow JS_GetProperty call.
JS::Rooted<JSObject*> newTarget(aCx, &aCallArgs.newTarget().toObject());
JS::Rooted<JSObject*> originalNewTarget(aCx, newTarget);
// See whether we have a known DOM constructor here, such that we can take a
// fast path.
prototypes::ID protoID = GetProtoIdForNewtarget(newTarget);
if (protoID == prototypes::id::_ID_Count) {
// We might still have a cross-compartment wrapper for a known DOM
// constructor.
newTarget = js::CheckedUnwrap(newTarget);
if (newTarget && newTarget != originalNewTarget) {
protoID = GetProtoIdForNewtarget(newTarget);
}
}
if (protoID != prototypes::id::_ID_Count) {
ProtoAndIfaceCache& protoAndIfaceCache =
*GetProtoAndIfaceCache(js::GetGlobalForObjectCrossCompartment(newTarget));
aDesiredProto.set(protoAndIfaceCache.EntrySlotMustExist(protoID));
if (newTarget != originalNewTarget) {
return JS_WrapObject(aCx, aDesiredProto);
}
return true;
}
// Slow path. This basically duplicates the ES6 spec's
// GetPrototypeFromConstructor except that instead of taking a string naming
// the fallback prototype we just fall back to using null and assume that our
// caller will then pick the right default. The actual defaulting behavior
// here still needs to be defined in the Web IDL specification.
//
// Note that it's very important to do this property get on originalNewTarget,
// not our unwrapped newTarget, since we want to get Xray behavior here as
// needed.
// XXXbz for speed purposes, using a preinterned id here sure would be nice.
JS::Rooted<JS::Value> protoVal(aCx);
if (!JS_GetProperty(aCx, originalNewTarget, "prototype", &protoVal)) {
return false;
}
if (!protoVal.isObject()) {
aDesiredProto.set(nullptr);
return true;
}
aDesiredProto.set(&protoVal.toObject());
return true;
}
CustomElementReactionsStack*
GetCustomElementReactionsStack(JS::Handle<JSObject*> aObj)
{
// This might not be the right object, if there are wrappers. Unwrap if we can.
JSObject* obj = js::CheckedUnwrap(aObj);
if (!obj) {
return nullptr;
}
nsGlobalWindow* window = xpc::WindowGlobalOrNull(obj);
if (!window) {
return nullptr;
}
DocGroup* docGroup = window->AsInner()->GetDocGroup();
if (!docGroup) {
return nullptr;
}
return docGroup->CustomElementReactionsStack();
}
// https://html.spec.whatwg.org/multipage/dom.html#htmlconstructor
already_AddRefed<nsGenericHTMLElement>
CreateHTMLElement(const GlobalObject& aGlobal, const JS::CallArgs& aCallArgs,
ErrorResult& aRv)
{
// Step 1.
nsCOMPtr<nsPIDOMWindowInner> window = do_QueryInterface(aGlobal.GetAsSupports());
if (!window) {
aRv.Throw(NS_ERROR_UNEXPECTED);
return nullptr;
}
nsIDocument* doc = window->GetExtantDoc();
if (!doc) {
aRv.Throw(NS_ERROR_UNEXPECTED);
return nullptr;
}
RefPtr<mozilla::dom::CustomElementRegistry> registry(window->CustomElements());
if (!registry) {
aRv.Throw(NS_ERROR_UNEXPECTED);
return nullptr;
}
// Step 2 is in the code output by CGClassConstructor.
// Step 3.
JSContext* cx = aGlobal.Context();
JS::Rooted<JSObject*> newTarget(cx, &aCallArgs.newTarget().toObject());
CustomElementDefinition* definition =
registry->LookupCustomElementDefinition(cx, newTarget);
if (!definition) {
aRv.ThrowTypeError<MSG_ILLEGAL_CONSTRUCTOR>();
return nullptr;
}
// The callee might be an Xray. Unwrap it to get actual callee.
JS::Rooted<JSObject*> callee(cx, js::CheckedUnwrap(&aCallArgs.callee()));
if (!callee) {
aRv.Throw(NS_ERROR_DOM_SECURITY_ERR);
return nullptr;
}
// And the actual callee might be in different compartment, so enter its
// compartment before getting the standard constructor object to compare to,
// so we get it from the same global as callee itself.
JSAutoCompartment ac(cx, callee);
int32_t tag = eHTMLTag_userdefined;
if (!definition->IsCustomBuiltIn()) {
// Step 4.
// If the definition is for an autonomous custom element, the active
// function should be HTMLElement.
JS::Rooted<JSObject*> constructor(cx, HTMLElementBinding::GetConstructorObject(cx));
if (!constructor) {
aRv.NoteJSContextException(cx);
return nullptr;
}
if (callee != constructor) {
aRv.ThrowTypeError<MSG_ILLEGAL_CONSTRUCTOR>();
return nullptr;
}
} else {
// Step 5.
// If the definition is for a customized built-in element, the localName
// should be defined in the specification.
nsIParserService* parserService = nsContentUtils::GetParserService();
if (!parserService) {
aRv.Throw(NS_ERROR_UNEXPECTED);
return nullptr;
}
tag = parserService->HTMLCaseSensitiveAtomTagToId(definition->mLocalName);
if (tag == eHTMLTag_userdefined) {
aRv.ThrowTypeError<MSG_ILLEGAL_CONSTRUCTOR>();
return nullptr;
}
MOZ_ASSERT(tag <= NS_HTML_TAG_MAX, "tag is out of bounds");
// If the definition is for a customized built-in element, the active
// function should be the localname's element interface.
constructorGetterCallback cb = sConstructorGetterCallback[tag];
if (!cb) {
aRv.ThrowTypeError<MSG_ILLEGAL_CONSTRUCTOR>();
return nullptr;
}
JS::Rooted<JSObject*> constructor(cx, cb(cx));
if (!constructor) {
aRv.NoteJSContextException(cx);
return nullptr;
}
if (callee != constructor) {
aRv.ThrowTypeError<MSG_ILLEGAL_CONSTRUCTOR>();
return nullptr;
}
}
RefPtr<mozilla::dom::NodeInfo> nodeInfo =
doc->NodeInfoManager()->GetNodeInfo(definition->mLocalName,
nullptr,
kNameSpaceID_XHTML,
nsIDOMNode::ELEMENT_NODE);
if (!nodeInfo) {
aRv.Throw(NS_ERROR_UNEXPECTED);
return nullptr;
}
// Step 6 and Step 7 are in the code output by CGClassConstructor.
// Step 8.
// Construction stack will be implemented in bug 1287348. So we always run
// "construction stack is empty" case for now.
RefPtr<nsGenericHTMLElement> element;
if (tag == eHTMLTag_userdefined) {
// Autonomous custom element.
element = NS_NewHTMLElement(nodeInfo.forget());
} else {
// Customized built-in element.
element = CreateHTMLElement(tag, nodeInfo.forget(), NOT_FROM_PARSER);
}
element->SetCustomElementData(
new CustomElementData(definition->mType, CustomElementData::State::eCustom));
return element.forget();
}
#ifdef DEBUG
namespace binding_detail {
void
AssertReflectorHasGivenProto(JSContext* aCx, JSObject* aReflector,
JS::Handle<JSObject*> aGivenProto)
{
if (!aGivenProto) {
// Nothing to assert here
return;
}
JS::Rooted<JSObject*> reflector(aCx, aReflector);
JSAutoCompartment ac(aCx, reflector);
JS::Rooted<JSObject*> reflectorProto(aCx);
bool ok = JS_GetPrototype(aCx, reflector, &reflectorProto);
MOZ_ASSERT(ok);
// aGivenProto may not be in the right compartment here, so we
// have to wrap it to compare.
JS::Rooted<JSObject*> givenProto(aCx, aGivenProto);
ok = JS_WrapObject(aCx, &givenProto);
MOZ_ASSERT(ok);
MOZ_ASSERT(givenProto == reflectorProto,
"How are we supposed to change the proto now?");
}
} // namespace binding_detail
#endif // DEBUG
void
SetDocumentAndPageUseCounter(JSContext* aCx, JSObject* aObject,
UseCounter aUseCounter)
{
nsGlobalWindow* win = xpc::WindowGlobalOrNull(js::UncheckedUnwrap(aObject));
if (win && win->GetDocument()) {
win->GetDocument()->SetDocumentAndPageUseCounter(aUseCounter);
}
}
namespace {
// This runnable is used to write a deprecation message from a worker to the
// console running on the main-thread.
class DeprecationWarningRunnable final : public WorkerProxyToMainThreadRunnable
{
nsIDocument::DeprecatedOperations mOperation;
public:
DeprecationWarningRunnable(WorkerPrivate* aWorkerPrivate,
nsIDocument::DeprecatedOperations aOperation)
: WorkerProxyToMainThreadRunnable(aWorkerPrivate)
, mOperation(aOperation)
{
MOZ_ASSERT(aWorkerPrivate);
aWorkerPrivate->AssertIsOnWorkerThread();
}
private:
void
RunOnMainThread() override
{
MOZ_ASSERT(NS_IsMainThread());
// Walk up to our containing page
WorkerPrivate* wp = mWorkerPrivate;
while (wp->GetParent()) {
wp = wp->GetParent();
}
nsPIDOMWindowInner* window = wp->GetWindow();
if (window && window->GetExtantDoc()) {
window->GetExtantDoc()->WarnOnceAbout(mOperation);
}
}
void
RunBackOnWorkerThreadForCleanup() override
{}
};
} // anonymous namespace
void
DeprecationWarning(JSContext* aCx, JSObject* aObject,
nsIDocument::DeprecatedOperations aOperation)
{
GlobalObject global(aCx, aObject);
if (global.Failed()) {
NS_ERROR("Could not create global for DeprecationWarning");
return;
}
DeprecationWarning(global, aOperation);
}
void
DeprecationWarning(const GlobalObject& aGlobal,
nsIDocument::DeprecatedOperations aOperation)
{
if (NS_IsMainThread()) {
nsCOMPtr<nsPIDOMWindowInner> window = do_QueryInterface(aGlobal.GetAsSupports());
if (window && window->GetExtantDoc()) {
window->GetExtantDoc()->WarnOnceAbout(aOperation);
}
return;
}
WorkerPrivate* workerPrivate = GetWorkerPrivateFromContext(aGlobal.Context());
if (!workerPrivate) {
return;
}
RefPtr<DeprecationWarningRunnable> runnable =
new DeprecationWarningRunnable(workerPrivate, aOperation);
runnable->Dispatch();
}
namespace binding_detail {
JSObject*
UnprivilegedJunkScopeOrWorkerGlobal()
{
if (NS_IsMainThread()) {
return xpc::UnprivilegedJunkScope();
}
return GetCurrentThreadWorkerGlobal();
}
} // namespace binding_detail
} // namespace dom
} // namespace mozilla