clang-1/lib/AST/Expr.cpp

2682 строки
92 KiB
C++
Исходник Обычный вид История

//===--- Expr.cpp - Expression AST Node Implementation --------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements the Expr class and subclasses.
//
//===----------------------------------------------------------------------===//
#include "clang/AST/Expr.h"
#include "clang/AST/ExprCXX.h"
#include "clang/AST/APValue.h"
#include "clang/AST/ASTContext.h"
#include "clang/AST/DeclObjC.h"
#include "clang/AST/DeclCXX.h"
#include "clang/AST/DeclTemplate.h"
#include "clang/AST/RecordLayout.h"
#include "clang/AST/StmtVisitor.h"
#include "clang/Basic/Builtins.h"
#include "clang/Basic/TargetInfo.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/raw_ostream.h"
#include <algorithm>
using namespace clang;
void Expr::ANCHOR() {} // key function for Expr class.
/// isKnownToHaveBooleanValue - Return true if this is an integer expression
/// that is known to return 0 or 1. This happens for _Bool/bool expressions
/// but also int expressions which are produced by things like comparisons in
/// C.
bool Expr::isKnownToHaveBooleanValue() const {
// If this value has _Bool type, it is obvious 0/1.
if (getType()->isBooleanType()) return true;
// If this is a non-scalar-integer type, we don't care enough to try.
if (!getType()->isIntegralOrEnumerationType()) return false;
if (const ParenExpr *PE = dyn_cast<ParenExpr>(this))
return PE->getSubExpr()->isKnownToHaveBooleanValue();
if (const UnaryOperator *UO = dyn_cast<UnaryOperator>(this)) {
switch (UO->getOpcode()) {
case UO_Plus:
case UO_Extension:
return UO->getSubExpr()->isKnownToHaveBooleanValue();
default:
return false;
}
}
// Only look through implicit casts. If the user writes
// '(int) (a && b)' treat it as an arbitrary int.
if (const ImplicitCastExpr *CE = dyn_cast<ImplicitCastExpr>(this))
return CE->getSubExpr()->isKnownToHaveBooleanValue();
if (const BinaryOperator *BO = dyn_cast<BinaryOperator>(this)) {
switch (BO->getOpcode()) {
default: return false;
case BO_LT: // Relational operators.
case BO_GT:
case BO_LE:
case BO_GE:
case BO_EQ: // Equality operators.
case BO_NE:
case BO_LAnd: // AND operator.
case BO_LOr: // Logical OR operator.
return true;
case BO_And: // Bitwise AND operator.
case BO_Xor: // Bitwise XOR operator.
case BO_Or: // Bitwise OR operator.
// Handle things like (x==2)|(y==12).
return BO->getLHS()->isKnownToHaveBooleanValue() &&
BO->getRHS()->isKnownToHaveBooleanValue();
case BO_Comma:
case BO_Assign:
return BO->getRHS()->isKnownToHaveBooleanValue();
}
}
if (const ConditionalOperator *CO = dyn_cast<ConditionalOperator>(this))
return CO->getTrueExpr()->isKnownToHaveBooleanValue() &&
CO->getFalseExpr()->isKnownToHaveBooleanValue();
return false;
}
//===----------------------------------------------------------------------===//
// Primary Expressions.
//===----------------------------------------------------------------------===//
void ExplicitTemplateArgumentList::initializeFrom(
const TemplateArgumentListInfo &Info) {
LAngleLoc = Info.getLAngleLoc();
RAngleLoc = Info.getRAngleLoc();
NumTemplateArgs = Info.size();
TemplateArgumentLoc *ArgBuffer = getTemplateArgs();
for (unsigned i = 0; i != NumTemplateArgs; ++i)
new (&ArgBuffer[i]) TemplateArgumentLoc(Info[i]);
}
void ExplicitTemplateArgumentList::copyInto(
TemplateArgumentListInfo &Info) const {
Info.setLAngleLoc(LAngleLoc);
Info.setRAngleLoc(RAngleLoc);
for (unsigned I = 0; I != NumTemplateArgs; ++I)
Info.addArgument(getTemplateArgs()[I]);
}
std::size_t ExplicitTemplateArgumentList::sizeFor(unsigned NumTemplateArgs) {
return sizeof(ExplicitTemplateArgumentList) +
sizeof(TemplateArgumentLoc) * NumTemplateArgs;
}
std::size_t ExplicitTemplateArgumentList::sizeFor(
const TemplateArgumentListInfo &Info) {
return sizeFor(Info.size());
}
void DeclRefExpr::computeDependence() {
ExprBits.TypeDependent = false;
ExprBits.ValueDependent = false;
NamedDecl *D = getDecl();
// (TD) C++ [temp.dep.expr]p3:
// An id-expression is type-dependent if it contains:
//
// and
//
// (VD) C++ [temp.dep.constexpr]p2:
// An identifier is value-dependent if it is:
// (TD) - an identifier that was declared with dependent type
// (VD) - a name declared with a dependent type,
if (getType()->isDependentType()) {
ExprBits.TypeDependent = true;
ExprBits.ValueDependent = true;
}
// (TD) - a conversion-function-id that specifies a dependent type
else if (D->getDeclName().getNameKind()
== DeclarationName::CXXConversionFunctionName &&
D->getDeclName().getCXXNameType()->isDependentType()) {
ExprBits.TypeDependent = true;
ExprBits.ValueDependent = true;
}
// (TD) - a template-id that is dependent,
else if (hasExplicitTemplateArgs() &&
TemplateSpecializationType::anyDependentTemplateArguments(
getTemplateArgs(),
getNumTemplateArgs())) {
ExprBits.TypeDependent = true;
ExprBits.ValueDependent = true;
}
// (VD) - the name of a non-type template parameter,
else if (isa<NonTypeTemplateParmDecl>(D))
ExprBits.ValueDependent = true;
// (VD) - a constant with integral or enumeration type and is
// initialized with an expression that is value-dependent.
else if (VarDecl *Var = dyn_cast<VarDecl>(D)) {
if (Var->getType()->isIntegralOrEnumerationType() &&
Var->getType().getCVRQualifiers() == Qualifiers::Const) {
if (const Expr *Init = Var->getAnyInitializer())
if (Init->isValueDependent())
ExprBits.ValueDependent = true;
}
// (VD) - FIXME: Missing from the standard:
// - a member function or a static data member of the current
// instantiation
else if (Var->isStaticDataMember() &&
Var->getDeclContext()->isDependentContext())
ExprBits.ValueDependent = true;
}
// (VD) - FIXME: Missing from the standard:
// - a member function or a static data member of the current
// instantiation
else if (isa<CXXMethodDecl>(D) && D->getDeclContext()->isDependentContext())
ExprBits.ValueDependent = true;
// (TD) - a nested-name-specifier or a qualified-id that names a
// member of an unknown specialization.
// (handled by DependentScopeDeclRefExpr)
}
DeclRefExpr::DeclRefExpr(NestedNameSpecifier *Qualifier,
SourceRange QualifierRange,
ValueDecl *D, SourceLocation NameLoc,
const TemplateArgumentListInfo *TemplateArgs,
QualType T)
: Expr(DeclRefExprClass, T, false, false),
DecoratedD(D,
(Qualifier? HasQualifierFlag : 0) |
(TemplateArgs ? HasExplicitTemplateArgumentListFlag : 0)),
Loc(NameLoc) {
if (Qualifier) {
NameQualifier *NQ = getNameQualifier();
NQ->NNS = Qualifier;
NQ->Range = QualifierRange;
}
if (TemplateArgs)
getExplicitTemplateArgs().initializeFrom(*TemplateArgs);
computeDependence();
}
DeclRefExpr::DeclRefExpr(NestedNameSpecifier *Qualifier,
SourceRange QualifierRange,
ValueDecl *D, const DeclarationNameInfo &NameInfo,
const TemplateArgumentListInfo *TemplateArgs,
QualType T)
: Expr(DeclRefExprClass, T, false, false),
DecoratedD(D,
(Qualifier? HasQualifierFlag : 0) |
(TemplateArgs ? HasExplicitTemplateArgumentListFlag : 0)),
Loc(NameInfo.getLoc()), DNLoc(NameInfo.getInfo()) {
if (Qualifier) {
NameQualifier *NQ = getNameQualifier();
NQ->NNS = Qualifier;
NQ->Range = QualifierRange;
}
if (TemplateArgs)
getExplicitTemplateArgs().initializeFrom(*TemplateArgs);
computeDependence();
}
DeclRefExpr *DeclRefExpr::Create(ASTContext &Context,
NestedNameSpecifier *Qualifier,
SourceRange QualifierRange,
ValueDecl *D,
SourceLocation NameLoc,
QualType T,
const TemplateArgumentListInfo *TemplateArgs) {
return Create(Context, Qualifier, QualifierRange, D,
DeclarationNameInfo(D->getDeclName(), NameLoc),
T, TemplateArgs);
}
DeclRefExpr *DeclRefExpr::Create(ASTContext &Context,
NestedNameSpecifier *Qualifier,
SourceRange QualifierRange,
ValueDecl *D,
const DeclarationNameInfo &NameInfo,
QualType T,
const TemplateArgumentListInfo *TemplateArgs) {
std::size_t Size = sizeof(DeclRefExpr);
if (Qualifier != 0)
Size += sizeof(NameQualifier);
if (TemplateArgs)
Size += ExplicitTemplateArgumentList::sizeFor(*TemplateArgs);
void *Mem = Context.Allocate(Size, llvm::alignOf<DeclRefExpr>());
return new (Mem) DeclRefExpr(Qualifier, QualifierRange, D, NameInfo,
TemplateArgs, T);
}
DeclRefExpr *DeclRefExpr::CreateEmpty(ASTContext &Context, bool HasQualifier,
unsigned NumTemplateArgs) {
std::size_t Size = sizeof(DeclRefExpr);
if (HasQualifier)
Size += sizeof(NameQualifier);
if (NumTemplateArgs)
Size += ExplicitTemplateArgumentList::sizeFor(NumTemplateArgs);
void *Mem = Context.Allocate(Size, llvm::alignOf<DeclRefExpr>());
return new (Mem) DeclRefExpr(EmptyShell());
}
SourceRange DeclRefExpr::getSourceRange() const {
SourceRange R = getNameInfo().getSourceRange();
if (hasQualifier())
R.setBegin(getQualifierRange().getBegin());
if (hasExplicitTemplateArgs())
R.setEnd(getRAngleLoc());
return R;
}
// FIXME: Maybe this should use DeclPrinter with a special "print predefined
// expr" policy instead.
std::string PredefinedExpr::ComputeName(IdentType IT, const Decl *CurrentDecl) {
ASTContext &Context = CurrentDecl->getASTContext();
if (const FunctionDecl *FD = dyn_cast<FunctionDecl>(CurrentDecl)) {
if (IT != PrettyFunction && IT != PrettyFunctionNoVirtual)
return FD->getNameAsString();
llvm::SmallString<256> Name;
llvm::raw_svector_ostream Out(Name);
if (const CXXMethodDecl *MD = dyn_cast<CXXMethodDecl>(FD)) {
if (MD->isVirtual() && IT != PrettyFunctionNoVirtual)
Out << "virtual ";
if (MD->isStatic())
Out << "static ";
}
PrintingPolicy Policy(Context.getLangOptions());
std::string Proto = FD->getQualifiedNameAsString(Policy);
const FunctionType *AFT = FD->getType()->getAs<FunctionType>();
const FunctionProtoType *FT = 0;
if (FD->hasWrittenPrototype())
FT = dyn_cast<FunctionProtoType>(AFT);
Proto += "(";
if (FT) {
llvm::raw_string_ostream POut(Proto);
for (unsigned i = 0, e = FD->getNumParams(); i != e; ++i) {
if (i) POut << ", ";
std::string Param;
FD->getParamDecl(i)->getType().getAsStringInternal(Param, Policy);
POut << Param;
}
if (FT->isVariadic()) {
if (FD->getNumParams()) POut << ", ";
POut << "...";
}
}
Proto += ")";
if (const CXXMethodDecl *MD = dyn_cast<CXXMethodDecl>(FD)) {
Qualifiers ThisQuals = Qualifiers::fromCVRMask(MD->getTypeQualifiers());
if (ThisQuals.hasConst())
Proto += " const";
if (ThisQuals.hasVolatile())
Proto += " volatile";
}
if (!isa<CXXConstructorDecl>(FD) && !isa<CXXDestructorDecl>(FD))
AFT->getResultType().getAsStringInternal(Proto, Policy);
Out << Proto;
Out.flush();
return Name.str().str();
}
if (const ObjCMethodDecl *MD = dyn_cast<ObjCMethodDecl>(CurrentDecl)) {
llvm::SmallString<256> Name;
llvm::raw_svector_ostream Out(Name);
Out << (MD->isInstanceMethod() ? '-' : '+');
Out << '[';
// For incorrect code, there might not be an ObjCInterfaceDecl. Do
// a null check to avoid a crash.
if (const ObjCInterfaceDecl *ID = MD->getClassInterface())
Out << ID;
if (const ObjCCategoryImplDecl *CID =
dyn_cast<ObjCCategoryImplDecl>(MD->getDeclContext()))
Out << '(' << CID << ')';
Out << ' ';
Out << MD->getSelector().getAsString();
Out << ']';
Out.flush();
return Name.str().str();
}
if (isa<TranslationUnitDecl>(CurrentDecl) && IT == PrettyFunction) {
// __PRETTY_FUNCTION__ -> "top level", the others produce an empty string.
return "top level";
}
return "";
}
void APNumericStorage::setIntValue(ASTContext &C, const llvm::APInt &Val) {
if (hasAllocation())
C.Deallocate(pVal);
BitWidth = Val.getBitWidth();
unsigned NumWords = Val.getNumWords();
const uint64_t* Words = Val.getRawData();
if (NumWords > 1) {
pVal = new (C) uint64_t[NumWords];
std::copy(Words, Words + NumWords, pVal);
} else if (NumWords == 1)
VAL = Words[0];
else
VAL = 0;
}
IntegerLiteral *
IntegerLiteral::Create(ASTContext &C, const llvm::APInt &V,
QualType type, SourceLocation l) {
return new (C) IntegerLiteral(C, V, type, l);
}
IntegerLiteral *
IntegerLiteral::Create(ASTContext &C, EmptyShell Empty) {
return new (C) IntegerLiteral(Empty);
}
FloatingLiteral *
FloatingLiteral::Create(ASTContext &C, const llvm::APFloat &V,
bool isexact, QualType Type, SourceLocation L) {
return new (C) FloatingLiteral(C, V, isexact, Type, L);
}
FloatingLiteral *
FloatingLiteral::Create(ASTContext &C, EmptyShell Empty) {
return new (C) FloatingLiteral(Empty);
}
/// getValueAsApproximateDouble - This returns the value as an inaccurate
/// double. Note that this may cause loss of precision, but is useful for
/// debugging dumps, etc.
double FloatingLiteral::getValueAsApproximateDouble() const {
llvm::APFloat V = getValue();
bool ignored;
V.convert(llvm::APFloat::IEEEdouble, llvm::APFloat::rmNearestTiesToEven,
&ignored);
return V.convertToDouble();
}
StringLiteral *StringLiteral::Create(ASTContext &C, const char *StrData,
unsigned ByteLength, bool Wide,
QualType Ty,
const SourceLocation *Loc,
unsigned NumStrs) {
// Allocate enough space for the StringLiteral plus an array of locations for
// any concatenated string tokens.
void *Mem = C.Allocate(sizeof(StringLiteral)+
sizeof(SourceLocation)*(NumStrs-1),
llvm::alignOf<StringLiteral>());
StringLiteral *SL = new (Mem) StringLiteral(Ty);
// OPTIMIZE: could allocate this appended to the StringLiteral.
char *AStrData = new (C, 1) char[ByteLength];
memcpy(AStrData, StrData, ByteLength);
SL->StrData = AStrData;
SL->ByteLength = ByteLength;
SL->IsWide = Wide;
SL->TokLocs[0] = Loc[0];
SL->NumConcatenated = NumStrs;
if (NumStrs != 1)
memcpy(&SL->TokLocs[1], Loc+1, sizeof(SourceLocation)*(NumStrs-1));
return SL;
}
StringLiteral *StringLiteral::CreateEmpty(ASTContext &C, unsigned NumStrs) {
void *Mem = C.Allocate(sizeof(StringLiteral)+
sizeof(SourceLocation)*(NumStrs-1),
llvm::alignOf<StringLiteral>());
StringLiteral *SL = new (Mem) StringLiteral(QualType());
SL->StrData = 0;
SL->ByteLength = 0;
SL->NumConcatenated = NumStrs;
return SL;
}
void StringLiteral::setString(ASTContext &C, llvm::StringRef Str) {
char *AStrData = new (C, 1) char[Str.size()];
memcpy(AStrData, Str.data(), Str.size());
StrData = AStrData;
ByteLength = Str.size();
}
/// getOpcodeStr - Turn an Opcode enum value into the punctuation char it
/// corresponds to, e.g. "sizeof" or "[pre]++".
const char *UnaryOperator::getOpcodeStr(Opcode Op) {
switch (Op) {
default: assert(0 && "Unknown unary operator");
case UO_PostInc: return "++";
case UO_PostDec: return "--";
case UO_PreInc: return "++";
case UO_PreDec: return "--";
case UO_AddrOf: return "&";
case UO_Deref: return "*";
case UO_Plus: return "+";
case UO_Minus: return "-";
case UO_Not: return "~";
case UO_LNot: return "!";
case UO_Real: return "__real";
case UO_Imag: return "__imag";
case UO_Extension: return "__extension__";
}
}
UnaryOperatorKind
UnaryOperator::getOverloadedOpcode(OverloadedOperatorKind OO, bool Postfix) {
switch (OO) {
default: assert(false && "No unary operator for overloaded function");
case OO_PlusPlus: return Postfix ? UO_PostInc : UO_PreInc;
case OO_MinusMinus: return Postfix ? UO_PostDec : UO_PreDec;
case OO_Amp: return UO_AddrOf;
case OO_Star: return UO_Deref;
case OO_Plus: return UO_Plus;
case OO_Minus: return UO_Minus;
case OO_Tilde: return UO_Not;
case OO_Exclaim: return UO_LNot;
}
}
OverloadedOperatorKind UnaryOperator::getOverloadedOperator(Opcode Opc) {
switch (Opc) {
case UO_PostInc: case UO_PreInc: return OO_PlusPlus;
case UO_PostDec: case UO_PreDec: return OO_MinusMinus;
case UO_AddrOf: return OO_Amp;
case UO_Deref: return OO_Star;
case UO_Plus: return OO_Plus;
case UO_Minus: return OO_Minus;
case UO_Not: return OO_Tilde;
case UO_LNot: return OO_Exclaim;
default: return OO_None;
}
}
//===----------------------------------------------------------------------===//
// Postfix Operators.
//===----------------------------------------------------------------------===//
CallExpr::CallExpr(ASTContext& C, StmtClass SC, Expr *fn, Expr **args,
Overhaul of Stmt allocation: - Made allocation of Stmt objects using vanilla new/delete a *compiler error* by making this new/delete "protected" within class Stmt. - Now the only way to allocate Stmt objects is by using the new operator that takes ASTContext& as an argument. This ensures that all Stmt nodes are allocated from the same (pool) allocator. - Naturally, these two changes required that *all* creation sites for AST nodes use new (ASTContext&). This is a large patch, but the majority of the changes are just this mechanical adjustment. - The above changes also mean that AST nodes can no longer be deallocated using 'delete'. Instead, one most do StmtObject->Destroy(ASTContext&) or do ASTContextObject.Deallocate(StmtObject) (the latter not running the 'Destroy' method). Along the way I also... - Made CompoundStmt allocate its array of Stmt* using the allocator in ASTContext (previously it used std::vector). There are a whole bunch of other Stmt classes that need to be similarly changed to ensure that all memory allocated for ASTs comes from the allocator in ASTContext. - Added a new smart pointer ExprOwningPtr to Sema.h. This replaces the uses of llvm::OwningPtr within Sema, as llvm::OwningPtr used 'delete' to free memory instead of a Stmt's 'Destroy' method. Big thanks to Doug Gregor for helping with the acrobatics of making 'new/delete' private and the new smart pointer ExprOwningPtr! git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@63997 91177308-0d34-0410-b5e6-96231b3b80d8
2009-02-07 04:47:29 +03:00
unsigned numargs, QualType t, SourceLocation rparenloc)
: Expr(SC, t,
fn->isTypeDependent() || hasAnyTypeDependentArguments(args, numargs),
fn->isValueDependent() || hasAnyValueDependentArguments(args,numargs)),
NumArgs(numargs) {
SubExprs = new (C) Stmt*[numargs+1];
SubExprs[FN] = fn;
for (unsigned i = 0; i != numargs; ++i)
SubExprs[i+ARGS_START] = args[i];
RParenLoc = rparenloc;
}
CallExpr::CallExpr(ASTContext& C, Expr *fn, Expr **args, unsigned numargs,
QualType t, SourceLocation rparenloc)
: Expr(CallExprClass, t,
fn->isTypeDependent() || hasAnyTypeDependentArguments(args, numargs),
fn->isValueDependent() || hasAnyValueDependentArguments(args,numargs)),
NumArgs(numargs) {
SubExprs = new (C) Stmt*[numargs+1];
SubExprs[FN] = fn;
for (unsigned i = 0; i != numargs; ++i)
SubExprs[i+ARGS_START] = args[i];
RParenLoc = rparenloc;
}
CallExpr::CallExpr(ASTContext &C, StmtClass SC, EmptyShell Empty)
: Expr(SC, Empty), SubExprs(0), NumArgs(0) {
SubExprs = new (C) Stmt*[1];
}
Decl *CallExpr::getCalleeDecl() {
Expr *CEE = getCallee()->IgnoreParenCasts();
// If we're calling a dereference, look at the pointer instead.
if (BinaryOperator *BO = dyn_cast<BinaryOperator>(CEE)) {
if (BO->isPtrMemOp())
CEE = BO->getRHS()->IgnoreParenCasts();
} else if (UnaryOperator *UO = dyn_cast<UnaryOperator>(CEE)) {
if (UO->getOpcode() == UO_Deref)
CEE = UO->getSubExpr()->IgnoreParenCasts();
}
if (DeclRefExpr *DRE = dyn_cast<DeclRefExpr>(CEE))
return DRE->getDecl();
if (MemberExpr *ME = dyn_cast<MemberExpr>(CEE))
return ME->getMemberDecl();
return 0;
}
FunctionDecl *CallExpr::getDirectCallee() {
return dyn_cast_or_null<FunctionDecl>(getCalleeDecl());
}
/// setNumArgs - This changes the number of arguments present in this call.
/// Any orphaned expressions are deleted by this, and any new operands are set
/// to null.
Overhaul of Stmt allocation: - Made allocation of Stmt objects using vanilla new/delete a *compiler error* by making this new/delete "protected" within class Stmt. - Now the only way to allocate Stmt objects is by using the new operator that takes ASTContext& as an argument. This ensures that all Stmt nodes are allocated from the same (pool) allocator. - Naturally, these two changes required that *all* creation sites for AST nodes use new (ASTContext&). This is a large patch, but the majority of the changes are just this mechanical adjustment. - The above changes also mean that AST nodes can no longer be deallocated using 'delete'. Instead, one most do StmtObject->Destroy(ASTContext&) or do ASTContextObject.Deallocate(StmtObject) (the latter not running the 'Destroy' method). Along the way I also... - Made CompoundStmt allocate its array of Stmt* using the allocator in ASTContext (previously it used std::vector). There are a whole bunch of other Stmt classes that need to be similarly changed to ensure that all memory allocated for ASTs comes from the allocator in ASTContext. - Added a new smart pointer ExprOwningPtr to Sema.h. This replaces the uses of llvm::OwningPtr within Sema, as llvm::OwningPtr used 'delete' to free memory instead of a Stmt's 'Destroy' method. Big thanks to Doug Gregor for helping with the acrobatics of making 'new/delete' private and the new smart pointer ExprOwningPtr! git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@63997 91177308-0d34-0410-b5e6-96231b3b80d8
2009-02-07 04:47:29 +03:00
void CallExpr::setNumArgs(ASTContext& C, unsigned NumArgs) {
// No change, just return.
if (NumArgs == getNumArgs()) return;
// If shrinking # arguments, just delete the extras and forgot them.
if (NumArgs < getNumArgs()) {
this->NumArgs = NumArgs;
return;
}
// Otherwise, we are growing the # arguments. New an bigger argument array.
Stmt **NewSubExprs = new (C) Stmt*[NumArgs+1];
// Copy over args.
for (unsigned i = 0; i != getNumArgs()+ARGS_START; ++i)
NewSubExprs[i] = SubExprs[i];
// Null out new args.
for (unsigned i = getNumArgs()+ARGS_START; i != NumArgs+ARGS_START; ++i)
NewSubExprs[i] = 0;
if (SubExprs) C.Deallocate(SubExprs);
SubExprs = NewSubExprs;
this->NumArgs = NumArgs;
}
/// isBuiltinCall - If this is a call to a builtin, return the builtin ID. If
/// not, return 0.
unsigned CallExpr::isBuiltinCall(ASTContext &Context) const {
// All simple function calls (e.g. func()) are implicitly cast to pointer to
// function. As a result, we try and obtain the DeclRefExpr from the
// ImplicitCastExpr.
const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(getCallee());
if (!ICE) // FIXME: deal with more complex calls (e.g. (func)(), (*func)()).
return 0;
const DeclRefExpr *DRE = dyn_cast<DeclRefExpr>(ICE->getSubExpr());
if (!DRE)
return 0;
const FunctionDecl *FDecl = dyn_cast<FunctionDecl>(DRE->getDecl());
if (!FDecl)
return 0;
if (!FDecl->getIdentifier())
return 0;
return FDecl->getBuiltinID();
}
QualType CallExpr::getCallReturnType() const {
QualType CalleeType = getCallee()->getType();
if (const PointerType *FnTypePtr = CalleeType->getAs<PointerType>())
CalleeType = FnTypePtr->getPointeeType();
else if (const BlockPointerType *BPT = CalleeType->getAs<BlockPointerType>())
CalleeType = BPT->getPointeeType();
else if (const MemberPointerType *MPT
= CalleeType->getAs<MemberPointerType>())
CalleeType = MPT->getPointeeType();
const FunctionType *FnType = CalleeType->getAs<FunctionType>();
return FnType->getResultType();
}
OffsetOfExpr *OffsetOfExpr::Create(ASTContext &C, QualType type,
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
SourceLocation OperatorLoc,
TypeSourceInfo *tsi,
OffsetOfNode* compsPtr, unsigned numComps,
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
Expr** exprsPtr, unsigned numExprs,
SourceLocation RParenLoc) {
void *Mem = C.Allocate(sizeof(OffsetOfExpr) +
sizeof(OffsetOfNode) * numComps +
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
sizeof(Expr*) * numExprs);
return new (Mem) OffsetOfExpr(C, type, OperatorLoc, tsi, compsPtr, numComps,
exprsPtr, numExprs, RParenLoc);
}
OffsetOfExpr *OffsetOfExpr::CreateEmpty(ASTContext &C,
unsigned numComps, unsigned numExprs) {
void *Mem = C.Allocate(sizeof(OffsetOfExpr) +
sizeof(OffsetOfNode) * numComps +
sizeof(Expr*) * numExprs);
return new (Mem) OffsetOfExpr(numComps, numExprs);
}
OffsetOfExpr::OffsetOfExpr(ASTContext &C, QualType type,
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
SourceLocation OperatorLoc, TypeSourceInfo *tsi,
OffsetOfNode* compsPtr, unsigned numComps,
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
Expr** exprsPtr, unsigned numExprs,
SourceLocation RParenLoc)
: Expr(OffsetOfExprClass, type, /*TypeDependent=*/false,
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
/*ValueDependent=*/tsi->getType()->isDependentType() ||
hasAnyTypeDependentArguments(exprsPtr, numExprs) ||
hasAnyValueDependentArguments(exprsPtr, numExprs)),
OperatorLoc(OperatorLoc), RParenLoc(RParenLoc), TSInfo(tsi),
NumComps(numComps), NumExprs(numExprs)
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
{
for(unsigned i = 0; i < numComps; ++i) {
setComponent(i, compsPtr[i]);
}
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
for(unsigned i = 0; i < numExprs; ++i) {
setIndexExpr(i, exprsPtr[i]);
}
}
IdentifierInfo *OffsetOfExpr::OffsetOfNode::getFieldName() const {
assert(getKind() == Field || getKind() == Identifier);
if (getKind() == Field)
return getField()->getIdentifier();
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
return reinterpret_cast<IdentifierInfo *> (Data & ~(uintptr_t)Mask);
}
MemberExpr *MemberExpr::Create(ASTContext &C, Expr *base, bool isarrow,
NestedNameSpecifier *qual,
SourceRange qualrange,
ValueDecl *memberdecl,
DeclAccessPair founddecl,
DeclarationNameInfo nameinfo,
const TemplateArgumentListInfo *targs,
QualType ty) {
std::size_t Size = sizeof(MemberExpr);
bool hasQualOrFound = (qual != 0 ||
founddecl.getDecl() != memberdecl ||
founddecl.getAccess() != memberdecl->getAccess());
if (hasQualOrFound)
Size += sizeof(MemberNameQualifier);
if (targs)
Size += ExplicitTemplateArgumentList::sizeFor(*targs);
void *Mem = C.Allocate(Size, llvm::alignOf<MemberExpr>());
MemberExpr *E = new (Mem) MemberExpr(base, isarrow, memberdecl, nameinfo, ty);
if (hasQualOrFound) {
if (qual && qual->isDependent()) {
E->setValueDependent(true);
E->setTypeDependent(true);
}
E->HasQualifierOrFoundDecl = true;
MemberNameQualifier *NQ = E->getMemberQualifier();
NQ->NNS = qual;
NQ->Range = qualrange;
NQ->FoundDecl = founddecl;
}
if (targs) {
E->HasExplicitTemplateArgumentList = true;
E->getExplicitTemplateArgs().initializeFrom(*targs);
}
return E;
}
const char *CastExpr::getCastKindName() const {
switch (getCastKind()) {
case CK_Unknown:
return "Unknown";
case CK_BitCast:
return "BitCast";
case CK_LValueBitCast:
return "LValueBitCast";
case CK_NoOp:
return "NoOp";
case CK_BaseToDerived:
return "BaseToDerived";
case CK_DerivedToBase:
return "DerivedToBase";
case CK_UncheckedDerivedToBase:
return "UncheckedDerivedToBase";
case CK_Dynamic:
return "Dynamic";
case CK_ToUnion:
return "ToUnion";
case CK_ArrayToPointerDecay:
return "ArrayToPointerDecay";
case CK_FunctionToPointerDecay:
return "FunctionToPointerDecay";
case CK_NullToMemberPointer:
return "NullToMemberPointer";
case CK_BaseToDerivedMemberPointer:
return "BaseToDerivedMemberPointer";
case CK_DerivedToBaseMemberPointer:
return "DerivedToBaseMemberPointer";
case CK_UserDefinedConversion:
return "UserDefinedConversion";
case CK_ConstructorConversion:
return "ConstructorConversion";
case CK_IntegralToPointer:
return "IntegralToPointer";
case CK_PointerToIntegral:
return "PointerToIntegral";
case CK_ToVoid:
return "ToVoid";
case CK_VectorSplat:
return "VectorSplat";
case CK_IntegralCast:
return "IntegralCast";
case CK_IntegralToFloating:
return "IntegralToFloating";
case CK_FloatingToIntegral:
return "FloatingToIntegral";
case CK_FloatingCast:
return "FloatingCast";
case CK_MemberPointerToBoolean:
return "MemberPointerToBoolean";
case CK_AnyPointerToObjCPointerCast:
return "AnyPointerToObjCPointerCast";
case CK_AnyPointerToBlockPointerCast:
return "AnyPointerToBlockPointerCast";
case CK_ObjCObjectLValueCast:
return "ObjCObjectLValueCast";
}
assert(0 && "Unhandled cast kind!");
return 0;
}
Expr *CastExpr::getSubExprAsWritten() {
Expr *SubExpr = 0;
CastExpr *E = this;
do {
SubExpr = E->getSubExpr();
// Skip any temporary bindings; they're implicit.
if (CXXBindTemporaryExpr *Binder = dyn_cast<CXXBindTemporaryExpr>(SubExpr))
SubExpr = Binder->getSubExpr();
// Conversions by constructor and conversion functions have a
// subexpression describing the call; strip it off.
if (E->getCastKind() == CK_ConstructorConversion)
SubExpr = cast<CXXConstructExpr>(SubExpr)->getArg(0);
else if (E->getCastKind() == CK_UserDefinedConversion)
SubExpr = cast<CXXMemberCallExpr>(SubExpr)->getImplicitObjectArgument();
// If the subexpression we're left with is an implicit cast, look
// through that, too.
} while ((E = dyn_cast<ImplicitCastExpr>(SubExpr)));
return SubExpr;
}
CXXBaseSpecifier **CastExpr::path_buffer() {
switch (getStmtClass()) {
#define ABSTRACT_STMT(x)
#define CASTEXPR(Type, Base) \
case Stmt::Type##Class: \
return reinterpret_cast<CXXBaseSpecifier**>(static_cast<Type*>(this)+1);
#define STMT(Type, Base)
#include "clang/AST/StmtNodes.inc"
default:
llvm_unreachable("non-cast expressions not possible here");
return 0;
}
}
void CastExpr::setCastPath(const CXXCastPath &Path) {
assert(Path.size() == path_size());
memcpy(path_buffer(), Path.data(), Path.size() * sizeof(CXXBaseSpecifier*));
}
ImplicitCastExpr *ImplicitCastExpr::Create(ASTContext &C, QualType T,
CastKind Kind, Expr *Operand,
const CXXCastPath *BasePath,
ExprValueKind VK) {
unsigned PathSize = (BasePath ? BasePath->size() : 0);
void *Buffer =
C.Allocate(sizeof(ImplicitCastExpr) + PathSize * sizeof(CXXBaseSpecifier*));
ImplicitCastExpr *E =
new (Buffer) ImplicitCastExpr(T, Kind, Operand, PathSize, VK);
if (PathSize) E->setCastPath(*BasePath);
return E;
}
ImplicitCastExpr *ImplicitCastExpr::CreateEmpty(ASTContext &C,
unsigned PathSize) {
void *Buffer =
C.Allocate(sizeof(ImplicitCastExpr) + PathSize * sizeof(CXXBaseSpecifier*));
return new (Buffer) ImplicitCastExpr(EmptyShell(), PathSize);
}
CStyleCastExpr *CStyleCastExpr::Create(ASTContext &C, QualType T,
CastKind K, Expr *Op,
const CXXCastPath *BasePath,
TypeSourceInfo *WrittenTy,
SourceLocation L, SourceLocation R) {
unsigned PathSize = (BasePath ? BasePath->size() : 0);
void *Buffer =
C.Allocate(sizeof(CStyleCastExpr) + PathSize * sizeof(CXXBaseSpecifier*));
CStyleCastExpr *E =
new (Buffer) CStyleCastExpr(T, K, Op, PathSize, WrittenTy, L, R);
if (PathSize) E->setCastPath(*BasePath);
return E;
}
CStyleCastExpr *CStyleCastExpr::CreateEmpty(ASTContext &C, unsigned PathSize) {
void *Buffer =
C.Allocate(sizeof(CStyleCastExpr) + PathSize * sizeof(CXXBaseSpecifier*));
return new (Buffer) CStyleCastExpr(EmptyShell(), PathSize);
}
/// getOpcodeStr - Turn an Opcode enum value into the punctuation char it
/// corresponds to, e.g. "<<=".
const char *BinaryOperator::getOpcodeStr(Opcode Op) {
switch (Op) {
case BO_PtrMemD: return ".*";
case BO_PtrMemI: return "->*";
case BO_Mul: return "*";
case BO_Div: return "/";
case BO_Rem: return "%";
case BO_Add: return "+";
case BO_Sub: return "-";
case BO_Shl: return "<<";
case BO_Shr: return ">>";
case BO_LT: return "<";
case BO_GT: return ">";
case BO_LE: return "<=";
case BO_GE: return ">=";
case BO_EQ: return "==";
case BO_NE: return "!=";
case BO_And: return "&";
case BO_Xor: return "^";
case BO_Or: return "|";
case BO_LAnd: return "&&";
case BO_LOr: return "||";
case BO_Assign: return "=";
case BO_MulAssign: return "*=";
case BO_DivAssign: return "/=";
case BO_RemAssign: return "%=";
case BO_AddAssign: return "+=";
case BO_SubAssign: return "-=";
case BO_ShlAssign: return "<<=";
case BO_ShrAssign: return ">>=";
case BO_AndAssign: return "&=";
case BO_XorAssign: return "^=";
case BO_OrAssign: return "|=";
case BO_Comma: return ",";
}
return "";
}
BinaryOperatorKind
BinaryOperator::getOverloadedOpcode(OverloadedOperatorKind OO) {
switch (OO) {
default: assert(false && "Not an overloadable binary operator");
case OO_Plus: return BO_Add;
case OO_Minus: return BO_Sub;
case OO_Star: return BO_Mul;
case OO_Slash: return BO_Div;
case OO_Percent: return BO_Rem;
case OO_Caret: return BO_Xor;
case OO_Amp: return BO_And;
case OO_Pipe: return BO_Or;
case OO_Equal: return BO_Assign;
case OO_Less: return BO_LT;
case OO_Greater: return BO_GT;
case OO_PlusEqual: return BO_AddAssign;
case OO_MinusEqual: return BO_SubAssign;
case OO_StarEqual: return BO_MulAssign;
case OO_SlashEqual: return BO_DivAssign;
case OO_PercentEqual: return BO_RemAssign;
case OO_CaretEqual: return BO_XorAssign;
case OO_AmpEqual: return BO_AndAssign;
case OO_PipeEqual: return BO_OrAssign;
case OO_LessLess: return BO_Shl;
case OO_GreaterGreater: return BO_Shr;
case OO_LessLessEqual: return BO_ShlAssign;
case OO_GreaterGreaterEqual: return BO_ShrAssign;
case OO_EqualEqual: return BO_EQ;
case OO_ExclaimEqual: return BO_NE;
case OO_LessEqual: return BO_LE;
case OO_GreaterEqual: return BO_GE;
case OO_AmpAmp: return BO_LAnd;
case OO_PipePipe: return BO_LOr;
case OO_Comma: return BO_Comma;
case OO_ArrowStar: return BO_PtrMemI;
}
}
OverloadedOperatorKind BinaryOperator::getOverloadedOperator(Opcode Opc) {
static const OverloadedOperatorKind OverOps[] = {
/* .* Cannot be overloaded */OO_None, OO_ArrowStar,
OO_Star, OO_Slash, OO_Percent,
OO_Plus, OO_Minus,
OO_LessLess, OO_GreaterGreater,
OO_Less, OO_Greater, OO_LessEqual, OO_GreaterEqual,
OO_EqualEqual, OO_ExclaimEqual,
OO_Amp,
OO_Caret,
OO_Pipe,
OO_AmpAmp,
OO_PipePipe,
OO_Equal, OO_StarEqual,
OO_SlashEqual, OO_PercentEqual,
OO_PlusEqual, OO_MinusEqual,
OO_LessLessEqual, OO_GreaterGreaterEqual,
OO_AmpEqual, OO_CaretEqual,
OO_PipeEqual,
OO_Comma
};
return OverOps[Opc];
}
InitListExpr::InitListExpr(ASTContext &C, SourceLocation lbraceloc,
Expr **initExprs, unsigned numInits,
SourceLocation rbraceloc)
: Expr(InitListExprClass, QualType(), false, false),
InitExprs(C, numInits),
LBraceLoc(lbraceloc), RBraceLoc(rbraceloc), SyntacticForm(0),
UnionFieldInit(0), HadArrayRangeDesignator(false)
{
for (unsigned I = 0; I != numInits; ++I) {
if (initExprs[I]->isTypeDependent())
ExprBits.TypeDependent = true;
if (initExprs[I]->isValueDependent())
ExprBits.ValueDependent = true;
}
InitExprs.insert(C, InitExprs.end(), initExprs, initExprs+numInits);
}
void InitListExpr::reserveInits(ASTContext &C, unsigned NumInits) {
if (NumInits > InitExprs.size())
InitExprs.reserve(C, NumInits);
}
void InitListExpr::resizeInits(ASTContext &C, unsigned NumInits) {
InitExprs.resize(C, NumInits, 0);
}
Expr *InitListExpr::updateInit(ASTContext &C, unsigned Init, Expr *expr) {
if (Init >= InitExprs.size()) {
InitExprs.insert(C, InitExprs.end(), Init - InitExprs.size() + 1, 0);
InitExprs.back() = expr;
return 0;
}
Expr *Result = cast_or_null<Expr>(InitExprs[Init]);
InitExprs[Init] = expr;
return Result;
}
SourceRange InitListExpr::getSourceRange() const {
if (SyntacticForm)
return SyntacticForm->getSourceRange();
SourceLocation Beg = LBraceLoc, End = RBraceLoc;
if (Beg.isInvalid()) {
// Find the first non-null initializer.
for (InitExprsTy::const_iterator I = InitExprs.begin(),
E = InitExprs.end();
I != E; ++I) {
if (Stmt *S = *I) {
Beg = S->getLocStart();
break;
}
}
}
if (End.isInvalid()) {
// Find the first non-null initializer from the end.
for (InitExprsTy::const_reverse_iterator I = InitExprs.rbegin(),
E = InitExprs.rend();
I != E; ++I) {
if (Stmt *S = *I) {
End = S->getSourceRange().getEnd();
break;
}
}
}
return SourceRange(Beg, End);
}
/// getFunctionType - Return the underlying function type for this block.
///
const FunctionType *BlockExpr::getFunctionType() const {
return getType()->getAs<BlockPointerType>()->
getPointeeType()->getAs<FunctionType>();
}
SourceLocation BlockExpr::getCaretLocation() const {
return TheBlock->getCaretLocation();
}
const Stmt *BlockExpr::getBody() const {
return TheBlock->getBody();
}
Stmt *BlockExpr::getBody() {
return TheBlock->getBody();
}
//===----------------------------------------------------------------------===//
// Generic Expression Routines
//===----------------------------------------------------------------------===//
/// isUnusedResultAWarning - Return true if this immediate expression should
/// be warned about if the result is unused. If so, fill in Loc and Ranges
/// with location to warn on and the source range[s] to report with the
/// warning.
bool Expr::isUnusedResultAWarning(SourceLocation &Loc, SourceRange &R1,
SourceRange &R2, ASTContext &Ctx) const {
// Don't warn if the expr is type dependent. The type could end up
// instantiating to void.
if (isTypeDependent())
return false;
switch (getStmtClass()) {
default:
if (getType()->isVoidType())
return false;
Loc = getExprLoc();
R1 = getSourceRange();
return true;
case ParenExprClass:
return cast<ParenExpr>(this)->getSubExpr()->
isUnusedResultAWarning(Loc, R1, R2, Ctx);
case UnaryOperatorClass: {
const UnaryOperator *UO = cast<UnaryOperator>(this);
switch (UO->getOpcode()) {
default: break;
case UO_PostInc:
case UO_PostDec:
case UO_PreInc:
case UO_PreDec: // ++/--
return false; // Not a warning.
case UO_Deref:
// Dereferencing a volatile pointer is a side-effect.
if (Ctx.getCanonicalType(getType()).isVolatileQualified())
return false;
break;
case UO_Real:
case UO_Imag:
// accessing a piece of a volatile complex is a side-effect.
if (Ctx.getCanonicalType(UO->getSubExpr()->getType())
.isVolatileQualified())
return false;
break;
case UO_Extension:
return UO->getSubExpr()->isUnusedResultAWarning(Loc, R1, R2, Ctx);
}
Loc = UO->getOperatorLoc();
R1 = UO->getSubExpr()->getSourceRange();
return true;
}
case BinaryOperatorClass: {
const BinaryOperator *BO = cast<BinaryOperator>(this);
switch (BO->getOpcode()) {
default:
break;
// Consider the RHS of comma for side effects. LHS was checked by
// Sema::CheckCommaOperands.
case BO_Comma:
// ((foo = <blah>), 0) is an idiom for hiding the result (and
// lvalue-ness) of an assignment written in a macro.
if (IntegerLiteral *IE =
dyn_cast<IntegerLiteral>(BO->getRHS()->IgnoreParens()))
if (IE->getValue() == 0)
return false;
return BO->getRHS()->isUnusedResultAWarning(Loc, R1, R2, Ctx);
// Consider '||', '&&' to have side effects if the LHS or RHS does.
case BO_LAnd:
case BO_LOr:
if (!BO->getLHS()->isUnusedResultAWarning(Loc, R1, R2, Ctx) ||
!BO->getRHS()->isUnusedResultAWarning(Loc, R1, R2, Ctx))
return false;
break;
}
if (BO->isAssignmentOp())
return false;
Loc = BO->getOperatorLoc();
R1 = BO->getLHS()->getSourceRange();
R2 = BO->getRHS()->getSourceRange();
return true;
}
case CompoundAssignOperatorClass:
case VAArgExprClass:
return false;
case ConditionalOperatorClass: {
// The condition must be evaluated, but if either the LHS or RHS is a
// warning, warn about them.
const ConditionalOperator *Exp = cast<ConditionalOperator>(this);
if (Exp->getLHS() &&
Exp->getLHS()->isUnusedResultAWarning(Loc, R1, R2, Ctx))
return true;
return Exp->getRHS()->isUnusedResultAWarning(Loc, R1, R2, Ctx);
}
case MemberExprClass:
// If the base pointer or element is to a volatile pointer/field, accessing
// it is a side effect.
if (Ctx.getCanonicalType(getType()).isVolatileQualified())
return false;
Loc = cast<MemberExpr>(this)->getMemberLoc();
R1 = SourceRange(Loc, Loc);
R2 = cast<MemberExpr>(this)->getBase()->getSourceRange();
return true;
case ArraySubscriptExprClass:
// If the base pointer or element is to a volatile pointer/field, accessing
// it is a side effect.
if (Ctx.getCanonicalType(getType()).isVolatileQualified())
return false;
Loc = cast<ArraySubscriptExpr>(this)->getRBracketLoc();
R1 = cast<ArraySubscriptExpr>(this)->getLHS()->getSourceRange();
R2 = cast<ArraySubscriptExpr>(this)->getRHS()->getSourceRange();
return true;
case CallExprClass:
case CXXOperatorCallExprClass:
case CXXMemberCallExprClass: {
// If this is a direct call, get the callee.
const CallExpr *CE = cast<CallExpr>(this);
if (const Decl *FD = CE->getCalleeDecl()) {
// If the callee has attribute pure, const, or warn_unused_result, warn
// about it. void foo() { strlen("bar"); } should warn.
//
// Note: If new cases are added here, DiagnoseUnusedExprResult should be
// updated to match for QoI.
if (FD->getAttr<WarnUnusedResultAttr>() ||
FD->getAttr<PureAttr>() || FD->getAttr<ConstAttr>()) {
Loc = CE->getCallee()->getLocStart();
R1 = CE->getCallee()->getSourceRange();
if (unsigned NumArgs = CE->getNumArgs())
R2 = SourceRange(CE->getArg(0)->getLocStart(),
CE->getArg(NumArgs-1)->getLocEnd());
return true;
}
}
return false;
}
case CXXTemporaryObjectExprClass:
case CXXConstructExprClass:
return false;
case ObjCMessageExprClass: {
const ObjCMessageExpr *ME = cast<ObjCMessageExpr>(this);
const ObjCMethodDecl *MD = ME->getMethodDecl();
if (MD && MD->getAttr<WarnUnusedResultAttr>()) {
Loc = getExprLoc();
return true;
}
return false;
}
case ObjCImplicitSetterGetterRefExprClass: { // Dot syntax for message send.
#if 0
const ObjCImplicitSetterGetterRefExpr *Ref =
cast<ObjCImplicitSetterGetterRefExpr>(this);
// FIXME: We really want the location of the '.' here.
Loc = Ref->getLocation();
R1 = SourceRange(Ref->getLocation(), Ref->getLocation());
if (Ref->getBase())
R2 = Ref->getBase()->getSourceRange();
#else
Loc = getExprLoc();
R1 = getSourceRange();
#endif
return true;
}
case StmtExprClass: {
// Statement exprs don't logically have side effects themselves, but are
// sometimes used in macros in ways that give them a type that is unused.
// For example ({ blah; foo(); }) will end up with a type if foo has a type.
// however, if the result of the stmt expr is dead, we don't want to emit a
// warning.
const CompoundStmt *CS = cast<StmtExpr>(this)->getSubStmt();
if (!CS->body_empty()) {
if (const Expr *E = dyn_cast<Expr>(CS->body_back()))
return E->isUnusedResultAWarning(Loc, R1, R2, Ctx);
if (const LabelStmt *Label = dyn_cast<LabelStmt>(CS->body_back()))
if (const Expr *E = dyn_cast<Expr>(Label->getSubStmt()))
return E->isUnusedResultAWarning(Loc, R1, R2, Ctx);
}
if (getType()->isVoidType())
return false;
Loc = cast<StmtExpr>(this)->getLParenLoc();
R1 = getSourceRange();
return true;
}
case CStyleCastExprClass:
// If this is an explicit cast to void, allow it. People do this when they
// think they know what they're doing :).
if (getType()->isVoidType())
return false;
Loc = cast<CStyleCastExpr>(this)->getLParenLoc();
R1 = cast<CStyleCastExpr>(this)->getSubExpr()->getSourceRange();
return true;
case CXXFunctionalCastExprClass: {
if (getType()->isVoidType())
return false;
const CastExpr *CE = cast<CastExpr>(this);
// If this is a cast to void or a constructor conversion, check the operand.
// Otherwise, the result of the cast is unused.
if (CE->getCastKind() == CK_ToVoid ||
CE->getCastKind() == CK_ConstructorConversion)
return (cast<CastExpr>(this)->getSubExpr()
->isUnusedResultAWarning(Loc, R1, R2, Ctx));
Loc = cast<CXXFunctionalCastExpr>(this)->getTypeBeginLoc();
R1 = cast<CXXFunctionalCastExpr>(this)->getSubExpr()->getSourceRange();
return true;
}
case ImplicitCastExprClass:
// Check the operand, since implicit casts are inserted by Sema
return (cast<ImplicitCastExpr>(this)
->getSubExpr()->isUnusedResultAWarning(Loc, R1, R2, Ctx));
case CXXDefaultArgExprClass:
return (cast<CXXDefaultArgExpr>(this)
->getExpr()->isUnusedResultAWarning(Loc, R1, R2, Ctx));
case CXXNewExprClass:
// FIXME: In theory, there might be new expressions that don't have side
// effects (e.g. a placement new with an uninitialized POD).
case CXXDeleteExprClass:
return false;
case CXXBindTemporaryExprClass:
return (cast<CXXBindTemporaryExpr>(this)
->getSubExpr()->isUnusedResultAWarning(Loc, R1, R2, Ctx));
case CXXExprWithTemporariesClass:
return (cast<CXXExprWithTemporaries>(this)
->getSubExpr()->isUnusedResultAWarning(Loc, R1, R2, Ctx));
}
}
/// isOBJCGCCandidate - Check if an expression is objc gc'able.
/// returns true, if it is; false otherwise.
bool Expr::isOBJCGCCandidate(ASTContext &Ctx) const {
switch (getStmtClass()) {
default:
return false;
case ObjCIvarRefExprClass:
return true;
case Expr::UnaryOperatorClass:
return cast<UnaryOperator>(this)->getSubExpr()->isOBJCGCCandidate(Ctx);
case ParenExprClass:
return cast<ParenExpr>(this)->getSubExpr()->isOBJCGCCandidate(Ctx);
case ImplicitCastExprClass:
return cast<ImplicitCastExpr>(this)->getSubExpr()->isOBJCGCCandidate(Ctx);
case CStyleCastExprClass:
return cast<CStyleCastExpr>(this)->getSubExpr()->isOBJCGCCandidate(Ctx);
case DeclRefExprClass: {
const Decl *D = cast<DeclRefExpr>(this)->getDecl();
if (const VarDecl *VD = dyn_cast<VarDecl>(D)) {
if (VD->hasGlobalStorage())
return true;
QualType T = VD->getType();
// dereferencing to a pointer is always a gc'able candidate,
// unless it is __weak.
return T->isPointerType() &&
(Ctx.getObjCGCAttrKind(T) != Qualifiers::Weak);
}
return false;
}
case MemberExprClass: {
const MemberExpr *M = cast<MemberExpr>(this);
return M->getBase()->isOBJCGCCandidate(Ctx);
}
case ArraySubscriptExprClass:
return cast<ArraySubscriptExpr>(this)->getBase()->isOBJCGCCandidate(Ctx);
}
}
bool Expr::isBoundMemberFunction(ASTContext &Ctx) const {
if (isTypeDependent())
return false;
return isLvalue(Ctx) == Expr::LV_MemberFunction;
}
static Expr::CanThrowResult MergeCanThrow(Expr::CanThrowResult CT1,
Expr::CanThrowResult CT2) {
// CanThrowResult constants are ordered so that the maximum is the correct
// merge result.
return CT1 > CT2 ? CT1 : CT2;
}
static Expr::CanThrowResult CanSubExprsThrow(ASTContext &C, const Expr *CE) {
Expr *E = const_cast<Expr*>(CE);
Expr::CanThrowResult R = Expr::CT_Cannot;
for (Expr::child_iterator I = E->child_begin(), IE = E->child_end();
I != IE && R != Expr::CT_Can; ++I) {
R = MergeCanThrow(R, cast<Expr>(*I)->CanThrow(C));
}
return R;
}
static Expr::CanThrowResult CanCalleeThrow(const Decl *D,
bool NullThrows = true) {
if (!D)
return NullThrows ? Expr::CT_Can : Expr::CT_Cannot;
// See if we can get a function type from the decl somehow.
const ValueDecl *VD = dyn_cast<ValueDecl>(D);
if (!VD) // If we have no clue what we're calling, assume the worst.
return Expr::CT_Can;
// As an extension, we assume that __attribute__((nothrow)) functions don't
// throw.
if (isa<FunctionDecl>(D) && D->hasAttr<NoThrowAttr>())
return Expr::CT_Cannot;
QualType T = VD->getType();
const FunctionProtoType *FT;
if ((FT = T->getAs<FunctionProtoType>())) {
} else if (const PointerType *PT = T->getAs<PointerType>())
FT = PT->getPointeeType()->getAs<FunctionProtoType>();
else if (const ReferenceType *RT = T->getAs<ReferenceType>())
FT = RT->getPointeeType()->getAs<FunctionProtoType>();
else if (const MemberPointerType *MT = T->getAs<MemberPointerType>())
FT = MT->getPointeeType()->getAs<FunctionProtoType>();
else if (const BlockPointerType *BT = T->getAs<BlockPointerType>())
FT = BT->getPointeeType()->getAs<FunctionProtoType>();
if (!FT)
return Expr::CT_Can;
return FT->hasEmptyExceptionSpec() ? Expr::CT_Cannot : Expr::CT_Can;
}
static Expr::CanThrowResult CanDynamicCastThrow(const CXXDynamicCastExpr *DC) {
if (DC->isTypeDependent())
return Expr::CT_Dependent;
if (!DC->getTypeAsWritten()->isReferenceType())
return Expr::CT_Cannot;
return DC->getCastKind() == clang::CK_Dynamic? Expr::CT_Can : Expr::CT_Cannot;
}
static Expr::CanThrowResult CanTypeidThrow(ASTContext &C,
const CXXTypeidExpr *DC) {
if (DC->isTypeOperand())
return Expr::CT_Cannot;
Expr *Op = DC->getExprOperand();
if (Op->isTypeDependent())
return Expr::CT_Dependent;
const RecordType *RT = Op->getType()->getAs<RecordType>();
if (!RT)
return Expr::CT_Cannot;
if (!cast<CXXRecordDecl>(RT->getDecl())->isPolymorphic())
return Expr::CT_Cannot;
if (Op->Classify(C).isPRValue())
return Expr::CT_Cannot;
return Expr::CT_Can;
}
Expr::CanThrowResult Expr::CanThrow(ASTContext &C) const {
// C++ [expr.unary.noexcept]p3:
// [Can throw] if in a potentially-evaluated context the expression would
// contain:
switch (getStmtClass()) {
case CXXThrowExprClass:
// - a potentially evaluated throw-expression
return CT_Can;
case CXXDynamicCastExprClass: {
// - a potentially evaluated dynamic_cast expression dynamic_cast<T>(v),
// where T is a reference type, that requires a run-time check
CanThrowResult CT = CanDynamicCastThrow(cast<CXXDynamicCastExpr>(this));
if (CT == CT_Can)
return CT;
return MergeCanThrow(CT, CanSubExprsThrow(C, this));
}
case CXXTypeidExprClass:
// - a potentially evaluated typeid expression applied to a glvalue
// expression whose type is a polymorphic class type
return CanTypeidThrow(C, cast<CXXTypeidExpr>(this));
// - a potentially evaluated call to a function, member function, function
// pointer, or member function pointer that does not have a non-throwing
// exception-specification
case CallExprClass:
case CXXOperatorCallExprClass:
case CXXMemberCallExprClass: {
CanThrowResult CT = CanCalleeThrow(cast<CallExpr>(this)->getCalleeDecl());
if (CT == CT_Can)
return CT;
return MergeCanThrow(CT, CanSubExprsThrow(C, this));
}
case CXXConstructExprClass:
case CXXTemporaryObjectExprClass: {
CanThrowResult CT = CanCalleeThrow(
cast<CXXConstructExpr>(this)->getConstructor());
if (CT == CT_Can)
return CT;
return MergeCanThrow(CT, CanSubExprsThrow(C, this));
}
case CXXNewExprClass: {
CanThrowResult CT = MergeCanThrow(
CanCalleeThrow(cast<CXXNewExpr>(this)->getOperatorNew()),
CanCalleeThrow(cast<CXXNewExpr>(this)->getConstructor(),
/*NullThrows*/false));
if (CT == CT_Can)
return CT;
return MergeCanThrow(CT, CanSubExprsThrow(C, this));
}
case CXXDeleteExprClass: {
CanThrowResult CT = CanCalleeThrow(
cast<CXXDeleteExpr>(this)->getOperatorDelete());
if (CT == CT_Can)
return CT;
const Expr *Arg = cast<CXXDeleteExpr>(this)->getArgument();
// Unwrap exactly one implicit cast, which converts all pointers to void*.
if (const ImplicitCastExpr *Cast = dyn_cast<ImplicitCastExpr>(Arg))
Arg = Cast->getSubExpr();
if (const PointerType *PT = Arg->getType()->getAs<PointerType>()) {
if (const RecordType *RT = PT->getPointeeType()->getAs<RecordType>()) {
CanThrowResult CT2 = CanCalleeThrow(
cast<CXXRecordDecl>(RT->getDecl())->getDestructor());
if (CT2 == CT_Can)
return CT2;
CT = MergeCanThrow(CT, CT2);
}
}
return MergeCanThrow(CT, CanSubExprsThrow(C, this));
}
case CXXBindTemporaryExprClass: {
// The bound temporary has to be destroyed again, which might throw.
CanThrowResult CT = CanCalleeThrow(
cast<CXXBindTemporaryExpr>(this)->getTemporary()->getDestructor());
if (CT == CT_Can)
return CT;
return MergeCanThrow(CT, CanSubExprsThrow(C, this));
}
// ObjC message sends are like function calls, but never have exception
// specs.
case ObjCMessageExprClass:
case ObjCPropertyRefExprClass:
case ObjCImplicitSetterGetterRefExprClass:
return CT_Can;
// Many other things have subexpressions, so we have to test those.
// Some are simple:
case ParenExprClass:
case MemberExprClass:
case CXXReinterpretCastExprClass:
case CXXConstCastExprClass:
case ConditionalOperatorClass:
case CompoundLiteralExprClass:
case ExtVectorElementExprClass:
case InitListExprClass:
case DesignatedInitExprClass:
case ParenListExprClass:
case VAArgExprClass:
case CXXDefaultArgExprClass:
case CXXExprWithTemporariesClass:
case ObjCIvarRefExprClass:
case ObjCIsaExprClass:
case ShuffleVectorExprClass:
return CanSubExprsThrow(C, this);
// Some might be dependent for other reasons.
case UnaryOperatorClass:
case ArraySubscriptExprClass:
case ImplicitCastExprClass:
case CStyleCastExprClass:
case CXXStaticCastExprClass:
case CXXFunctionalCastExprClass:
case BinaryOperatorClass:
case CompoundAssignOperatorClass: {
CanThrowResult CT = isTypeDependent() ? CT_Dependent : CT_Cannot;
return MergeCanThrow(CT, CanSubExprsThrow(C, this));
}
// FIXME: We should handle StmtExpr, but that opens a MASSIVE can of worms.
case StmtExprClass:
return CT_Can;
case ChooseExprClass:
if (isTypeDependent() || isValueDependent())
return CT_Dependent;
return cast<ChooseExpr>(this)->getChosenSubExpr(C)->CanThrow(C);
// Some expressions are always dependent.
case DependentScopeDeclRefExprClass:
case CXXUnresolvedConstructExprClass:
case CXXDependentScopeMemberExprClass:
return CT_Dependent;
default:
// All other expressions don't have subexpressions, or else they are
// unevaluated.
return CT_Cannot;
}
}
Expr* Expr::IgnoreParens() {
Expr* E = this;
while (true) {
if (ParenExpr* P = dyn_cast<ParenExpr>(E)) {
E = P->getSubExpr();
continue;
}
if (UnaryOperator* P = dyn_cast<UnaryOperator>(E)) {
if (P->getOpcode() == UO_Extension) {
E = P->getSubExpr();
continue;
}
}
return E;
}
}
/// IgnoreParenCasts - Ignore parentheses and casts. Strip off any ParenExpr
/// or CastExprs or ImplicitCastExprs, returning their operand.
Expr *Expr::IgnoreParenCasts() {
Expr *E = this;
while (true) {
if (ParenExpr* P = dyn_cast<ParenExpr>(E)) {
E = P->getSubExpr();
continue;
}
if (CastExpr *P = dyn_cast<CastExpr>(E)) {
E = P->getSubExpr();
continue;
}
if (UnaryOperator* P = dyn_cast<UnaryOperator>(E)) {
if (P->getOpcode() == UO_Extension) {
E = P->getSubExpr();
continue;
}
}
return E;
}
}
Expr *Expr::IgnoreParenImpCasts() {
Expr *E = this;
while (true) {
if (ParenExpr *P = dyn_cast<ParenExpr>(E)) {
E = P->getSubExpr();
continue;
}
if (ImplicitCastExpr *P = dyn_cast<ImplicitCastExpr>(E)) {
E = P->getSubExpr();
continue;
}
if (UnaryOperator* P = dyn_cast<UnaryOperator>(E)) {
if (P->getOpcode() == UO_Extension) {
E = P->getSubExpr();
continue;
}
}
return E;
}
}
/// IgnoreParenNoopCasts - Ignore parentheses and casts that do not change the
/// value (including ptr->int casts of the same size). Strip off any
/// ParenExpr or CastExprs, returning their operand.
Expr *Expr::IgnoreParenNoopCasts(ASTContext &Ctx) {
Expr *E = this;
while (true) {
if (ParenExpr *P = dyn_cast<ParenExpr>(E)) {
E = P->getSubExpr();
continue;
}
if (CastExpr *P = dyn_cast<CastExpr>(E)) {
// We ignore integer <-> casts that are of the same width, ptr<->ptr and
// ptr<->int casts of the same width. We also ignore all identity casts.
Expr *SE = P->getSubExpr();
if (Ctx.hasSameUnqualifiedType(E->getType(), SE->getType())) {
E = SE;
continue;
}
if ((E->getType()->isPointerType() ||
E->getType()->isIntegralType(Ctx)) &&
(SE->getType()->isPointerType() ||
SE->getType()->isIntegralType(Ctx)) &&
Ctx.getTypeSize(E->getType()) == Ctx.getTypeSize(SE->getType())) {
E = SE;
continue;
}
}
if (UnaryOperator* P = dyn_cast<UnaryOperator>(E)) {
if (P->getOpcode() == UO_Extension) {
E = P->getSubExpr();
continue;
}
}
return E;
}
}
bool Expr::isDefaultArgument() const {
const Expr *E = this;
while (const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(E))
E = ICE->getSubExprAsWritten();
return isa<CXXDefaultArgExpr>(E);
}
2010-04-02 22:24:57 +04:00
/// \brief Skip over any no-op casts and any temporary-binding
/// expressions.
static const Expr *skipTemporaryBindingsAndNoOpCasts(const Expr *E) {
while (const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(E)) {
if (ICE->getCastKind() == CK_NoOp)
2010-04-02 22:24:57 +04:00
E = ICE->getSubExpr();
else
break;
}
while (const CXXBindTemporaryExpr *BE = dyn_cast<CXXBindTemporaryExpr>(E))
E = BE->getSubExpr();
while (const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(E)) {
if (ICE->getCastKind() == CK_NoOp)
2010-04-02 22:24:57 +04:00
E = ICE->getSubExpr();
else
break;
}
2010-04-02 22:24:57 +04:00
return E;
}
/// isTemporaryObject - Determines if this expression produces a
/// temporary of the given class type.
bool Expr::isTemporaryObject(ASTContext &C, const CXXRecordDecl *TempTy) const {
if (!C.hasSameUnqualifiedType(getType(), C.getTypeDeclType(TempTy)))
return false;
2010-04-02 22:24:57 +04:00
const Expr *E = skipTemporaryBindingsAndNoOpCasts(this);
2010-04-02 22:24:57 +04:00
// Temporaries are by definition pr-values of class type.
if (!E->Classify(C).isPRValue()) {
// In this context, property reference is a message call and is pr-value.
if (!isa<ObjCPropertyRefExpr>(E) &&
!isa<ObjCImplicitSetterGetterRefExpr>(E))
return false;
}
2010-04-02 22:24:57 +04:00
// Black-list a few cases which yield pr-values of class type that don't
// refer to temporaries of that type:
// - implicit derived-to-base conversions
if (isa<ImplicitCastExpr>(E)) {
switch (cast<ImplicitCastExpr>(E)->getCastKind()) {
case CK_DerivedToBase:
case CK_UncheckedDerivedToBase:
return false;
default:
break;
}
2010-04-02 22:24:57 +04:00
}
// - member expressions (all)
if (isa<MemberExpr>(E))
return false;
return true;
2010-04-02 22:24:57 +04:00
}
/// hasAnyTypeDependentArguments - Determines if any of the expressions
/// in Exprs is type-dependent.
bool Expr::hasAnyTypeDependentArguments(Expr** Exprs, unsigned NumExprs) {
for (unsigned I = 0; I < NumExprs; ++I)
if (Exprs[I]->isTypeDependent())
return true;
return false;
}
/// hasAnyValueDependentArguments - Determines if any of the expressions
/// in Exprs is value-dependent.
bool Expr::hasAnyValueDependentArguments(Expr** Exprs, unsigned NumExprs) {
for (unsigned I = 0; I < NumExprs; ++I)
if (Exprs[I]->isValueDependent())
return true;
return false;
}
bool Expr::isConstantInitializer(ASTContext &Ctx, bool IsForRef) const {
// This function is attempting whether an expression is an initializer
// which can be evaluated at compile-time. isEvaluatable handles most
// of the cases, but it can't deal with some initializer-specific
// expressions, and it can't deal with aggregates; we deal with those here,
// and fall back to isEvaluatable for the other cases.
// If we ever capture reference-binding directly in the AST, we can
// kill the second parameter.
if (IsForRef) {
EvalResult Result;
return EvaluateAsLValue(Result, Ctx) && !Result.HasSideEffects;
}
switch (getStmtClass()) {
default: break;
case StringLiteralClass:
This patch includes a conceptually simple, but very intrusive/pervasive change. The idea is to segregate Objective-C "object" pointers from general C pointers (utilizing the recently added ObjCObjectPointerType). The fun starts in Sema::GetTypeForDeclarator(), where "SomeInterface *" is now represented by a single AST node (rather than a PointerType whose Pointee is an ObjCInterfaceType). Since a significant amount of code assumed ObjC object pointers where based on C pointers/structs, this patch is very tedious. It should also explain why it is hard to accomplish this in smaller, self-contained patches. This patch does most of the "heavy lifting" related to moving from PointerType->ObjCObjectPointerType. It doesn't include all potential "cleanups". The good news is additional cleanups can be done later (some are noted in the code). This patch is so large that I didn't want to include any changes that are purely aesthetic. By making the ObjC types truly built-in, they are much easier to work with (and require fewer "hacks"). For example, there is no need for ASTContext::isObjCIdStructType() or ASTContext::isObjCClassStructType()! We believe this change (and the follow-up cleanups) will pay dividends over time. Given the amount of code change, I do expect some fallout from this change (though it does pass all of the clang tests). If you notice any problems, please let us know asap! Thanks. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@75314 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-11 03:34:53 +04:00
case ObjCStringLiteralClass:
case ObjCEncodeExprClass:
return true;
case CXXTemporaryObjectExprClass:
case CXXConstructExprClass: {
const CXXConstructExpr *CE = cast<CXXConstructExpr>(this);
// Only if it's
// 1) an application of the trivial default constructor or
if (!CE->getConstructor()->isTrivial()) return false;
if (!CE->getNumArgs()) return true;
// 2) an elidable trivial copy construction of an operand which is
// itself a constant initializer. Note that we consider the
// operand on its own, *not* as a reference binding.
return CE->isElidable() &&
CE->getArg(0)->isConstantInitializer(Ctx, false);
}
case CompoundLiteralExprClass: {
// This handles gcc's extension that allows global initializers like
// "struct x {int x;} x = (struct x) {};".
// FIXME: This accepts other cases it shouldn't!
const Expr *Exp = cast<CompoundLiteralExpr>(this)->getInitializer();
return Exp->isConstantInitializer(Ctx, false);
}
case InitListExprClass: {
// FIXME: This doesn't deal with fields with reference types correctly.
// FIXME: This incorrectly allows pointers cast to integers to be assigned
// to bitfields.
const InitListExpr *Exp = cast<InitListExpr>(this);
unsigned numInits = Exp->getNumInits();
for (unsigned i = 0; i < numInits; i++) {
if (!Exp->getInit(i)->isConstantInitializer(Ctx, false))
return false;
}
return true;
}
case ImplicitValueInitExprClass:
return true;
case ParenExprClass:
return cast<ParenExpr>(this)->getSubExpr()
->isConstantInitializer(Ctx, IsForRef);
case ChooseExprClass:
return cast<ChooseExpr>(this)->getChosenSubExpr(Ctx)
->isConstantInitializer(Ctx, IsForRef);
case UnaryOperatorClass: {
const UnaryOperator* Exp = cast<UnaryOperator>(this);
if (Exp->getOpcode() == UO_Extension)
return Exp->getSubExpr()->isConstantInitializer(Ctx, false);
break;
}
case BinaryOperatorClass: {
// Special case &&foo - &&bar. It would be nice to generalize this somehow
// but this handles the common case.
const BinaryOperator *Exp = cast<BinaryOperator>(this);
if (Exp->getOpcode() == BO_Sub &&
isa<AddrLabelExpr>(Exp->getLHS()->IgnoreParenNoopCasts(Ctx)) &&
isa<AddrLabelExpr>(Exp->getRHS()->IgnoreParenNoopCasts(Ctx)))
return true;
break;
}
case CXXFunctionalCastExprClass:
case CXXStaticCastExprClass:
case ImplicitCastExprClass:
case CStyleCastExprClass:
// Handle casts with a destination that's a struct or union; this
// deals with both the gcc no-op struct cast extension and the
// cast-to-union extension.
if (getType()->isRecordType())
return cast<CastExpr>(this)->getSubExpr()
->isConstantInitializer(Ctx, false);
// Integer->integer casts can be handled here, which is important for
// things like (int)(&&x-&&y). Scary but true.
if (getType()->isIntegerType() &&
cast<CastExpr>(this)->getSubExpr()->getType()->isIntegerType())
return cast<CastExpr>(this)->getSubExpr()
->isConstantInitializer(Ctx, false);
break;
}
return isEvaluatable(Ctx);
}
/// isNullPointerConstant - C99 6.3.2.3p3 - Return true if this is either an
/// integer constant expression with the value zero, or if this is one that is
/// cast to void*.
bool Expr::isNullPointerConstant(ASTContext &Ctx,
NullPointerConstantValueDependence NPC) const {
if (isValueDependent()) {
switch (NPC) {
case NPC_NeverValueDependent:
assert(false && "Unexpected value dependent expression!");
// If the unthinkable happens, fall through to the safest alternative.
case NPC_ValueDependentIsNull:
return isTypeDependent() || getType()->isIntegralType(Ctx);
case NPC_ValueDependentIsNotNull:
return false;
}
}
// Strip off a cast to void*, if it exists. Except in C++.
if (const ExplicitCastExpr *CE = dyn_cast<ExplicitCastExpr>(this)) {
if (!Ctx.getLangOptions().CPlusPlus) {
// Check that it is a cast to void*.
if (const PointerType *PT = CE->getType()->getAs<PointerType>()) {
QualType Pointee = PT->getPointeeType();
if (!Pointee.hasQualifiers() &&
Pointee->isVoidType() && // to void*
CE->getSubExpr()->getType()->isIntegerType()) // from int.
return CE->getSubExpr()->isNullPointerConstant(Ctx, NPC);
}
}
} else if (const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(this)) {
// Ignore the ImplicitCastExpr type entirely.
return ICE->getSubExpr()->isNullPointerConstant(Ctx, NPC);
} else if (const ParenExpr *PE = dyn_cast<ParenExpr>(this)) {
// Accept ((void*)0) as a null pointer constant, as many other
// implementations do.
return PE->getSubExpr()->isNullPointerConstant(Ctx, NPC);
} else if (const CXXDefaultArgExpr *DefaultArg
= dyn_cast<CXXDefaultArgExpr>(this)) {
// See through default argument expressions
return DefaultArg->getExpr()->isNullPointerConstant(Ctx, NPC);
} else if (isa<GNUNullExpr>(this)) {
// The GNU __null extension is always a null pointer constant.
return true;
}
// C++0x nullptr_t is always a null pointer constant.
if (getType()->isNullPtrType())
return true;
if (const RecordType *UT = getType()->getAsUnionType())
if (UT && UT->getDecl()->hasAttr<TransparentUnionAttr>())
if (const CompoundLiteralExpr *CLE = dyn_cast<CompoundLiteralExpr>(this)){
const Expr *InitExpr = CLE->getInitializer();
if (const InitListExpr *ILE = dyn_cast<InitListExpr>(InitExpr))
return ILE->getInit(0)->isNullPointerConstant(Ctx, NPC);
}
// This expression must be an integer type.
if (!getType()->isIntegerType() ||
(Ctx.getLangOptions().CPlusPlus && getType()->isEnumeralType()))
return false;
// If we have an integer constant expression, we need to *evaluate* it and
// test for the value 0.
llvm::APSInt Result;
return isIntegerConstantExpr(Result, Ctx) && Result == 0;
}
FieldDecl *Expr::getBitField() {
Expr *E = this->IgnoreParens();
while (ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(E)) {
if (ICE->getValueKind() != VK_RValue &&
ICE->getCastKind() == CK_NoOp)
E = ICE->getSubExpr()->IgnoreParens();
else
break;
}
if (MemberExpr *MemRef = dyn_cast<MemberExpr>(E))
if (FieldDecl *Field = dyn_cast<FieldDecl>(MemRef->getMemberDecl()))
if (Field->isBitField())
return Field;
if (DeclRefExpr *DeclRef = dyn_cast<DeclRefExpr>(E))
if (FieldDecl *Field = dyn_cast<FieldDecl>(DeclRef->getDecl()))
if (Field->isBitField())
return Field;
if (BinaryOperator *BinOp = dyn_cast<BinaryOperator>(E))
if (BinOp->isAssignmentOp() && BinOp->getLHS())
return BinOp->getLHS()->getBitField();
return 0;
}
bool Expr::refersToVectorElement() const {
const Expr *E = this->IgnoreParens();
while (const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(E)) {
if (ICE->getValueKind() != VK_RValue &&
ICE->getCastKind() == CK_NoOp)
E = ICE->getSubExpr()->IgnoreParens();
else
break;
}
if (const ArraySubscriptExpr *ASE = dyn_cast<ArraySubscriptExpr>(E))
return ASE->getBase()->getType()->isVectorType();
if (isa<ExtVectorElementExpr>(E))
return true;
return false;
}
/// isArrow - Return true if the base expression is a pointer to vector,
/// return false if the base expression is a vector.
bool ExtVectorElementExpr::isArrow() const {
return getBase()->getType()->isPointerType();
}
unsigned ExtVectorElementExpr::getNumElements() const {
if (const VectorType *VT = getType()->getAs<VectorType>())
return VT->getNumElements();
return 1;
}
/// containsDuplicateElements - Return true if any element access is repeated.
bool ExtVectorElementExpr::containsDuplicateElements() const {
// FIXME: Refactor this code to an accessor on the AST node which returns the
// "type" of component access, and share with code below and in Sema.
llvm::StringRef Comp = Accessor->getName();
// Halving swizzles do not contain duplicate elements.
if (Comp == "hi" || Comp == "lo" || Comp == "even" || Comp == "odd")
return false;
// Advance past s-char prefix on hex swizzles.
if (Comp[0] == 's' || Comp[0] == 'S')
Comp = Comp.substr(1);
for (unsigned i = 0, e = Comp.size(); i != e; ++i)
if (Comp.substr(i + 1).find(Comp[i]) != llvm::StringRef::npos)
return true;
return false;
}
/// getEncodedElementAccess - We encode the fields as a llvm ConstantArray.
void ExtVectorElementExpr::getEncodedElementAccess(
llvm::SmallVectorImpl<unsigned> &Elts) const {
llvm::StringRef Comp = Accessor->getName();
if (Comp[0] == 's' || Comp[0] == 'S')
Comp = Comp.substr(1);
bool isHi = Comp == "hi";
bool isLo = Comp == "lo";
bool isEven = Comp == "even";
bool isOdd = Comp == "odd";
for (unsigned i = 0, e = getNumElements(); i != e; ++i) {
uint64_t Index;
if (isHi)
Index = e + i;
else if (isLo)
Index = i;
else if (isEven)
Index = 2 * i;
else if (isOdd)
Index = 2 * i + 1;
else
Index = ExtVectorType::getAccessorIdx(Comp[i]);
Elts.push_back(Index);
}
}
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
ObjCMessageExpr::ObjCMessageExpr(QualType T,
SourceLocation LBracLoc,
SourceLocation SuperLoc,
bool IsInstanceSuper,
QualType SuperType,
Selector Sel,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
ObjCMethodDecl *Method,
Expr **Args, unsigned NumArgs,
SourceLocation RBracLoc)
: Expr(ObjCMessageExprClass, T, /*TypeDependent=*/false,
/*ValueDependent=*/false),
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
NumArgs(NumArgs), Kind(IsInstanceSuper? SuperInstance : SuperClass),
HasMethod(Method != 0), SuperLoc(SuperLoc),
SelectorOrMethod(reinterpret_cast<uintptr_t>(Method? Method
: Sel.getAsOpaquePtr())),
LBracLoc(LBracLoc), RBracLoc(RBracLoc)
{
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
setReceiverPointer(SuperType.getAsOpaquePtr());
if (NumArgs)
memcpy(getArgs(), Args, NumArgs * sizeof(Expr *));
}
ObjCMessageExpr::ObjCMessageExpr(QualType T,
SourceLocation LBracLoc,
TypeSourceInfo *Receiver,
Selector Sel,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
ObjCMethodDecl *Method,
Expr **Args, unsigned NumArgs,
SourceLocation RBracLoc)
: Expr(ObjCMessageExprClass, T, T->isDependentType(),
(T->isDependentType() ||
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
hasAnyValueDependentArguments(Args, NumArgs))),
NumArgs(NumArgs), Kind(Class), HasMethod(Method != 0),
SelectorOrMethod(reinterpret_cast<uintptr_t>(Method? Method
: Sel.getAsOpaquePtr())),
LBracLoc(LBracLoc), RBracLoc(RBracLoc)
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
{
setReceiverPointer(Receiver);
if (NumArgs)
memcpy(getArgs(), Args, NumArgs * sizeof(Expr *));
}
ObjCMessageExpr::ObjCMessageExpr(QualType T,
SourceLocation LBracLoc,
Expr *Receiver,
Selector Sel,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
ObjCMethodDecl *Method,
Expr **Args, unsigned NumArgs,
SourceLocation RBracLoc)
: Expr(ObjCMessageExprClass, T, Receiver->isTypeDependent(),
(Receiver->isTypeDependent() ||
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
hasAnyValueDependentArguments(Args, NumArgs))),
NumArgs(NumArgs), Kind(Instance), HasMethod(Method != 0),
SelectorOrMethod(reinterpret_cast<uintptr_t>(Method? Method
: Sel.getAsOpaquePtr())),
LBracLoc(LBracLoc), RBracLoc(RBracLoc)
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
{
setReceiverPointer(Receiver);
if (NumArgs)
memcpy(getArgs(), Args, NumArgs * sizeof(Expr *));
}
ObjCMessageExpr *ObjCMessageExpr::Create(ASTContext &Context, QualType T,
SourceLocation LBracLoc,
SourceLocation SuperLoc,
bool IsInstanceSuper,
QualType SuperType,
Selector Sel,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
ObjCMethodDecl *Method,
Expr **Args, unsigned NumArgs,
SourceLocation RBracLoc) {
unsigned Size = sizeof(ObjCMessageExpr) + sizeof(void *) +
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
NumArgs * sizeof(Expr *);
void *Mem = Context.Allocate(Size, llvm::AlignOf<ObjCMessageExpr>::Alignment);
return new (Mem) ObjCMessageExpr(T, LBracLoc, SuperLoc, IsInstanceSuper,
SuperType, Sel, Method, Args, NumArgs,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
RBracLoc);
}
ObjCMessageExpr *ObjCMessageExpr::Create(ASTContext &Context, QualType T,
SourceLocation LBracLoc,
TypeSourceInfo *Receiver,
Selector Sel,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
ObjCMethodDecl *Method,
Expr **Args, unsigned NumArgs,
SourceLocation RBracLoc) {
unsigned Size = sizeof(ObjCMessageExpr) + sizeof(void *) +
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
NumArgs * sizeof(Expr *);
void *Mem = Context.Allocate(Size, llvm::AlignOf<ObjCMessageExpr>::Alignment);
return new (Mem) ObjCMessageExpr(T, LBracLoc, Receiver, Sel, Method, Args,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
NumArgs, RBracLoc);
}
ObjCMessageExpr *ObjCMessageExpr::Create(ASTContext &Context, QualType T,
SourceLocation LBracLoc,
Expr *Receiver,
Selector Sel,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
ObjCMethodDecl *Method,
Expr **Args, unsigned NumArgs,
SourceLocation RBracLoc) {
unsigned Size = sizeof(ObjCMessageExpr) + sizeof(void *) +
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
NumArgs * sizeof(Expr *);
void *Mem = Context.Allocate(Size, llvm::AlignOf<ObjCMessageExpr>::Alignment);
return new (Mem) ObjCMessageExpr(T, LBracLoc, Receiver, Sel, Method, Args,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
NumArgs, RBracLoc);
}
ObjCMessageExpr *ObjCMessageExpr::CreateEmpty(ASTContext &Context,
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
unsigned NumArgs) {
unsigned Size = sizeof(ObjCMessageExpr) + sizeof(void *) +
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
NumArgs * sizeof(Expr *);
void *Mem = Context.Allocate(Size, llvm::AlignOf<ObjCMessageExpr>::Alignment);
return new (Mem) ObjCMessageExpr(EmptyShell(), NumArgs);
}
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
Selector ObjCMessageExpr::getSelector() const {
if (HasMethod)
return reinterpret_cast<const ObjCMethodDecl *>(SelectorOrMethod)
->getSelector();
return Selector(SelectorOrMethod);
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
}
ObjCInterfaceDecl *ObjCMessageExpr::getReceiverInterface() const {
switch (getReceiverKind()) {
case Instance:
if (const ObjCObjectPointerType *Ptr
= getInstanceReceiver()->getType()->getAs<ObjCObjectPointerType>())
return Ptr->getInterfaceDecl();
break;
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
case Class:
if (const ObjCObjectType *Ty
= getClassReceiver()->getAs<ObjCObjectType>())
return Ty->getInterface();
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
break;
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
case SuperInstance:
if (const ObjCObjectPointerType *Ptr
= getSuperType()->getAs<ObjCObjectPointerType>())
return Ptr->getInterfaceDecl();
break;
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
case SuperClass:
if (const ObjCObjectPointerType *Iface
= getSuperType()->getAs<ObjCObjectPointerType>())
return Iface->getInterfaceDecl();
break;
}
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
return 0;
}
bool ChooseExpr::isConditionTrue(ASTContext &C) const {
return getCond()->EvaluateAsInt(C) != 0;
}
void ShuffleVectorExpr::setExprs(ASTContext &C, Expr ** Exprs,
unsigned NumExprs) {
if (SubExprs) C.Deallocate(SubExprs);
SubExprs = new (C) Stmt* [NumExprs];
this->NumExprs = NumExprs;
memcpy(SubExprs, Exprs, sizeof(Expr *) * NumExprs);
}
//===----------------------------------------------------------------------===//
// DesignatedInitExpr
//===----------------------------------------------------------------------===//
IdentifierInfo *DesignatedInitExpr::Designator::getFieldName() {
assert(Kind == FieldDesignator && "Only valid on a field designator");
if (Field.NameOrField & 0x01)
return reinterpret_cast<IdentifierInfo *>(Field.NameOrField&~0x01);
else
return getField()->getIdentifier();
}
DesignatedInitExpr::DesignatedInitExpr(ASTContext &C, QualType Ty,
unsigned NumDesignators,
const Designator *Designators,
SourceLocation EqualOrColonLoc,
bool GNUSyntax,
Expr **IndexExprs,
unsigned NumIndexExprs,
Expr *Init)
: Expr(DesignatedInitExprClass, Ty,
Init->isTypeDependent(), Init->isValueDependent()),
EqualOrColonLoc(EqualOrColonLoc), GNUSyntax(GNUSyntax),
NumDesignators(NumDesignators), NumSubExprs(NumIndexExprs + 1) {
this->Designators = new (C) Designator[NumDesignators];
// Record the initializer itself.
child_iterator Child = child_begin();
*Child++ = Init;
// Copy the designators and their subexpressions, computing
// value-dependence along the way.
unsigned IndexIdx = 0;
for (unsigned I = 0; I != NumDesignators; ++I) {
this->Designators[I] = Designators[I];
if (this->Designators[I].isArrayDesignator()) {
// Compute type- and value-dependence.
Expr *Index = IndexExprs[IndexIdx];
ExprBits.ValueDependent = ExprBits.ValueDependent ||
Index->isTypeDependent() || Index->isValueDependent();
// Copy the index expressions into permanent storage.
*Child++ = IndexExprs[IndexIdx++];
} else if (this->Designators[I].isArrayRangeDesignator()) {
// Compute type- and value-dependence.
Expr *Start = IndexExprs[IndexIdx];
Expr *End = IndexExprs[IndexIdx + 1];
ExprBits.ValueDependent = ExprBits.ValueDependent ||
Start->isTypeDependent() || Start->isValueDependent() ||
End->isTypeDependent() || End->isValueDependent();
// Copy the start/end expressions into permanent storage.
*Child++ = IndexExprs[IndexIdx++];
*Child++ = IndexExprs[IndexIdx++];
}
}
assert(IndexIdx == NumIndexExprs && "Wrong number of index expressions");
}
DesignatedInitExpr *
DesignatedInitExpr::Create(ASTContext &C, Designator *Designators,
unsigned NumDesignators,
Expr **IndexExprs, unsigned NumIndexExprs,
SourceLocation ColonOrEqualLoc,
bool UsesColonSyntax, Expr *Init) {
void *Mem = C.Allocate(sizeof(DesignatedInitExpr) +
sizeof(Stmt *) * (NumIndexExprs + 1), 8);
return new (Mem) DesignatedInitExpr(C, C.VoidTy, NumDesignators, Designators,
ColonOrEqualLoc, UsesColonSyntax,
IndexExprs, NumIndexExprs, Init);
}
DesignatedInitExpr *DesignatedInitExpr::CreateEmpty(ASTContext &C,
unsigned NumIndexExprs) {
void *Mem = C.Allocate(sizeof(DesignatedInitExpr) +
sizeof(Stmt *) * (NumIndexExprs + 1), 8);
return new (Mem) DesignatedInitExpr(NumIndexExprs + 1);
}
void DesignatedInitExpr::setDesignators(ASTContext &C,
const Designator *Desigs,
unsigned NumDesigs) {
Designators = new (C) Designator[NumDesigs];
NumDesignators = NumDesigs;
for (unsigned I = 0; I != NumDesigs; ++I)
Designators[I] = Desigs[I];
}
SourceRange DesignatedInitExpr::getSourceRange() const {
SourceLocation StartLoc;
Designator &First =
*const_cast<DesignatedInitExpr*>(this)->designators_begin();
if (First.isFieldDesignator()) {
if (GNUSyntax)
StartLoc = SourceLocation::getFromRawEncoding(First.Field.FieldLoc);
else
StartLoc = SourceLocation::getFromRawEncoding(First.Field.DotLoc);
} else
StartLoc =
SourceLocation::getFromRawEncoding(First.ArrayOrRange.LBracketLoc);
return SourceRange(StartLoc, getInit()->getSourceRange().getEnd());
}
Expr *DesignatedInitExpr::getArrayIndex(const Designator& D) {
assert(D.Kind == Designator::ArrayDesignator && "Requires array designator");
char* Ptr = static_cast<char*>(static_cast<void *>(this));
Ptr += sizeof(DesignatedInitExpr);
Stmt **SubExprs = reinterpret_cast<Stmt**>(reinterpret_cast<void**>(Ptr));
return cast<Expr>(*(SubExprs + D.ArrayOrRange.Index + 1));
}
Expr *DesignatedInitExpr::getArrayRangeStart(const Designator& D) {
assert(D.Kind == Designator::ArrayRangeDesignator &&
"Requires array range designator");
char* Ptr = static_cast<char*>(static_cast<void *>(this));
Ptr += sizeof(DesignatedInitExpr);
Stmt **SubExprs = reinterpret_cast<Stmt**>(reinterpret_cast<void**>(Ptr));
return cast<Expr>(*(SubExprs + D.ArrayOrRange.Index + 1));
}
Expr *DesignatedInitExpr::getArrayRangeEnd(const Designator& D) {
assert(D.Kind == Designator::ArrayRangeDesignator &&
"Requires array range designator");
char* Ptr = static_cast<char*>(static_cast<void *>(this));
Ptr += sizeof(DesignatedInitExpr);
Stmt **SubExprs = reinterpret_cast<Stmt**>(reinterpret_cast<void**>(Ptr));
return cast<Expr>(*(SubExprs + D.ArrayOrRange.Index + 2));
}
/// \brief Replaces the designator at index @p Idx with the series
/// of designators in [First, Last).
void DesignatedInitExpr::ExpandDesignator(ASTContext &C, unsigned Idx,
const Designator *First,
const Designator *Last) {
unsigned NumNewDesignators = Last - First;
if (NumNewDesignators == 0) {
std::copy_backward(Designators + Idx + 1,
Designators + NumDesignators,
Designators + Idx);
--NumNewDesignators;
return;
} else if (NumNewDesignators == 1) {
Designators[Idx] = *First;
return;
}
Designator *NewDesignators
= new (C) Designator[NumDesignators - 1 + NumNewDesignators];
std::copy(Designators, Designators + Idx, NewDesignators);
std::copy(First, Last, NewDesignators + Idx);
std::copy(Designators + Idx + 1, Designators + NumDesignators,
NewDesignators + Idx + NumNewDesignators);
Designators = NewDesignators;
NumDesignators = NumDesignators - 1 + NumNewDesignators;
}
ParenListExpr::ParenListExpr(ASTContext& C, SourceLocation lparenloc,
Expr **exprs, unsigned nexprs,
SourceLocation rparenloc)
: Expr(ParenListExprClass, QualType(),
hasAnyTypeDependentArguments(exprs, nexprs),
hasAnyValueDependentArguments(exprs, nexprs)),
NumExprs(nexprs), LParenLoc(lparenloc), RParenLoc(rparenloc) {
Exprs = new (C) Stmt*[nexprs];
for (unsigned i = 0; i != nexprs; ++i)
Exprs[i] = exprs[i];
}
//===----------------------------------------------------------------------===//
// ExprIterator.
//===----------------------------------------------------------------------===//
Expr* ExprIterator::operator[](size_t idx) { return cast<Expr>(I[idx]); }
Expr* ExprIterator::operator*() const { return cast<Expr>(*I); }
Expr* ExprIterator::operator->() const { return cast<Expr>(*I); }
const Expr* ConstExprIterator::operator[](size_t idx) const {
return cast<Expr>(I[idx]);
}
const Expr* ConstExprIterator::operator*() const { return cast<Expr>(*I); }
const Expr* ConstExprIterator::operator->() const { return cast<Expr>(*I); }
//===----------------------------------------------------------------------===//
// Child Iterators for iterating over subexpressions/substatements
//===----------------------------------------------------------------------===//
// DeclRefExpr
Stmt::child_iterator DeclRefExpr::child_begin() { return child_iterator(); }
Stmt::child_iterator DeclRefExpr::child_end() { return child_iterator(); }
// ObjCIvarRefExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator ObjCIvarRefExpr::child_begin() { return &Base; }
Stmt::child_iterator ObjCIvarRefExpr::child_end() { return &Base+1; }
// ObjCPropertyRefExpr
Stmt::child_iterator ObjCPropertyRefExpr::child_begin()
{
if (BaseExprOrSuperType.is<Stmt*>()) {
// Hack alert!
return reinterpret_cast<Stmt**> (&BaseExprOrSuperType);
}
return child_iterator();
}
Stmt::child_iterator ObjCPropertyRefExpr::child_end()
{ return BaseExprOrSuperType.is<Stmt*>() ?
reinterpret_cast<Stmt**> (&BaseExprOrSuperType)+1 :
child_iterator();
}
// ObjCImplicitSetterGetterRefExpr
Stmt::child_iterator ObjCImplicitSetterGetterRefExpr::child_begin() {
// If this is accessing a class member or super, skip that entry.
// Technically, 2nd condition is sufficient. But I want to be verbose
if (isSuperReceiver() || !Base)
return child_iterator();
return &Base;
}
Stmt::child_iterator ObjCImplicitSetterGetterRefExpr::child_end() {
if (isSuperReceiver() || !Base)
return child_iterator();
return &Base+1;
}
// ObjCIsaExpr
Stmt::child_iterator ObjCIsaExpr::child_begin() { return &Base; }
Stmt::child_iterator ObjCIsaExpr::child_end() { return &Base+1; }
// PredefinedExpr
Stmt::child_iterator PredefinedExpr::child_begin() { return child_iterator(); }
Stmt::child_iterator PredefinedExpr::child_end() { return child_iterator(); }
// IntegerLiteral
Stmt::child_iterator IntegerLiteral::child_begin() { return child_iterator(); }
Stmt::child_iterator IntegerLiteral::child_end() { return child_iterator(); }
// CharacterLiteral
Stmt::child_iterator CharacterLiteral::child_begin() { return child_iterator();}
Stmt::child_iterator CharacterLiteral::child_end() { return child_iterator(); }
// FloatingLiteral
Stmt::child_iterator FloatingLiteral::child_begin() { return child_iterator(); }
Stmt::child_iterator FloatingLiteral::child_end() { return child_iterator(); }
// ImaginaryLiteral
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator ImaginaryLiteral::child_begin() { return &Val; }
Stmt::child_iterator ImaginaryLiteral::child_end() { return &Val+1; }
// StringLiteral
Stmt::child_iterator StringLiteral::child_begin() { return child_iterator(); }
Stmt::child_iterator StringLiteral::child_end() { return child_iterator(); }
// ParenExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator ParenExpr::child_begin() { return &Val; }
Stmt::child_iterator ParenExpr::child_end() { return &Val+1; }
// UnaryOperator
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator UnaryOperator::child_begin() { return &Val; }
Stmt::child_iterator UnaryOperator::child_end() { return &Val+1; }
Completely reimplement __builtin_offsetof, based on a patch by Roberto Amadini. This change introduces a new expression node type, OffsetOfExpr, that describes __builtin_offsetof. Previously, __builtin_offsetof was implemented using a unary operator whose subexpression involved various synthesized array-subscript and member-reference expressions, which was ugly and made it very hard to instantiate as a template. OffsetOfExpr represents the AST more faithfully, with proper type source information and a more compact representation. OffsetOfExpr also has support for dependent __builtin_offsetof expressions; it can be value-dependent, but will never be type-dependent (like sizeof or alignof). This commit introduces template instantiation for __builtin_offsetof as well. There are two major caveats to this patch: 1) CodeGen cannot handle the case where __builtin_offsetof is not a constant expression, so it produces an error. So, to avoid regressing in C, we retain the old UnaryOperator-based __builtin_offsetof implementation in C while using the shiny new OffsetOfExpr implementation in C++. The old implementation can go away once we have proper CodeGen support for this case, which we expect won't cause much trouble in C++. 2) __builtin_offsetof doesn't work well with non-POD class types, particularly when the designated field is found within a base class. I will address this in a subsequent patch. Fixes PR5880 and a bunch of assertions when building Boost.Python tests. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@102542 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-29 02:16:22 +04:00
// OffsetOfExpr
Stmt::child_iterator OffsetOfExpr::child_begin() {
return reinterpret_cast<Stmt **> (reinterpret_cast<OffsetOfNode *> (this + 1)
+ NumComps);
}
Stmt::child_iterator OffsetOfExpr::child_end() {
return child_iterator(&*child_begin() + NumExprs);
}
// SizeOfAlignOfExpr
Stmt::child_iterator SizeOfAlignOfExpr::child_begin() {
// If this is of a type and the type is a VLA type (and not a typedef), the
// size expression of the VLA needs to be treated as an executable expression.
// Why isn't this weirdness documented better in StmtIterator?
if (isArgumentType()) {
if (VariableArrayType* T = dyn_cast<VariableArrayType>(
getArgumentType().getTypePtr()))
return child_iterator(T);
return child_iterator();
}
return child_iterator(&Argument.Ex);
}
Stmt::child_iterator SizeOfAlignOfExpr::child_end() {
if (isArgumentType())
return child_iterator();
return child_iterator(&Argument.Ex + 1);
}
// ArraySubscriptExpr
Stmt::child_iterator ArraySubscriptExpr::child_begin() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0];
}
Stmt::child_iterator ArraySubscriptExpr::child_end() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0]+END_EXPR;
}
// CallExpr
Stmt::child_iterator CallExpr::child_begin() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0];
}
Stmt::child_iterator CallExpr::child_end() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0]+NumArgs+ARGS_START;
}
// MemberExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator MemberExpr::child_begin() { return &Base; }
Stmt::child_iterator MemberExpr::child_end() { return &Base+1; }
// ExtVectorElementExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator ExtVectorElementExpr::child_begin() { return &Base; }
Stmt::child_iterator ExtVectorElementExpr::child_end() { return &Base+1; }
// CompoundLiteralExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator CompoundLiteralExpr::child_begin() { return &Init; }
Stmt::child_iterator CompoundLiteralExpr::child_end() { return &Init+1; }
// CastExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator CastExpr::child_begin() { return &Op; }
Stmt::child_iterator CastExpr::child_end() { return &Op+1; }
// BinaryOperator
Stmt::child_iterator BinaryOperator::child_begin() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0];
}
Stmt::child_iterator BinaryOperator::child_end() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0]+END_EXPR;
}
// ConditionalOperator
Stmt::child_iterator ConditionalOperator::child_begin() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0];
}
Stmt::child_iterator ConditionalOperator::child_end() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0]+END_EXPR;
}
// AddrLabelExpr
Stmt::child_iterator AddrLabelExpr::child_begin() { return child_iterator(); }
Stmt::child_iterator AddrLabelExpr::child_end() { return child_iterator(); }
// StmtExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator StmtExpr::child_begin() { return &SubStmt; }
Stmt::child_iterator StmtExpr::child_end() { return &SubStmt+1; }
// TypesCompatibleExpr
Stmt::child_iterator TypesCompatibleExpr::child_begin() {
return child_iterator();
}
Stmt::child_iterator TypesCompatibleExpr::child_end() {
return child_iterator();
}
// ChooseExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator ChooseExpr::child_begin() { return &SubExprs[0]; }
Stmt::child_iterator ChooseExpr::child_end() { return &SubExprs[0]+END_EXPR; }
// GNUNullExpr
Stmt::child_iterator GNUNullExpr::child_begin() { return child_iterator(); }
Stmt::child_iterator GNUNullExpr::child_end() { return child_iterator(); }
// ShuffleVectorExpr
Stmt::child_iterator ShuffleVectorExpr::child_begin() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0];
}
Stmt::child_iterator ShuffleVectorExpr::child_end() {
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
return &SubExprs[0]+NumExprs;
}
// VAArgExpr
This patch is motivated by numerous strict-aliasing warnings when compiling clang as a Release build. The big change is that all AST nodes (subclasses of Stmt) whose children are Expr* store their children as Stmt* or arrays of Stmt*. This is to remove strict-aliasing warnings when using StmtIterator. None of the interfaces of any of the classes have changed (except those with arg_iterators, see below), as the accessor methods introduce the needed casts (via cast<>). While this extra casting may seem cumbersome, it actually adds some important sanity checks throughout the codebase, as clients using StmtIterator can potentially overwrite children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts provide extra sanity checks that are operational in debug builds to catch invariant violations such as these. For classes that have arg_iterators (e.g., CallExpr), the definition of arg_iterator has been replaced. Instead of it being Expr**, it is an actual class (called ExprIterator) that wraps a Stmt**, and provides the necessary operators for iteration. The nice thing about this class is that it also uses cast<> to type-checking, which introduces extra sanity checks throughout the codebase that are useful for debugging. A few of the CodeGen functions that use arg_iterator (especially from OverloadExpr) have been modified to take begin and end iterators instead of a base Expr** and the number of arguments. This matches more with the abstraction of iteration. This still needs to be cleaned up a little bit, as clients expect that ExprIterator is a RandomAccessIterator (which we may or may not wish to allow for efficiency of representation). This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c, which was already broken) on both a Debug and Release build, but it should obviously be reviewed. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@52378 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-17 06:43:46 +04:00
Stmt::child_iterator VAArgExpr::child_begin() { return &Val; }
Stmt::child_iterator VAArgExpr::child_end() { return &Val+1; }
// InitListExpr
Stmt::child_iterator InitListExpr::child_begin() {
return InitExprs.size() ? &InitExprs[0] : 0;
}
Stmt::child_iterator InitListExpr::child_end() {
return InitExprs.size() ? &InitExprs[0] + InitExprs.size() : 0;
}
// DesignatedInitExpr
Stmt::child_iterator DesignatedInitExpr::child_begin() {
char* Ptr = static_cast<char*>(static_cast<void *>(this));
Ptr += sizeof(DesignatedInitExpr);
return reinterpret_cast<Stmt**>(reinterpret_cast<void**>(Ptr));
}
Stmt::child_iterator DesignatedInitExpr::child_end() {
return child_iterator(&*child_begin() + NumSubExprs);
}
// ImplicitValueInitExpr
Stmt::child_iterator ImplicitValueInitExpr::child_begin() {
return child_iterator();
}
Stmt::child_iterator ImplicitValueInitExpr::child_end() {
return child_iterator();
}
// ParenListExpr
Stmt::child_iterator ParenListExpr::child_begin() {
return &Exprs[0];
}
Stmt::child_iterator ParenListExpr::child_end() {
return &Exprs[0]+NumExprs;
}
// ObjCStringLiteral
Stmt::child_iterator ObjCStringLiteral::child_begin() {
return &String;
}
Stmt::child_iterator ObjCStringLiteral::child_end() {
return &String+1;
}
// ObjCEncodeExpr
Stmt::child_iterator ObjCEncodeExpr::child_begin() { return child_iterator(); }
Stmt::child_iterator ObjCEncodeExpr::child_end() { return child_iterator(); }
// ObjCSelectorExpr
Stmt::child_iterator ObjCSelectorExpr::child_begin() {
return child_iterator();
}
Stmt::child_iterator ObjCSelectorExpr::child_end() {
return child_iterator();
}
// ObjCProtocolExpr
Stmt::child_iterator ObjCProtocolExpr::child_begin() {
return child_iterator();
}
Stmt::child_iterator ObjCProtocolExpr::child_end() {
return child_iterator();
}
// ObjCMessageExpr
Stmt::child_iterator ObjCMessageExpr::child_begin() {
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
if (getReceiverKind() == Instance)
return reinterpret_cast<Stmt **>(this + 1);
return getArgs();
}
Stmt::child_iterator ObjCMessageExpr::child_end() {
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101972 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 04:45:42 +04:00
return getArgs() + getNumArgs();
}
// Blocks
Stmt::child_iterator BlockExpr::child_begin() { return child_iterator(); }
Stmt::child_iterator BlockExpr::child_end() { return child_iterator(); }
Stmt::child_iterator BlockDeclRefExpr::child_begin() { return child_iterator();}
Stmt::child_iterator BlockDeclRefExpr::child_end() { return child_iterator(); }