diff options
Diffstat (limited to 'clang/lib')
53 files changed, 1979 insertions, 341 deletions
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp index 6b6275f..16cf114 100644 --- a/clang/lib/AST/ASTContext.cpp +++ b/clang/lib/AST/ASTContext.cpp @@ -940,7 +940,6 @@ ASTContext::ASTContext(LangOptions &LOpts, SourceManager &SM, FunctionProtoTypes(this_(), FunctionProtoTypesLog2InitSize), DependentTypeOfExprTypes(this_()), DependentDecltypeTypes(this_()), DependentPackIndexingTypes(this_()), TemplateSpecializationTypes(this_()), - DependentTemplateSpecializationTypes(this_()), DependentBitIntTypes(this_()), SubstTemplateTemplateParmPacks(this_()), DeducedTemplates(this_()), ArrayParameterTypes(this_()), CanonTemplateTemplateParms(this_()), SourceMgr(SM), LangOpts(LOpts), @@ -5979,10 +5978,9 @@ QualType ASTContext::getDependentTemplateSpecializationType( llvm::FoldingSetNodeID ID; DependentTemplateSpecializationType::Profile(ID, *this, Keyword, Name, Args); - void *InsertPos = nullptr; - if (auto *T = DependentTemplateSpecializationTypes.FindNodeOrInsertPos( - ID, InsertPos)) - return QualType(T, 0); + if (auto const T_iter = DependentTemplateSpecializationTypes.find(ID); + T_iter != DependentTemplateSpecializationTypes.end()) + return QualType(T_iter->getSecond(), 0); NestedNameSpecifier *NNS = Name.getQualifier(); @@ -6001,11 +5999,6 @@ QualType ASTContext::getDependentTemplateSpecializationType( CanonKeyword, {CanonNNS, Name.getName(), /*HasTemplateKeyword=*/true}, CanonArgs, /*IsCanonical=*/true); - // Find the insert position again. - [[maybe_unused]] auto *Nothing = - DependentTemplateSpecializationTypes.FindNodeOrInsertPos(ID, - InsertPos); - assert(!Nothing && "canonical type broken"); } } else { assert(Keyword == getCanonicalElaboratedTypeKeyword(Keyword)); @@ -6021,8 +6014,13 @@ QualType ASTContext::getDependentTemplateSpecializationType( alignof(DependentTemplateSpecializationType)); auto *T = new (Mem) DependentTemplateSpecializationType(Keyword, Name, Args, Canon); +#ifndef NDEBUG + llvm::FoldingSetNodeID InsertedID; + T->Profile(InsertedID, *this); + assert(InsertedID == ID && "ID does not match"); +#endif Types.push_back(T); - DependentTemplateSpecializationTypes.InsertNode(T, InsertPos); + DependentTemplateSpecializationTypes.try_emplace(ID, T); return QualType(T, 0); } diff --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp index 0d12161..9808298 100644 --- a/clang/lib/AST/ExprConstant.cpp +++ b/clang/lib/AST/ExprConstant.cpp @@ -14636,7 +14636,9 @@ EvaluateComparisonBinaryOperator(EvalInfo &Info, const BinaryOperator *E, if (!LHSDesignator.Invalid && !RHSDesignator.Invalid && IsRelational) { bool WasArrayIndex; unsigned Mismatch = FindDesignatorMismatch( - getType(LHSValue.Base), LHSDesignator, RHSDesignator, WasArrayIndex); + LHSValue.Base.isNull() ? QualType() + : getType(LHSValue.Base).getNonReferenceType(), + LHSDesignator, RHSDesignator, WasArrayIndex); // At the point where the designators diverge, the comparison has a // specified value if: // - we are comparing array indices @@ -14680,7 +14682,7 @@ EvaluateComparisonBinaryOperator(EvalInfo &Info, const BinaryOperator *E, // compare pointers within the object in question; otherwise, the result // depends on where the object is located in memory. if (!LHSValue.Base.isNull() && IsRelational) { - QualType BaseTy = getType(LHSValue.Base); + QualType BaseTy = getType(LHSValue.Base).getNonReferenceType(); if (BaseTy->isIncompleteType()) return Error(E); CharUnits Size = Info.Ctx.getTypeSizeInChars(BaseTy); diff --git a/clang/lib/AST/OSLog.cpp b/clang/lib/AST/OSLog.cpp index b777d4d..91f8410 100644 --- a/clang/lib/AST/OSLog.cpp +++ b/clang/lib/AST/OSLog.cpp @@ -1,4 +1,16 @@ -// TODO: header template +//===--- OSLog.cpp - OS log format string analysis ------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +/// +/// \file +/// This file implements analysis functions for OS log format strings and +/// buffer layout computation for __builtin_os_log_format and related builtins. +/// +//===----------------------------------------------------------------------===// #include "clang/AST/OSLog.h" #include "clang/AST/Attr.h" @@ -137,8 +149,8 @@ public: for (auto &Data : ArgsData) { if (!Data.MaskType.empty()) { CharUnits Size = CharUnits::fromQuantity(8); - Layout.Items.emplace_back(OSLogBufferItem::MaskKind, nullptr, - Size, 0, Data.MaskType); + Layout.Items.emplace_back(OSLogBufferItem::MaskKind, nullptr, Size, 0, + Data.MaskType); } if (Data.FieldWidth) { diff --git a/clang/lib/Basic/Targets/WebAssembly.cpp b/clang/lib/Basic/Targets/WebAssembly.cpp index af25d25..e362350e 100644 --- a/clang/lib/Basic/Targets/WebAssembly.cpp +++ b/clang/lib/Basic/Targets/WebAssembly.cpp @@ -64,6 +64,7 @@ bool WebAssemblyTargetInfo::hasFeature(StringRef Feature) const { .Case("mutable-globals", HasMutableGlobals) .Case("nontrapping-fptoint", HasNontrappingFPToInt) .Case("reference-types", HasReferenceTypes) + .Case("gc", HasGC) .Case("relaxed-simd", SIMDLevel >= RelaxedSIMD) .Case("sign-ext", HasSignExt) .Case("simd128", SIMDLevel >= SIMD128) @@ -106,6 +107,8 @@ void WebAssemblyTargetInfo::getTargetDefines(const LangOptions &Opts, Builder.defineMacro("__wasm_nontrapping_fptoint__"); if (HasReferenceTypes) Builder.defineMacro("__wasm_reference_types__"); + if (HasGC) + Builder.defineMacro("__wasm_gc__"); if (SIMDLevel >= RelaxedSIMD) Builder.defineMacro("__wasm_relaxed_simd__"); if (HasSignExt) @@ -307,6 +310,14 @@ bool WebAssemblyTargetInfo::handleTargetFeatures( HasReferenceTypes = false; continue; } + if (Feature == "+gc") { + HasGC = true; + continue; + } + if (Feature == "-gc") { + HasGC = false; + continue; + } if (Feature == "+relaxed-simd") { SIMDLevel = std::max(SIMDLevel, RelaxedSIMD); continue; @@ -353,6 +364,11 @@ bool WebAssemblyTargetInfo::handleTargetFeatures( return false; } + // gc implies reference-types + if (HasGC) { + HasReferenceTypes = true; + } + // bulk-memory-opt is a subset of bulk-memory. if (HasBulkMemory) { HasBulkMemoryOpt = true; diff --git a/clang/lib/Basic/Targets/WebAssembly.h b/clang/lib/Basic/Targets/WebAssembly.h index 57b366c..c47c8cc 100644 --- a/clang/lib/Basic/Targets/WebAssembly.h +++ b/clang/lib/Basic/Targets/WebAssembly.h @@ -69,6 +69,7 @@ class LLVM_LIBRARY_VISIBILITY WebAssemblyTargetInfo : public TargetInfo { bool HasMutableGlobals = false; bool HasNontrappingFPToInt = false; bool HasReferenceTypes = false; + bool HasGC = false; bool HasSignExt = false; bool HasTailCall = false; bool HasWideArithmetic = false; diff --git a/clang/lib/CIR/CodeGen/CIRGenClass.cpp b/clang/lib/CIR/CodeGen/CIRGenClass.cpp index fbf53db..50cca0e 100644 --- a/clang/lib/CIR/CodeGen/CIRGenClass.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenClass.cpp @@ -12,6 +12,7 @@ #include "CIRGenCXXABI.h" #include "CIRGenFunction.h" +#include "CIRGenValue.h" #include "clang/AST/ExprCXX.h" #include "clang/AST/RecordLayout.h" @@ -311,6 +312,116 @@ void CIRGenFunction::emitInitializerForField(FieldDecl *field, LValue lhs, assert(!cir::MissingFeatures::requiresCleanups()); } +/// Emit a loop to call a particular constructor for each of several members +/// of an array. +/// +/// \param ctor the constructor to call for each element +/// \param arrayType the type of the array to initialize +/// \param arrayBegin an arrayType* +/// \param zeroInitialize true if each element should be +/// zero-initialized before it is constructed +void CIRGenFunction::emitCXXAggrConstructorCall( + const CXXConstructorDecl *ctor, const clang::ArrayType *arrayType, + Address arrayBegin, const CXXConstructExpr *e, bool newPointerIsChecked, + bool zeroInitialize) { + QualType elementType; + mlir::Value numElements = emitArrayLength(arrayType, elementType, arrayBegin); + emitCXXAggrConstructorCall(ctor, numElements, arrayBegin, e, + newPointerIsChecked, zeroInitialize); +} + +/// Emit a loop to call a particular constructor for each of several members +/// of an array. +/// +/// \param ctor the constructor to call for each element +/// \param numElements the number of elements in the array; +/// may be zero +/// \param arrayBase a T*, where T is the type constructed by ctor +/// \param zeroInitialize true if each element should be +/// zero-initialized before it is constructed +void CIRGenFunction::emitCXXAggrConstructorCall( + const CXXConstructorDecl *ctor, mlir::Value numElements, Address arrayBase, + const CXXConstructExpr *e, bool newPointerIsChecked, bool zeroInitialize) { + // It's legal for numElements to be zero. This can happen both + // dynamically, because x can be zero in 'new A[x]', and statically, + // because of GCC extensions that permit zero-length arrays. There + // are probably legitimate places where we could assume that this + // doesn't happen, but it's not clear that it's worth it. + + // Optimize for a constant count. + auto constantCount = dyn_cast<cir::ConstantOp>(numElements.getDefiningOp()); + if (constantCount) { + auto constIntAttr = mlir::dyn_cast<cir::IntAttr>(constantCount.getValue()); + // Just skip out if the constant count is zero. + if (constIntAttr && constIntAttr.getUInt() == 0) + return; + } else { + // Otherwise, emit the check. + cgm.errorNYI(e->getSourceRange(), "dynamic-length array expression"); + } + + auto arrayTy = mlir::cast<cir::ArrayType>(arrayBase.getElementType()); + mlir::Type elementType = arrayTy.getElementType(); + cir::PointerType ptrToElmType = builder.getPointerTo(elementType); + + // Tradional LLVM codegen emits a loop here. CIR lowers to a loop as part of + // LoweringPrepare. + + // The alignment of the base, adjusted by the size of a single element, + // provides a conservative estimate of the alignment of every element. + // (This assumes we never start tracking offsetted alignments.) + // + // Note that these are complete objects and so we don't need to + // use the non-virtual size or alignment. + QualType type = getContext().getTypeDeclType(ctor->getParent()); + CharUnits eltAlignment = arrayBase.getAlignment().alignmentOfArrayElement( + getContext().getTypeSizeInChars(type)); + + // Zero initialize the storage, if requested. + if (zeroInitialize) + emitNullInitialization(*currSrcLoc, arrayBase, type); + + // C++ [class.temporary]p4: + // There are two contexts in which temporaries are destroyed at a different + // point than the end of the full-expression. The first context is when a + // default constructor is called to initialize an element of an array. + // If the constructor has one or more default arguments, the destruction of + // every temporary created in a default argument expression is sequenced + // before the construction of the next array element, if any. + { + assert(!cir::MissingFeatures::runCleanupsScope()); + + // Evaluate the constructor and its arguments in a regular + // partial-destroy cleanup. + if (getLangOpts().Exceptions && + !ctor->getParent()->hasTrivialDestructor()) { + cgm.errorNYI(e->getSourceRange(), "partial array cleanups"); + } + + // Emit the constructor call that will execute for every array element. + mlir::Value arrayOp = + builder.createPtrBitcast(arrayBase.getPointer(), arrayTy); + builder.create<cir::ArrayCtor>( + *currSrcLoc, arrayOp, [&](mlir::OpBuilder &b, mlir::Location loc) { + mlir::BlockArgument arg = + b.getInsertionBlock()->addArgument(ptrToElmType, loc); + Address curAddr = Address(arg, elementType, eltAlignment); + assert(!cir::MissingFeatures::sanitizers()); + auto currAVS = AggValueSlot::forAddr( + curAddr, type.getQualifiers(), AggValueSlot::IsDestructed, + AggValueSlot::IsNotAliased, AggValueSlot::DoesNotOverlap, + AggValueSlot::IsNotZeroed); + emitCXXConstructorCall(ctor, Ctor_Complete, + /*ForVirtualBase=*/false, + /*Delegating=*/false, currAVS, e); + builder.create<cir::YieldOp>(loc); + }); + } + + if (constantCount.use_empty()) + constantCount.erase(); +} + void CIRGenFunction::emitDelegateCXXConstructorCall( const CXXConstructorDecl *ctor, CXXCtorType ctorType, const FunctionArgList &args, SourceLocation loc) { @@ -369,6 +480,19 @@ void CIRGenFunction::emitImplicitAssignmentOperatorBody(FunctionArgList &args) { s->getStmtClassName()); } +void CIRGenFunction::destroyCXXObject(CIRGenFunction &cgf, Address addr, + QualType type) { + const RecordType *rtype = type->castAs<RecordType>(); + const CXXRecordDecl *record = cast<CXXRecordDecl>(rtype->getDecl()); + const CXXDestructorDecl *dtor = record->getDestructor(); + // TODO(cir): Unlike traditional codegen, CIRGen should actually emit trivial + // dtors which shall be removed on later CIR passes. However, only remove this + // assertion after we have a test case to exercise this path. + assert(!dtor->isTrivial()); + cgf.emitCXXDestructorCall(dtor, Dtor_Complete, /*forVirtualBase*/ false, + /*delegating=*/false, addr, type); +} + void CIRGenFunction::emitDelegatingCXXConstructorCall( const CXXConstructorDecl *ctor, const FunctionArgList &args) { assert(ctor->isDelegatingConstructor()); diff --git a/clang/lib/CIR/CodeGen/CIRGenCleanup.cpp b/clang/lib/CIR/CodeGen/CIRGenCleanup.cpp new file mode 100644 index 0000000..be21ce9 --- /dev/null +++ b/clang/lib/CIR/CodeGen/CIRGenCleanup.cpp @@ -0,0 +1,69 @@ +//===--- CIRGenCleanup.cpp - Bookkeeping and code emission for cleanups ---===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file contains code dealing with the IR generation for cleanups +// and related information. +// +// A "cleanup" is a piece of code which needs to be executed whenever +// control transfers out of a particular scope. This can be +// conditionalized to occur only on exceptional control flow, only on +// normal control flow, or both. +// +//===----------------------------------------------------------------------===// + +#include "CIRGenFunction.h" + +#include "clang/CIR/MissingFeatures.h" + +using namespace clang; +using namespace clang::CIRGen; + +//===----------------------------------------------------------------------===// +// CIRGenFunction cleanup related +//===----------------------------------------------------------------------===// + +//===----------------------------------------------------------------------===// +// EHScopeStack +//===----------------------------------------------------------------------===// + +void EHScopeStack::Cleanup::anchor() {} + +static mlir::Block *getCurCleanupBlock(CIRGenFunction &cgf) { + mlir::OpBuilder::InsertionGuard guard(cgf.getBuilder()); + mlir::Block *cleanup = + cgf.curLexScope->getOrCreateCleanupBlock(cgf.getBuilder()); + return cleanup; +} + +/// Pops a cleanup block. If the block includes a normal cleanup, the +/// current insertion point is threaded through the cleanup, as are +/// any branch fixups on the cleanup. +void CIRGenFunction::popCleanupBlock() { + assert(!ehStack.cleanupStack.empty() && "cleanup stack is empty!"); + mlir::OpBuilder::InsertionGuard guard(builder); + std::unique_ptr<EHScopeStack::Cleanup> cleanup = + ehStack.cleanupStack.pop_back_val(); + + assert(!cir::MissingFeatures::ehCleanupFlags()); + mlir::Block *cleanupEntry = getCurCleanupBlock(*this); + builder.setInsertionPointToEnd(cleanupEntry); + cleanup->emit(*this); +} + +/// Pops cleanup blocks until the given savepoint is reached. +void CIRGenFunction::popCleanupBlocks(size_t oldCleanupStackDepth) { + assert(!cir::MissingFeatures::ehstackBranches()); + + assert(ehStack.getStackDepth() >= oldCleanupStackDepth); + + // Pop cleanup blocks until we reach the base stack depth for the + // current scope. + while (ehStack.getStackDepth() > oldCleanupStackDepth) { + popCleanupBlock(); + } +} diff --git a/clang/lib/CIR/CodeGen/CIRGenDecl.cpp b/clang/lib/CIR/CodeGen/CIRGenDecl.cpp index afbe92a..a28ac3c 100644 --- a/clang/lib/CIR/CodeGen/CIRGenDecl.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenDecl.cpp @@ -183,8 +183,8 @@ void CIRGenFunction::emitAutoVarCleanups( const VarDecl &d = *emission.Variable; // Check the type for a cleanup. - if (d.needsDestruction(getContext())) - cgm.errorNYI(d.getSourceRange(), "emitAutoVarCleanups: type cleanup"); + if (QualType::DestructionKind dtorKind = d.needsDestruction(getContext())) + emitAutoVarTypeCleanup(emission, dtorKind); assert(!cir::MissingFeatures::opAllocaPreciseLifetime()); @@ -648,3 +648,96 @@ void CIRGenFunction::emitNullabilityCheck(LValue lhs, mlir::Value rhs, assert(!cir::MissingFeatures::sanitizers()); } + +/// Immediately perform the destruction of the given object. +/// +/// \param addr - the address of the object; a type* +/// \param type - the type of the object; if an array type, all +/// objects are destroyed in reverse order +/// \param destroyer - the function to call to destroy individual +/// elements +void CIRGenFunction::emitDestroy(Address addr, QualType type, + Destroyer *destroyer) { + if (getContext().getAsArrayType(type)) + cgm.errorNYI("emitDestroy: array type"); + + return destroyer(*this, addr, type); +} + +CIRGenFunction::Destroyer * +CIRGenFunction::getDestroyer(QualType::DestructionKind kind) { + switch (kind) { + case QualType::DK_none: + llvm_unreachable("no destroyer for trivial dtor"); + case QualType::DK_cxx_destructor: + return destroyCXXObject; + case QualType::DK_objc_strong_lifetime: + case QualType::DK_objc_weak_lifetime: + case QualType::DK_nontrivial_c_struct: + cgm.errorNYI("getDestroyer: other destruction kind"); + return nullptr; + } + llvm_unreachable("Unknown DestructionKind"); +} + +namespace { +struct DestroyObject final : EHScopeStack::Cleanup { + DestroyObject(Address addr, QualType type, + CIRGenFunction::Destroyer *destroyer) + : addr(addr), type(type), destroyer(destroyer) {} + + Address addr; + QualType type; + CIRGenFunction::Destroyer *destroyer; + + void emit(CIRGenFunction &cgf) override { + cgf.emitDestroy(addr, type, destroyer); + } +}; +} // namespace + +/// Enter a destroy cleanup for the given local variable. +void CIRGenFunction::emitAutoVarTypeCleanup( + const CIRGenFunction::AutoVarEmission &emission, + QualType::DestructionKind dtorKind) { + assert(dtorKind != QualType::DK_none); + + // Note that for __block variables, we want to destroy the + // original stack object, not the possibly forwarded object. + Address addr = emission.getObjectAddress(*this); + + const VarDecl *var = emission.Variable; + QualType type = var->getType(); + + CleanupKind cleanupKind = NormalAndEHCleanup; + CIRGenFunction::Destroyer *destroyer = nullptr; + + switch (dtorKind) { + case QualType::DK_none: + llvm_unreachable("no cleanup for trivially-destructible variable"); + + case QualType::DK_cxx_destructor: + // If there's an NRVO flag on the emission, we need a different + // cleanup. + if (emission.NRVOFlag) { + cgm.errorNYI(var->getSourceRange(), "emitAutoVarTypeCleanup: NRVO"); + return; + } + // Otherwise, this is handled below. + break; + + case QualType::DK_objc_strong_lifetime: + case QualType::DK_objc_weak_lifetime: + case QualType::DK_nontrivial_c_struct: + cgm.errorNYI(var->getSourceRange(), + "emitAutoVarTypeCleanup: other dtor kind"); + return; + } + + // If we haven't chosen a more specific destroyer, use the default. + if (!destroyer) + destroyer = getDestroyer(dtorKind); + + assert(!cir::MissingFeatures::ehCleanupFlags()); + ehStack.pushCleanup<DestroyObject>(cleanupKind, addr, type, destroyer); +} diff --git a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp index 1f64801..7ff5f26 100644 --- a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp @@ -1657,37 +1657,38 @@ void CIRGenFunction::emitCXXConstructExpr(const CXXConstructExpr *e, return; } - if (getContext().getAsArrayType(e->getType())) { - cgm.errorNYI(e->getSourceRange(), "emitCXXConstructExpr: array type"); - return; - } + if (const ArrayType *arrayType = getContext().getAsArrayType(e->getType())) { + assert(!cir::MissingFeatures::sanitizers()); + emitCXXAggrConstructorCall(cd, arrayType, dest.getAddress(), e, false); + } else { - clang::CXXCtorType type = Ctor_Complete; - bool forVirtualBase = false; - bool delegating = false; - - switch (e->getConstructionKind()) { - case CXXConstructionKind::Complete: - type = Ctor_Complete; - break; - case CXXConstructionKind::Delegating: - // We should be emitting a constructor; GlobalDecl will assert this - type = curGD.getCtorType(); - delegating = true; - break; - case CXXConstructionKind::VirtualBase: - // This should just set 'forVirtualBase' to true and fall through, but - // virtual base class support is otherwise missing, so this needs to wait - // until it can be tested. - cgm.errorNYI(e->getSourceRange(), - "emitCXXConstructExpr: virtual base constructor"); - return; - case CXXConstructionKind::NonVirtualBase: - type = Ctor_Base; - break; - } + clang::CXXCtorType type = Ctor_Complete; + bool forVirtualBase = false; + bool delegating = false; - emitCXXConstructorCall(cd, type, forVirtualBase, delegating, dest, e); + switch (e->getConstructionKind()) { + case CXXConstructionKind::Complete: + type = Ctor_Complete; + break; + case CXXConstructionKind::Delegating: + // We should be emitting a constructor; GlobalDecl will assert this + type = curGD.getCtorType(); + delegating = true; + break; + case CXXConstructionKind::VirtualBase: + // This should just set 'forVirtualBase' to true and fall through, but + // virtual base class support is otherwise missing, so this needs to wait + // until it can be tested. + cgm.errorNYI(e->getSourceRange(), + "emitCXXConstructExpr: virtual base constructor"); + return; + case CXXConstructionKind::NonVirtualBase: + type = Ctor_Base; + break; + } + + emitCXXConstructorCall(cd, type, forVirtualBase, delegating, dest, e); + } } RValue CIRGenFunction::emitReferenceBindingToExpr(const Expr *e) { diff --git a/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp b/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp index 0d12c5c..51aab95 100644 --- a/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp @@ -357,10 +357,97 @@ void AggExprEmitter::visitCXXParenListOrInitListExpr( emitArrayInit(dest.getAddress(), arrayTy, e->getType(), e, args, arrayFiller); return; + } else if (e->getType()->isVariableArrayType()) { + cgf.cgm.errorNYI(e->getSourceRange(), + "visitCXXParenListOrInitListExpr variable array type"); + return; + } + + if (e->getType()->isArrayType()) { + cgf.cgm.errorNYI(e->getSourceRange(), + "visitCXXParenListOrInitListExpr array type"); + return; + } + + assert(e->getType()->isRecordType() && "Only support structs/unions here!"); + + // Do struct initialization; this code just sets each individual member + // to the approprate value. This makes bitfield support automatic; + // the disadvantage is that the generated code is more difficult for + // the optimizer, especially with bitfields. + unsigned numInitElements = args.size(); + RecordDecl *record = e->getType()->castAs<RecordType>()->getDecl(); + + // We'll need to enter cleanup scopes in case any of the element + // initializers throws an exception. + assert(!cir::MissingFeatures::requiresCleanups()); + + unsigned curInitIndex = 0; + + // Emit initialization of base classes. + if (auto *cxxrd = dyn_cast<CXXRecordDecl>(record)) { + assert(numInitElements >= cxxrd->getNumBases() && + "missing initializer for base class"); + if (cxxrd->getNumBases() > 0) { + cgf.cgm.errorNYI(e->getSourceRange(), + "visitCXXParenListOrInitListExpr base class init"); + return; + } + } + + LValue destLV = cgf.makeAddrLValue(dest.getAddress(), e->getType()); + + if (record->isUnion()) { + cgf.cgm.errorNYI(e->getSourceRange(), + "visitCXXParenListOrInitListExpr union type"); + return; } - cgf.cgm.errorNYI( - "visitCXXParenListOrInitListExpr Record or VariableSizeArray type"); + // Here we iterate over the fields; this makes it simpler to both + // default-initialize fields and skip over unnamed fields. + for (const FieldDecl *field : record->fields()) { + // We're done once we hit the flexible array member. + if (field->getType()->isIncompleteArrayType()) + break; + + // Always skip anonymous bitfields. + if (field->isUnnamedBitField()) + continue; + + // We're done if we reach the end of the explicit initializers, we + // have a zeroed object, and the rest of the fields are + // zero-initializable. + if (curInitIndex == numInitElements && dest.isZeroed() && + cgf.getTypes().isZeroInitializable(e->getType())) + break; + LValue lv = + cgf.emitLValueForFieldInitialization(destLV, field, field->getName()); + // We never generate write-barriers for initialized fields. + assert(!cir::MissingFeatures::setNonGC()); + + if (curInitIndex < numInitElements) { + // Store the initializer into the field. + CIRGenFunction::SourceLocRAIIObject loc{ + cgf, cgf.getLoc(record->getSourceRange())}; + emitInitializationToLValue(args[curInitIndex++], lv); + } else { + // We're out of initializers; default-initialize to null + emitNullInitializationToLValue(cgf.getLoc(e->getSourceRange()), lv); + } + + // Push a destructor if necessary. + // FIXME: if we have an array of structures, all explicitly + // initialized, we can end up pushing a linear number of cleanups. + if (field->getType().isDestructedType()) { + cgf.cgm.errorNYI(e->getSourceRange(), + "visitCXXParenListOrInitListExpr destructor"); + return; + } + + // From classic codegen, maybe not useful for CIR: + // If the GEP didn't get used because of a dead zero init or something + // else, clean it up for -O0 builds and general tidiness. + } } // TODO(cir): This could be shared with classic codegen. diff --git a/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp b/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp index d52da0c..02685a3 100644 --- a/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp @@ -60,7 +60,7 @@ public: mlir::Value VisitDeclRefExpr(DeclRefExpr *e); mlir::Value VisitGenericSelectionExpr(GenericSelectionExpr *e); mlir::Value VisitImplicitCastExpr(ImplicitCastExpr *e); - mlir::Value VisitInitListExpr(const InitListExpr *e); + mlir::Value VisitInitListExpr(InitListExpr *e); mlir::Value VisitCompoundLiteralExpr(CompoundLiteralExpr *e) { return emitLoadOfLValue(e); @@ -194,8 +194,12 @@ mlir::Value ComplexExprEmitter::emitCast(CastKind ck, Expr *op, } case CK_LValueToRValueBitCast: { - cgf.cgm.errorNYI("ComplexExprEmitter::emitCast CK_LValueToRValueBitCast"); - return {}; + LValue sourceLVal = cgf.emitLValue(op); + Address addr = sourceLVal.getAddress().withElementType( + builder, cgf.convertTypeForMem(destTy)); + LValue destLV = cgf.makeAddrLValue(addr, destTy); + assert(!cir::MissingFeatures::opTBAA()); + return emitLoadOfLValue(destLV, op->getExprLoc()); } case CK_BitCast: @@ -448,7 +452,7 @@ mlir::Value ComplexExprEmitter::VisitImplicitCastExpr(ImplicitCastExpr *e) { return emitCast(e->getCastKind(), e->getSubExpr(), e->getType()); } -mlir::Value ComplexExprEmitter::VisitInitListExpr(const InitListExpr *e) { +mlir::Value ComplexExprEmitter::VisitInitListExpr(InitListExpr *e) { mlir::Location loc = cgf.getLoc(e->getExprLoc()); if (e->getNumInits() == 2) { mlir::Value real = cgf.emitScalarExpr(e->getInit(0)); @@ -456,10 +460,8 @@ mlir::Value ComplexExprEmitter::VisitInitListExpr(const InitListExpr *e) { return builder.createComplexCreate(loc, real, imag); } - if (e->getNumInits() == 1) { - cgf.cgm.errorNYI("Create Complex with InitList with size 1"); - return {}; - } + if (e->getNumInits() == 1) + return Visit(e->getInit(0)); assert(e->getNumInits() == 0 && "Unexpected number of inits"); mlir::Type complexTy = cgf.convertType(e->getType()); diff --git a/clang/lib/CIR/CodeGen/CIRGenFunction.cpp b/clang/lib/CIR/CodeGen/CIRGenFunction.cpp index 3e69e56..b4b95d6 100644 --- a/clang/lib/CIR/CodeGen/CIRGenFunction.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenFunction.cpp @@ -26,7 +26,11 @@ namespace clang::CIRGen { CIRGenFunction::CIRGenFunction(CIRGenModule &cgm, CIRGenBuilderTy &builder, bool suppressNewContext) - : CIRGenTypeCache(cgm), cgm{cgm}, builder(builder) {} + : CIRGenTypeCache(cgm), cgm{cgm}, builder(builder) { + ehStack.setCGF(this); + currentCleanupStackDepth = 0; + assert(ehStack.getStackDepth() == 0); +} CIRGenFunction::~CIRGenFunction() {} @@ -227,6 +231,14 @@ void CIRGenFunction::LexicalScope::cleanup() { CIRGenBuilderTy &builder = cgf.builder; LexicalScope *localScope = cgf.curLexScope; + auto applyCleanup = [&]() { + if (performCleanup) { + // ApplyDebugLocation + assert(!cir::MissingFeatures::generateDebugInfo()); + forceCleanup(); + } + }; + if (returnBlock != nullptr) { // Write out the return block, which loads the value from `__retval` and // issues the `cir.return`. @@ -235,32 +247,42 @@ void CIRGenFunction::LexicalScope::cleanup() { (void)emitReturn(*returnLoc); } - mlir::Block *curBlock = builder.getBlock(); - if (isGlobalInit() && !curBlock) - return; - if (curBlock->mightHaveTerminator() && curBlock->getTerminator()) - return; - - // Get rid of any empty block at the end of the scope. - bool entryBlock = builder.getInsertionBlock()->isEntryBlock(); - if (!entryBlock && curBlock->empty()) { - curBlock->erase(); - if (returnBlock != nullptr && returnBlock->getUses().empty()) - returnBlock->erase(); - return; - } - - // Reached the end of the scope. - { + auto insertCleanupAndLeave = [&](mlir::Block *insPt) { mlir::OpBuilder::InsertionGuard guard(builder); - builder.setInsertionPointToEnd(curBlock); + builder.setInsertionPointToEnd(insPt); + + // If we still don't have a cleanup block, it means that `applyCleanup` + // below might be able to get us one. + mlir::Block *cleanupBlock = localScope->getCleanupBlock(builder); + + // Leverage and defers to RunCleanupsScope's dtor and scope handling. + applyCleanup(); + + // If we now have one after `applyCleanup`, hook it up properly. + if (!cleanupBlock && localScope->getCleanupBlock(builder)) { + cleanupBlock = localScope->getCleanupBlock(builder); + builder.create<cir::BrOp>(insPt->back().getLoc(), cleanupBlock); + if (!cleanupBlock->mightHaveTerminator()) { + mlir::OpBuilder::InsertionGuard guard(builder); + builder.setInsertionPointToEnd(cleanupBlock); + builder.create<cir::YieldOp>(localScope->endLoc); + } + } if (localScope->depth == 0) { // Reached the end of the function. if (returnBlock != nullptr) { - if (returnBlock->getUses().empty()) + if (returnBlock->getUses().empty()) { returnBlock->erase(); - else { + } else { + // Thread return block via cleanup block. + if (cleanupBlock) { + for (mlir::BlockOperand &blockUse : returnBlock->getUses()) { + cir::BrOp brOp = mlir::cast<cir::BrOp>(blockUse.getOwner()); + brOp.setSuccessor(cleanupBlock); + } + } + builder.create<cir::BrOp>(*returnLoc, returnBlock); return; } @@ -268,13 +290,50 @@ void CIRGenFunction::LexicalScope::cleanup() { emitImplicitReturn(); return; } - // Reached the end of a non-function scope. Some scopes, such as those - // used with the ?: operator, can return a value. - if (!localScope->isTernary() && !curBlock->mightHaveTerminator()) { + + // End of any local scope != function + // Ternary ops have to deal with matching arms for yielding types + // and do return a value, it must do its own cir.yield insertion. + if (!localScope->isTernary() && !insPt->mightHaveTerminator()) { !retVal ? builder.create<cir::YieldOp>(localScope->endLoc) : builder.create<cir::YieldOp>(localScope->endLoc, retVal); } + }; + + // If a cleanup block has been created at some point, branch to it + // and set the insertion point to continue at the cleanup block. + // Terminators are then inserted either in the cleanup block or + // inline in this current block. + mlir::Block *cleanupBlock = localScope->getCleanupBlock(builder); + if (cleanupBlock) + insertCleanupAndLeave(cleanupBlock); + + // Now deal with any pending block wrap up like implicit end of + // scope. + + mlir::Block *curBlock = builder.getBlock(); + if (isGlobalInit() && !curBlock) + return; + if (curBlock->mightHaveTerminator() && curBlock->getTerminator()) + return; + + // Get rid of any empty block at the end of the scope. + bool entryBlock = builder.getInsertionBlock()->isEntryBlock(); + if (!entryBlock && curBlock->empty()) { + curBlock->erase(); + if (returnBlock != nullptr && returnBlock->getUses().empty()) + returnBlock->erase(); + return; } + + // If there's a cleanup block, branch to it, nothing else to do. + if (cleanupBlock) { + builder.create<cir::BrOp>(curBlock->back().getLoc(), cleanupBlock); + return; + } + + // No pre-existent cleanup block, emit cleanup code and yield/return. + insertCleanupAndLeave(curBlock); } cir::ReturnOp CIRGenFunction::LexicalScope::emitReturn(mlir::Location loc) { @@ -408,7 +467,19 @@ void CIRGenFunction::startFunction(GlobalDecl gd, QualType returnType, } } -void CIRGenFunction::finishFunction(SourceLocation endLoc) {} +void CIRGenFunction::finishFunction(SourceLocation endLoc) { + // Pop any cleanups that might have been associated with the + // parameters. Do this in whatever block we're currently in; it's + // important to do this before we enter the return block or return + // edges will be *really* confused. + // TODO(cir): Use prologueCleanupDepth here. + bool hasCleanups = ehStack.getStackDepth() != currentCleanupStackDepth; + if (hasCleanups) { + assert(!cir::MissingFeatures::generateDebugInfo()); + // FIXME(cir): should we clearInsertionPoint? breaks many testcases + popCleanupBlocks(currentCleanupStackDepth); + } +} mlir::LogicalResult CIRGenFunction::emitFunctionBody(const clang::Stmt *body) { auto result = mlir::LogicalResult::success(); @@ -808,4 +879,48 @@ bool CIRGenFunction::shouldNullCheckClassCastValue(const CastExpr *ce) { return true; } +/// Computes the length of an array in elements, as well as the base +/// element type and a properly-typed first element pointer. +mlir::Value +CIRGenFunction::emitArrayLength(const clang::ArrayType *origArrayType, + QualType &baseType, Address &addr) { + const clang::ArrayType *arrayType = origArrayType; + + // If it's a VLA, we have to load the stored size. Note that + // this is the size of the VLA in bytes, not its size in elements. + if (isa<VariableArrayType>(arrayType)) { + assert(cir::MissingFeatures::vlas()); + cgm.errorNYI(*currSrcLoc, "VLAs"); + return builder.getConstInt(*currSrcLoc, SizeTy, 0); + } + + uint64_t countFromCLAs = 1; + QualType eltType; + + auto cirArrayType = mlir::dyn_cast<cir::ArrayType>(addr.getElementType()); + + while (cirArrayType) { + assert(isa<ConstantArrayType>(arrayType)); + countFromCLAs *= cirArrayType.getSize(); + eltType = arrayType->getElementType(); + + cirArrayType = + mlir::dyn_cast<cir::ArrayType>(cirArrayType.getElementType()); + + arrayType = getContext().getAsArrayType(arrayType->getElementType()); + assert((!cirArrayType || arrayType) && + "CIR and Clang types are out-of-sync"); + } + + if (arrayType) { + // From this point onwards, the Clang array type has been emitted + // as some other type (probably a packed struct). Compute the array + // size, and just emit the 'begin' expression as a bitcast. + cgm.errorNYI(*currSrcLoc, "length for non-array underlying types"); + } + + baseType = eltType; + return builder.getConstInt(*currSrcLoc, SizeTy, countFromCLAs); +} + } // namespace clang::CIRGen diff --git a/clang/lib/CIR/CodeGen/CIRGenFunction.h b/clang/lib/CIR/CodeGen/CIRGenFunction.h index 2aceeef..4891c74 100644 --- a/clang/lib/CIR/CodeGen/CIRGenFunction.h +++ b/clang/lib/CIR/CodeGen/CIRGenFunction.h @@ -18,6 +18,7 @@ #include "CIRGenModule.h" #include "CIRGenTypeCache.h" #include "CIRGenValue.h" +#include "EHScopeStack.h" #include "Address.h" @@ -61,6 +62,9 @@ public: /// The compiler-generated variable that holds the return value. std::optional<mlir::Value> fnRetAlloca; + /// Tracks function scope overall cleanup handling. + EHScopeStack ehStack; + /// CXXThisDecl - When generating code for a C++ member function, /// this will hold the implicit 'this' declaration. ImplicitParamDecl *cxxabiThisDecl = nullptr; @@ -595,14 +599,65 @@ public: FunctionArgList args, clang::SourceLocation loc, clang::SourceLocation startLoc); + /// Takes the old cleanup stack size and emits the cleanup blocks + /// that have been added. + void popCleanupBlocks(size_t oldCleanupStackDepth); + void popCleanupBlock(); + + /// Enters a new scope for capturing cleanups, all of which + /// will be executed once the scope is exited. + class RunCleanupsScope { + size_t cleanupStackDepth, oldCleanupStackDepth; + + protected: + bool performCleanup; + + private: + RunCleanupsScope(const RunCleanupsScope &) = delete; + void operator=(const RunCleanupsScope &) = delete; + + protected: + CIRGenFunction &cgf; + + /// Enter a new cleanup scope. + explicit RunCleanupsScope(CIRGenFunction &cgf) + : performCleanup(true), cgf(cgf) { + cleanupStackDepth = cgf.ehStack.getStackDepth(); + oldCleanupStackDepth = cgf.currentCleanupStackDepth; + cgf.currentCleanupStackDepth = cleanupStackDepth; + } + + /// Exit this cleanup scope, emitting any accumulated cleanups. + ~RunCleanupsScope() { + if (performCleanup) + forceCleanup(); + } + + /// Force the emission of cleanups now, instead of waiting + /// until this object is destroyed. + void forceCleanup() { + assert(performCleanup && "Already forced cleanup"); + { + mlir::OpBuilder::InsertionGuard guard(cgf.getBuilder()); + cgf.popCleanupBlocks(cleanupStackDepth); + performCleanup = false; + cgf.currentCleanupStackDepth = oldCleanupStackDepth; + } + } + }; + + // Cleanup stack depth of the RunCleanupsScope that was pushed most recently. + size_t currentCleanupStackDepth; + +public: /// Represents a scope, including function bodies, compound statements, and /// the substatements of if/while/do/for/switch/try statements. This class /// handles any automatic cleanup, along with the return value. - struct LexicalScope { + struct LexicalScope : public RunCleanupsScope { private: - // TODO(CIR): This will live in the base class RunCleanupScope once that - // class is upstreamed. - CIRGenFunction &cgf; + // Block containing cleanup code for things initialized in this + // lexical context (scope). + mlir::Block *cleanupBlock = nullptr; // Points to the scope entry block. This is useful, for instance, for // helping to insert allocas before finalizing any recursive CodeGen from @@ -632,8 +687,8 @@ public: unsigned depth = 0; LexicalScope(CIRGenFunction &cgf, mlir::Location loc, mlir::Block *eb) - : cgf(cgf), entryBlock(eb), parentScope(cgf.curLexScope), beginLoc(loc), - endLoc(loc) { + : RunCleanupsScope(cgf), entryBlock(eb), parentScope(cgf.curLexScope), + beginLoc(loc), endLoc(loc) { assert(entryBlock && "LexicalScope requires an entry block"); cgf.curLexScope = this; @@ -671,6 +726,27 @@ public: void setAsSwitch() { scopeKind = Kind::Switch; } void setAsTernary() { scopeKind = Kind::Ternary; } + // Lazy create cleanup block or return what's available. + mlir::Block *getOrCreateCleanupBlock(mlir::OpBuilder &builder) { + if (cleanupBlock) + return cleanupBlock; + cleanupBlock = createCleanupBlock(builder); + return cleanupBlock; + } + + mlir::Block *getCleanupBlock(mlir::OpBuilder &builder) { + return cleanupBlock; + } + + mlir::Block *createCleanupBlock(mlir::OpBuilder &builder) { + // Create the cleanup block but dont hook it up around just yet. + mlir::OpBuilder::InsertionGuard guard(builder); + mlir::Region *r = builder.getBlock() ? builder.getBlock()->getParent() + : &cgf.curFn->getRegion(0); + cleanupBlock = builder.createBlock(r); + return cleanupBlock; + } + // --- // Return handling. // --- @@ -721,6 +797,12 @@ public: LexicalScope *curLexScope = nullptr; + typedef void Destroyer(CIRGenFunction &cgf, Address addr, QualType ty); + + static Destroyer destroyCXXObject; + + Destroyer *getDestroyer(clang::QualType::DestructionKind kind); + /// ---------------------- /// CIR emit functions /// ---------------------- @@ -766,6 +848,8 @@ public: /// even if no aggregate location is provided. RValue emitAnyExprToTemp(const clang::Expr *e); + mlir::Value emitArrayLength(const clang::ArrayType *arrayType, + QualType &baseType, Address &addr); LValue emitArraySubscriptExpr(const clang::ArraySubscriptExpr *e); Address emitArrayToPointerDecay(const Expr *array); @@ -779,6 +863,8 @@ public: void emitAutoVarCleanups(const AutoVarEmission &emission); void emitAutoVarInit(const AutoVarEmission &emission); + void emitAutoVarTypeCleanup(const AutoVarEmission &emission, + clang::QualType::DestructionKind dtorKind); void emitBaseInitializer(mlir::Location loc, const CXXRecordDecl *classDecl, CXXCtorInitializer *baseInit); @@ -836,6 +922,9 @@ public: LValue emitCompoundLiteralLValue(const CompoundLiteralExpr *e); void emitConstructorBody(FunctionArgList &args); + + void emitDestroy(Address addr, QualType type, Destroyer *destroyer); + void emitDestructorBody(FunctionArgList &args); mlir::LogicalResult emitContinueStmt(const clang::ContinueStmt &s); @@ -843,6 +932,16 @@ public: void emitCXXConstructExpr(const clang::CXXConstructExpr *e, AggValueSlot dest); + void emitCXXAggrConstructorCall(const CXXConstructorDecl *ctor, + const clang::ArrayType *arrayType, + Address arrayBegin, const CXXConstructExpr *e, + bool newPointerIsChecked, + bool zeroInitialize = false); + void emitCXXAggrConstructorCall(const CXXConstructorDecl *ctor, + mlir::Value numElements, Address arrayBase, + const CXXConstructExpr *e, + bool newPointerIsChecked, + bool zeroInitialize); void emitCXXConstructorCall(const clang::CXXConstructorDecl *d, clang::CXXCtorType type, bool forVirtualBase, bool delegating, AggValueSlot thisAVS, diff --git a/clang/lib/CIR/CodeGen/CIRGenItaniumCXXABI.cpp b/clang/lib/CIR/CodeGen/CIRGenItaniumCXXABI.cpp index 6577f5f..e5e4c68 100644 --- a/clang/lib/CIR/CodeGen/CIRGenItaniumCXXABI.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenItaniumCXXABI.cpp @@ -113,8 +113,6 @@ static StructorCIRGen getCIRGenToUse(CIRGenModule &cgm, GlobalDecl aliasDecl; if (const auto *dd = dyn_cast<CXXDestructorDecl>(md)) { - // The assignment is correct here, but other support for this is NYI. - cgm.errorNYI(md->getSourceRange(), "getCIRGenToUse: dtor"); aliasDecl = GlobalDecl(dd, Dtor_Complete); } else { const auto *cd = cast<CXXConstructorDecl>(md); diff --git a/clang/lib/CIR/CodeGen/CIRGenStmt.cpp b/clang/lib/CIR/CodeGen/CIRGenStmt.cpp index 9193f6f..21bee33 100644 --- a/clang/lib/CIR/CodeGen/CIRGenStmt.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenStmt.cpp @@ -409,7 +409,10 @@ mlir::LogicalResult CIRGenFunction::emitReturnStmt(const ReturnStmt &s) { } auto *retBlock = curLexScope->getOrCreateRetBlock(*this, loc); + // This should emit a branch through the cleanup block if one exists. builder.create<cir::BrOp>(loc, retBlock); + if (ehStack.getStackDepth() != currentCleanupStackDepth) + cgm.errorNYI(s.getSourceRange(), "return with cleanup stack"); builder.createBlock(builder.getBlock()->getParent()); return mlir::success(); diff --git a/clang/lib/CIR/CodeGen/CMakeLists.txt b/clang/lib/CIR/CodeGen/CMakeLists.txt index 03ea60c..ca3a329 100644 --- a/clang/lib/CIR/CodeGen/CMakeLists.txt +++ b/clang/lib/CIR/CodeGen/CMakeLists.txt @@ -11,6 +11,7 @@ add_clang_library(clangCIR CIRGenBuilder.cpp CIRGenCall.cpp CIRGenClass.cpp + CIRGenCleanup.cpp CIRGenCXX.cpp CIRGenCXXABI.cpp CIRGenCXXExpr.cpp diff --git a/clang/lib/CIR/CodeGen/EHScopeStack.h b/clang/lib/CIR/CodeGen/EHScopeStack.h new file mode 100644 index 0000000..22750ac --- /dev/null +++ b/clang/lib/CIR/CodeGen/EHScopeStack.h @@ -0,0 +1,99 @@ +//===-- EHScopeStack.h - Stack for cleanup CIR generation -------*- C++ -*-===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// These classes should be the minimum interface required for other parts of +// CIR CodeGen to emit cleanups. The implementation is in CIRGenCleanup.cpp and +// other implemenentation details that are not widely needed are in +// CIRGenCleanup.h. +// +// TODO(cir): this header should be shared between LLVM and CIR codegen. +// +//===----------------------------------------------------------------------===// + +#ifndef CLANG_LIB_CIR_CODEGEN_EHSCOPESTACK_H +#define CLANG_LIB_CIR_CODEGEN_EHSCOPESTACK_H + +#include "llvm/ADT/SmallVector.h" + +namespace clang::CIRGen { + +class CIRGenFunction; + +enum CleanupKind : unsigned { + /// Denotes a cleanup that should run when a scope is exited using exceptional + /// control flow (a throw statement leading to stack unwinding, ). + EHCleanup = 0x1, + + /// Denotes a cleanup that should run when a scope is exited using normal + /// control flow (falling off the end of the scope, return, goto, ...). + NormalCleanup = 0x2, + + NormalAndEHCleanup = EHCleanup | NormalCleanup, + + LifetimeMarker = 0x8, + NormalEHLifetimeMarker = LifetimeMarker | NormalAndEHCleanup, +}; + +/// A stack of scopes which respond to exceptions, including cleanups +/// and catch blocks. +class EHScopeStack { +public: + /// Information for lazily generating a cleanup. Subclasses must be + /// POD-like: cleanups will not be destructed, and they will be + /// allocated on the cleanup stack and freely copied and moved + /// around. + /// + /// Cleanup implementations should generally be declared in an + /// anonymous namespace. + class Cleanup { + // Anchor the construction vtable. + virtual void anchor(); + + public: + Cleanup(const Cleanup &) = default; + Cleanup(Cleanup &&) {} + Cleanup() = default; + + virtual ~Cleanup() = default; + + /// Emit the cleanup. For normal cleanups, this is run in the + /// same EH context as when the cleanup was pushed, i.e. the + /// immediately-enclosing context of the cleanup scope. For + /// EH cleanups, this is run in a terminate context. + /// + // \param flags cleanup kind. + virtual void emit(CIRGenFunction &cgf) = 0; + }; + + // Classic codegen has a finely tuned custom allocator and a complex stack + // management scheme. We'll probably eventually want to find a way to share + // that implementation. For now, we will use a very simplified implementation + // to get cleanups working. + llvm::SmallVector<std::unique_ptr<Cleanup>, 8> cleanupStack; + +private: + /// The CGF this Stack belong to + CIRGenFunction *cgf = nullptr; + +public: + EHScopeStack() = default; + ~EHScopeStack() = default; + + /// Push a lazily-created cleanup on the stack. + template <class T, class... As> void pushCleanup(CleanupKind kind, As... a) { + cleanupStack.push_back(std::make_unique<T>(a...)); + } + + void setCGF(CIRGenFunction *inCGF) { cgf = inCGF; } + + size_t getStackDepth() const { return cleanupStack.size(); } +}; + +} // namespace clang::CIRGen + +#endif // CLANG_LIB_CIR_CODEGEN_EHSCOPESTACK_H diff --git a/clang/lib/CIR/Dialect/IR/CIRDialect.cpp b/clang/lib/CIR/Dialect/IR/CIRDialect.cpp index cd77166..2213c75 100644 --- a/clang/lib/CIR/Dialect/IR/CIRDialect.cpp +++ b/clang/lib/CIR/Dialect/IR/CIRDialect.cpp @@ -17,6 +17,7 @@ #include "mlir/Interfaces/ControlFlowInterfaces.h" #include "mlir/Interfaces/FunctionImplementation.h" +#include "mlir/Support/LLVM.h" #include "clang/CIR/Dialect/IR/CIROpsDialect.cpp.inc" #include "clang/CIR/Dialect/IR/CIROpsEnums.cpp.inc" @@ -338,7 +339,7 @@ static LogicalResult checkConstantTypes(mlir::Operation *op, mlir::Type opType, } if (mlir::isa<cir::ConstArrayAttr, cir::ConstVectorAttr, - cir::ConstComplexAttr>(attrType)) + cir::ConstComplexAttr, cir::PoisonAttr>(attrType)) return success(); assert(isa<TypedAttr>(attrType) && "What else could we be looking at here?"); @@ -628,6 +629,11 @@ static Value tryFoldCastChain(cir::CastOp op) { } OpFoldResult cir::CastOp::fold(FoldAdaptor adaptor) { + if (mlir::isa_and_present<cir::PoisonAttr>(adaptor.getSrc())) { + // Propagate poison value + return cir::PoisonAttr::get(getContext(), getType()); + } + if (getSrc().getType() == getType()) { switch (getKind()) { case cir::CastKind::integral: { @@ -1782,6 +1788,12 @@ static bool isBoolNot(cir::UnaryOp op) { // // and the argument of the first one (%0) will be used instead. OpFoldResult cir::UnaryOp::fold(FoldAdaptor adaptor) { + if (auto poison = + mlir::dyn_cast_if_present<cir::PoisonAttr>(adaptor.getInput())) { + // Propagate poison values + return poison; + } + if (isBoolNot(*this)) if (auto previous = dyn_cast_or_null<UnaryOp>(getInput().getDefiningOp())) if (isBoolNot(previous)) @@ -2231,6 +2243,134 @@ LogicalResult cir::ComplexImagPtrOp::verify() { } //===----------------------------------------------------------------------===// +// Bit manipulation operations +//===----------------------------------------------------------------------===// + +static OpFoldResult +foldUnaryBitOp(mlir::Attribute inputAttr, + llvm::function_ref<llvm::APInt(const llvm::APInt &)> func, + bool poisonZero = false) { + if (mlir::isa_and_present<cir::PoisonAttr>(inputAttr)) { + // Propagate poison value + return inputAttr; + } + + auto input = mlir::dyn_cast_if_present<IntAttr>(inputAttr); + if (!input) + return nullptr; + + llvm::APInt inputValue = input.getValue(); + if (poisonZero && inputValue.isZero()) + return cir::PoisonAttr::get(input.getType()); + + llvm::APInt resultValue = func(inputValue); + return IntAttr::get(input.getType(), resultValue); +} + +OpFoldResult BitClrsbOp::fold(FoldAdaptor adaptor) { + return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) { + unsigned resultValue = + inputValue.getBitWidth() - inputValue.getSignificantBits(); + return llvm::APInt(inputValue.getBitWidth(), resultValue); + }); +} + +OpFoldResult BitClzOp::fold(FoldAdaptor adaptor) { + return foldUnaryBitOp( + adaptor.getInput(), + [](const llvm::APInt &inputValue) { + unsigned resultValue = inputValue.countLeadingZeros(); + return llvm::APInt(inputValue.getBitWidth(), resultValue); + }, + getPoisonZero()); +} + +OpFoldResult BitCtzOp::fold(FoldAdaptor adaptor) { + return foldUnaryBitOp( + adaptor.getInput(), + [](const llvm::APInt &inputValue) { + return llvm::APInt(inputValue.getBitWidth(), + inputValue.countTrailingZeros()); + }, + getPoisonZero()); +} + +OpFoldResult BitParityOp::fold(FoldAdaptor adaptor) { + return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) { + return llvm::APInt(inputValue.getBitWidth(), inputValue.popcount() % 2); + }); +} + +OpFoldResult BitPopcountOp::fold(FoldAdaptor adaptor) { + return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) { + return llvm::APInt(inputValue.getBitWidth(), inputValue.popcount()); + }); +} + +OpFoldResult BitReverseOp::fold(FoldAdaptor adaptor) { + return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) { + return inputValue.reverseBits(); + }); +} + +OpFoldResult ByteSwapOp::fold(FoldAdaptor adaptor) { + return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) { + return inputValue.byteSwap(); + }); +} + +OpFoldResult RotateOp::fold(FoldAdaptor adaptor) { + if (mlir::isa_and_present<cir::PoisonAttr>(adaptor.getInput()) || + mlir::isa_and_present<cir::PoisonAttr>(adaptor.getAmount())) { + // Propagate poison values + return cir::PoisonAttr::get(getType()); + } + + auto input = mlir::dyn_cast_if_present<IntAttr>(adaptor.getInput()); + auto amount = mlir::dyn_cast_if_present<IntAttr>(adaptor.getAmount()); + if (!input && !amount) + return nullptr; + + // We could fold cir.rotate even if one of its two operands is not a constant: + // - `cir.rotate left/right %0, 0` could be folded into just %0 even if %0 + // is not a constant. + // - `cir.rotate left/right 0/0b111...111, %0` could be folded into 0 or + // 0b111...111 even if %0 is not a constant. + + llvm::APInt inputValue; + if (input) { + inputValue = input.getValue(); + if (inputValue.isZero() || inputValue.isAllOnes()) { + // An input value of all 0s or all 1s will not change after rotation + return input; + } + } + + uint64_t amountValue; + if (amount) { + amountValue = amount.getValue().urem(getInput().getType().getWidth()); + if (amountValue == 0) { + // A shift amount of 0 will not change the input value + return getInput(); + } + } + + if (!input || !amount) + return nullptr; + + assert(inputValue.getBitWidth() == getInput().getType().getWidth() && + "input value must have the same bit width as the input type"); + + llvm::APInt resultValue; + if (isRotateLeft()) + resultValue = inputValue.rotl(amountValue); + else + resultValue = inputValue.rotr(amountValue); + + return IntAttr::get(input.getContext(), input.getType(), resultValue); +} + +//===----------------------------------------------------------------------===// // TableGen'd op method definitions //===----------------------------------------------------------------------===// diff --git a/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp b/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp index e505db5..2143f16 100644 --- a/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp +++ b/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp @@ -143,7 +143,8 @@ void CIRCanonicalizePass::runOnOperation() { if (isa<BrOp, BrCondOp, CastOp, ScopeOp, SwitchOp, SelectOp, UnaryOp, ComplexCreateOp, ComplexImagOp, ComplexRealOp, VecCmpOp, VecCreateOp, VecExtractOp, VecShuffleOp, VecShuffleDynamicOp, - VecTernaryOp>(op)) + VecTernaryOp, BitClrsbOp, BitClzOp, BitCtzOp, BitParityOp, + BitPopcountOp, BitReverseOp, ByteSwapOp, RotateOp>(op)) ops.push_back(op); }); diff --git a/clang/lib/CIR/Dialect/Transforms/LoweringPrepare.cpp b/clang/lib/CIR/Dialect/Transforms/LoweringPrepare.cpp index 0f13417..cef83ea 100644 --- a/clang/lib/CIR/Dialect/Transforms/LoweringPrepare.cpp +++ b/clang/lib/CIR/Dialect/Transforms/LoweringPrepare.cpp @@ -8,10 +8,12 @@ #include "PassDetail.h" #include "clang/AST/ASTContext.h" +#include "clang/AST/CharUnits.h" #include "clang/CIR/Dialect/Builder/CIRBaseBuilder.h" #include "clang/CIR/Dialect/IR/CIRDialect.h" #include "clang/CIR/Dialect/IR/CIROpsEnums.h" #include "clang/CIR/Dialect/Passes.h" +#include "clang/CIR/MissingFeatures.h" #include <iostream> #include <memory> @@ -27,6 +29,15 @@ struct LoweringPreparePass : public LoweringPrepareBase<LoweringPreparePass> { void runOnOp(mlir::Operation *op); void lowerCastOp(cir::CastOp op); void lowerUnaryOp(cir::UnaryOp op); + void lowerArrayCtor(ArrayCtor op); + + /// + /// AST related + /// ----------- + + clang::ASTContext *astCtx; + + void setASTContext(clang::ASTContext *c) { astCtx = c; } }; } // namespace @@ -158,8 +169,83 @@ void LoweringPreparePass::lowerUnaryOp(cir::UnaryOp op) { op.erase(); } +static void lowerArrayDtorCtorIntoLoop(cir::CIRBaseBuilderTy &builder, + clang::ASTContext *astCtx, + mlir::Operation *op, mlir::Type eltTy, + mlir::Value arrayAddr, + uint64_t arrayLen) { + // Generate loop to call into ctor/dtor for every element. + mlir::Location loc = op->getLoc(); + + // TODO: instead of fixed integer size, create alias for PtrDiffTy and unify + // with CIRGen stuff. + const unsigned sizeTypeSize = + astCtx->getTypeSize(astCtx->getSignedSizeType()); + auto ptrDiffTy = + cir::IntType::get(builder.getContext(), sizeTypeSize, /*isSigned=*/false); + mlir::Value numArrayElementsConst = builder.getUnsignedInt(loc, arrayLen, 64); + + auto begin = builder.create<cir::CastOp>( + loc, eltTy, cir::CastKind::array_to_ptrdecay, arrayAddr); + mlir::Value end = builder.create<cir::PtrStrideOp>(loc, eltTy, begin, + numArrayElementsConst); + + mlir::Value tmpAddr = builder.createAlloca( + loc, /*addr type*/ builder.getPointerTo(eltTy), + /*var type*/ eltTy, "__array_idx", builder.getAlignmentAttr(1)); + builder.createStore(loc, begin, tmpAddr); + + cir::DoWhileOp loop = builder.createDoWhile( + loc, + /*condBuilder=*/ + [&](mlir::OpBuilder &b, mlir::Location loc) { + auto currentElement = b.create<cir::LoadOp>(loc, eltTy, tmpAddr); + mlir::Type boolTy = cir::BoolType::get(b.getContext()); + auto cmp = builder.create<cir::CmpOp>(loc, boolTy, cir::CmpOpKind::ne, + currentElement, end); + builder.createCondition(cmp); + }, + /*bodyBuilder=*/ + [&](mlir::OpBuilder &b, mlir::Location loc) { + auto currentElement = b.create<cir::LoadOp>(loc, eltTy, tmpAddr); + + cir::CallOp ctorCall; + op->walk([&](cir::CallOp c) { ctorCall = c; }); + assert(ctorCall && "expected ctor call"); + + auto one = builder.create<cir::ConstantOp>( + loc, ptrDiffTy, cir::IntAttr::get(ptrDiffTy, 1)); + + ctorCall->moveAfter(one); + ctorCall->setOperand(0, currentElement); + + // Advance pointer and store them to temporary variable + auto nextElement = + builder.create<cir::PtrStrideOp>(loc, eltTy, currentElement, one); + builder.createStore(loc, nextElement, tmpAddr); + builder.createYield(loc); + }); + + op->replaceAllUsesWith(loop); + op->erase(); +} + +void LoweringPreparePass::lowerArrayCtor(cir::ArrayCtor op) { + cir::CIRBaseBuilderTy builder(getContext()); + builder.setInsertionPointAfter(op.getOperation()); + + mlir::Type eltTy = op->getRegion(0).getArgument(0).getType(); + assert(!cir::MissingFeatures::vlas()); + auto arrayLen = + mlir::cast<cir::ArrayType>(op.getAddr().getType().getPointee()).getSize(); + lowerArrayDtorCtorIntoLoop(builder, astCtx, op, eltTy, op.getAddr(), + arrayLen); +} + void LoweringPreparePass::runOnOp(mlir::Operation *op) { - if (auto cast = mlir::dyn_cast<cir::CastOp>(op)) + if (auto arrayCtor = dyn_cast<ArrayCtor>(op)) + lowerArrayCtor(arrayCtor); + else if (auto cast = mlir::dyn_cast<cir::CastOp>(op)) lowerCastOp(cast); else if (auto unary = mlir::dyn_cast<cir::UnaryOp>(op)) lowerUnaryOp(unary); @@ -171,7 +257,7 @@ void LoweringPreparePass::runOnOperation() { llvm::SmallVector<mlir::Operation *> opsToTransform; op->walk([&](mlir::Operation *op) { - if (mlir::isa<cir::CastOp, cir::UnaryOp>(op)) + if (mlir::isa<cir::ArrayCtor, cir::CastOp, cir::UnaryOp>(op)) opsToTransform.push_back(op); }); @@ -182,3 +268,10 @@ void LoweringPreparePass::runOnOperation() { std::unique_ptr<Pass> mlir::createLoweringPreparePass() { return std::make_unique<LoweringPreparePass>(); } + +std::unique_ptr<Pass> +mlir::createLoweringPreparePass(clang::ASTContext *astCtx) { + auto pass = std::make_unique<LoweringPreparePass>(); + pass->setASTContext(astCtx); + return std::move(pass); +} diff --git a/clang/lib/CIR/Lowering/CIRPasses.cpp b/clang/lib/CIR/Lowering/CIRPasses.cpp index 5607abc..bb9781b 100644 --- a/clang/lib/CIR/Lowering/CIRPasses.cpp +++ b/clang/lib/CIR/Lowering/CIRPasses.cpp @@ -31,7 +31,7 @@ mlir::LogicalResult runCIRToCIRPasses(mlir::ModuleOp theModule, if (enableCIRSimplify) pm.addPass(mlir::createCIRSimplifyPass()); - pm.addPass(mlir::createLoweringPreparePass()); + pm.addPass(mlir::createLoweringPreparePass(&astContext)); pm.enableVerifier(enableVerifier); (void)mlir::applyPassManagerCLOptions(pm); diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp index 3cd7de0..c27b889 100644 --- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp +++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp @@ -1027,6 +1027,12 @@ mlir::LogicalResult CIRToLLVMConstantOpLowering::matchAndRewrite( mlir::ConversionPatternRewriter &rewriter) const { mlir::Attribute attr = op.getValue(); + if (mlir::isa<cir::PoisonAttr>(attr)) { + rewriter.replaceOpWithNewOp<mlir::LLVM::PoisonOp>( + op, getTypeConverter()->convertType(op.getType())); + return mlir::success(); + } + if (mlir::isa<mlir::IntegerType>(op.getType())) { // Verified cir.const operations cannot actually be of these types, but the // lowering pass may generate temporary cir.const operations with these diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp index 1b72578..0b8b824 100644 --- a/clang/lib/CodeGen/BackendUtil.cpp +++ b/clang/lib/CodeGen/BackendUtil.cpp @@ -1027,12 +1027,6 @@ void EmitAssemblyHelper::RunOptimizationPipeline( MPM.addPass( createModuleToFunctionPassAdaptor(ObjCARCExpandPass())); }); - PB.registerPipelineEarlySimplificationEPCallback( - [](ModulePassManager &MPM, OptimizationLevel Level, - ThinOrFullLTOPhase) { - if (Level != OptimizationLevel::O0) - MPM.addPass(ObjCARCAPElimPass()); - }); PB.registerScalarOptimizerLateEPCallback( [](FunctionPassManager &FPM, OptimizationLevel Level) { if (Level != OptimizationLevel::O0) diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp index 0bceece..d9bd443 100644 --- a/clang/lib/CodeGen/CGCall.cpp +++ b/clang/lib/CodeGen/CGCall.cpp @@ -2852,20 +2852,28 @@ void CodeGenModule::ConstructAttributeList(StringRef Name, if (AI.getInReg()) Attrs.addAttribute(llvm::Attribute::InReg); - // Depending on the ABI, this may be either a byval or a dead_on_return - // argument. - if (AI.getIndirectByVal()) { - Attrs.addByValAttr(getTypes().ConvertTypeForMem(ParamType)); - } else { - // Add dead_on_return when the object's lifetime ends in the callee. - // This includes trivially-destructible objects, as well as objects - // whose destruction / clean-up is carried out within the callee (e.g., - // Obj-C ARC-managed structs, MSVC callee-destroyed objects). - if (!ParamType.isDestructedType() || !ParamType->isRecordType() || - ParamType->castAs<RecordType>() - ->getDecl() - ->isParamDestroyedInCallee()) - Attrs.addAttribute(llvm::Attribute::DeadOnReturn); + // HLSL out and inout parameters must not be marked with ByVal or + // DeadOnReturn attributes because stores to these parameters by the + // callee are visible to the caller. + if (auto ParamABI = FI.getExtParameterInfo(ArgNo).getABI(); + ParamABI != ParameterABI::HLSLOut && + ParamABI != ParameterABI::HLSLInOut) { + + // Depending on the ABI, this may be either a byval or a dead_on_return + // argument. + if (AI.getIndirectByVal()) { + Attrs.addByValAttr(getTypes().ConvertTypeForMem(ParamType)); + } else { + // Add dead_on_return when the object's lifetime ends in the callee. + // This includes trivially-destructible objects, as well as objects + // whose destruction / clean-up is carried out within the callee + // (e.g., Obj-C ARC-managed structs, MSVC callee-destroyed objects). + if (!ParamType.isDestructedType() || !ParamType->isRecordType() || + ParamType->castAs<RecordType>() + ->getDecl() + ->isParamDestroyedInCallee()) + Attrs.addAttribute(llvm::Attribute::DeadOnReturn); + } } auto *Decl = ParamType->getAsRecordDecl(); diff --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp index a371b67..77fc3a2 100644 --- a/clang/lib/CodeGen/CGDebugInfo.cpp +++ b/clang/lib/CodeGen/CGDebugInfo.cpp @@ -6435,7 +6435,7 @@ CodeGenFunction::LexicalScope::~LexicalScope() { static std::string SanitizerHandlerToCheckLabel(SanitizerHandler Handler) { std::string Label; switch (Handler) { -#define SANITIZER_CHECK(Enum, Name, Version) \ +#define SANITIZER_CHECK(Enum, Name, Version, Msg) \ case Enum: \ Label = "__ubsan_check_" #Name; \ break; diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp index 85c7688..90aed79 100644 --- a/clang/lib/CodeGen/CGExpr.cpp +++ b/clang/lib/CodeGen/CGExpr.cpp @@ -85,6 +85,16 @@ enum VariableTypeDescriptorKind : uint16_t { // Miscellaneous Helper Methods //===--------------------------------------------------------------------===// +static llvm::StringRef GetUBSanTrapForHandler(SanitizerHandler ID) { + switch (ID) { +#define SANITIZER_CHECK(Enum, Name, Version, Msg) \ + case SanitizerHandler::Enum: \ + return Msg; + LIST_SANITIZER_CHECKS +#undef SANITIZER_CHECK + } +} + /// CreateTempAlloca - This creates a alloca and inserts it into the entry /// block. RawAddress @@ -3649,7 +3659,7 @@ struct SanitizerHandlerInfo { } const SanitizerHandlerInfo SanitizerHandlers[] = { -#define SANITIZER_CHECK(Enum, Name, Version) {#Name, Version}, +#define SANITIZER_CHECK(Enum, Name, Version, Msg) {#Name, Version}, LIST_SANITIZER_CHECKS #undef SANITIZER_CHECK }; @@ -3954,6 +3964,8 @@ void CodeGenFunction::EmitCfiCheckFail() { StartFunction(GlobalDecl(), CGM.getContext().VoidTy, F, FI, Args, SourceLocation()); + ApplyDebugLocation ADL = ApplyDebugLocation::CreateArtificial(*this); + // This function is not affected by NoSanitizeList. This function does // not have a source location, but "src:*" would still apply. Revert any // changes to SanOpts made in StartFunction. @@ -4051,6 +4063,15 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked, llvm::BasicBlock *&TrapBB = TrapBBs[CheckHandlerID]; + llvm::DILocation *TrapLocation = Builder.getCurrentDebugLocation(); + llvm::StringRef TrapMessage = GetUBSanTrapForHandler(CheckHandlerID); + + if (getDebugInfo() && !TrapMessage.empty() && + CGM.getCodeGenOpts().SanitizeDebugTrapReasons && TrapLocation) { + TrapLocation = getDebugInfo()->CreateTrapFailureMessageFor( + TrapLocation, "Undefined Behavior Sanitizer", TrapMessage); + } + NoMerge = NoMerge || !CGM.getCodeGenOpts().OptimizationLevel || (CurCodeDecl && CurCodeDecl->hasAttr<OptimizeNoneAttr>()); @@ -4059,8 +4080,8 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked, auto Call = TrapBB->begin(); assert(isa<llvm::CallInst>(Call) && "Expected call in trap BB"); - Call->applyMergedLocation(Call->getDebugLoc(), - Builder.getCurrentDebugLocation()); + Call->applyMergedLocation(Call->getDebugLoc(), TrapLocation); + Builder.CreateCondBr(Checked, Cont, TrapBB, MDHelper.createLikelyBranchWeights()); } else { @@ -4069,6 +4090,8 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked, MDHelper.createLikelyBranchWeights()); EmitBlock(TrapBB); + ApplyDebugLocation applyTrapDI(*this, TrapLocation); + llvm::CallInst *TrapCall = Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::ubsantrap), llvm::ConstantInt::get(CGM.Int8Ty, CheckHandlerID)); diff --git a/clang/lib/CodeGen/CGExprCXX.cpp b/clang/lib/CodeGen/CGExprCXX.cpp index 359e30c..b8238a4 100644 --- a/clang/lib/CodeGen/CGExprCXX.cpp +++ b/clang/lib/CodeGen/CGExprCXX.cpp @@ -2146,30 +2146,9 @@ void CodeGenFunction::EmitCXXDeleteExpr(const CXXDeleteExpr *E) { return; } - // We might be deleting a pointer to array. If so, GEP down to the - // first non-array element. - // (this assumes that A(*)[3][7] is converted to [3 x [7 x %A]]*) - if (DeleteTy->isConstantArrayType()) { - llvm::Value *Zero = Builder.getInt32(0); - SmallVector<llvm::Value*,8> GEP; - - GEP.push_back(Zero); // point at the outermost array - - // For each layer of array type we're pointing at: - while (const ConstantArrayType *Arr - = getContext().getAsConstantArrayType(DeleteTy)) { - // 1. Unpeel the array type. - DeleteTy = Arr->getElementType(); - - // 2. GEP to the first element of the array. - GEP.push_back(Zero); - } - - Ptr = Builder.CreateInBoundsGEP(Ptr, GEP, ConvertTypeForMem(DeleteTy), - Ptr.getAlignment(), "del.first"); - } - - assert(ConvertTypeForMem(DeleteTy) == Ptr.getElementType()); + // We might be deleting a pointer to array. + DeleteTy = getContext().getBaseElementType(DeleteTy); + Ptr = Ptr.withElementType(ConvertTypeForMem(DeleteTy)); if (E->isArrayForm()) { EmitArrayDelete(*this, E, Ptr, DeleteTy); diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp index f1698a0..91237cf 100644 --- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp +++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp @@ -7108,8 +7108,8 @@ private: Address BP, Address LB, bool IsNonContiguous, uint64_t DimSize) : CGF(CGF), CombinedInfo(CombinedInfo), Flags(Flags), MapDecl(MapDecl), - MapExpr(MapExpr), BP(BP), LB(LB), IsNonContiguous(IsNonContiguous), - DimSize(DimSize) {} + MapExpr(MapExpr), BP(BP), IsNonContiguous(IsNonContiguous), + DimSize(DimSize), LB(LB) {} void processField( const OMPClauseMappableExprCommon::MappableComponent &MC, diff --git a/clang/lib/CodeGen/SanitizerHandler.h b/clang/lib/CodeGen/SanitizerHandler.h index bb42e39..a66e7ab 100644 --- a/clang/lib/CodeGen/SanitizerHandler.h +++ b/clang/lib/CodeGen/SanitizerHandler.h @@ -14,35 +14,69 @@ #define LLVM_CLANG_LIB_CODEGEN_SANITIZER_HANDLER_H #define LIST_SANITIZER_CHECKS \ - SANITIZER_CHECK(AddOverflow, add_overflow, 0) \ - SANITIZER_CHECK(BuiltinUnreachable, builtin_unreachable, 0) \ - SANITIZER_CHECK(CFICheckFail, cfi_check_fail, 0) \ - SANITIZER_CHECK(DivremOverflow, divrem_overflow, 0) \ - SANITIZER_CHECK(DynamicTypeCacheMiss, dynamic_type_cache_miss, 0) \ - SANITIZER_CHECK(FloatCastOverflow, float_cast_overflow, 0) \ - SANITIZER_CHECK(FunctionTypeMismatch, function_type_mismatch, 0) \ - SANITIZER_CHECK(ImplicitConversion, implicit_conversion, 0) \ - SANITIZER_CHECK(InvalidBuiltin, invalid_builtin, 0) \ - SANITIZER_CHECK(InvalidObjCCast, invalid_objc_cast, 0) \ - SANITIZER_CHECK(LoadInvalidValue, load_invalid_value, 0) \ - SANITIZER_CHECK(MissingReturn, missing_return, 0) \ - SANITIZER_CHECK(MulOverflow, mul_overflow, 0) \ - SANITIZER_CHECK(NegateOverflow, negate_overflow, 0) \ - SANITIZER_CHECK(NullabilityArg, nullability_arg, 0) \ - SANITIZER_CHECK(NullabilityReturn, nullability_return, 1) \ - SANITIZER_CHECK(NonnullArg, nonnull_arg, 0) \ - SANITIZER_CHECK(NonnullReturn, nonnull_return, 1) \ - SANITIZER_CHECK(OutOfBounds, out_of_bounds, 0) \ - SANITIZER_CHECK(PointerOverflow, pointer_overflow, 0) \ - SANITIZER_CHECK(ShiftOutOfBounds, shift_out_of_bounds, 0) \ - SANITIZER_CHECK(SubOverflow, sub_overflow, 0) \ - SANITIZER_CHECK(TypeMismatch, type_mismatch, 1) \ - SANITIZER_CHECK(AlignmentAssumption, alignment_assumption, 0) \ - SANITIZER_CHECK(VLABoundNotPositive, vla_bound_not_positive, 0) \ - SANITIZER_CHECK(BoundsSafety, bounds_safety, 0) + SANITIZER_CHECK(AddOverflow, add_overflow, 0, "Integer addition overflowed") \ + SANITIZER_CHECK(BuiltinUnreachable, builtin_unreachable, 0, \ + "_builtin_unreachable(), execution reached an unreachable " \ + "program point") \ + SANITIZER_CHECK(CFICheckFail, cfi_check_fail, 0, \ + "Control flow integrity check failed") \ + SANITIZER_CHECK(DivremOverflow, divrem_overflow, 0, \ + "Integer divide or remainder overflowed") \ + SANITIZER_CHECK(DynamicTypeCacheMiss, dynamic_type_cache_miss, 0, \ + "Dynamic type cache miss, member call made on an object " \ + "whose dynamic type differs from the expected type") \ + SANITIZER_CHECK(FloatCastOverflow, float_cast_overflow, 0, \ + "Floating-point to integer conversion overflowed") \ + SANITIZER_CHECK(FunctionTypeMismatch, function_type_mismatch, 0, \ + "Function called with mismatched signature") \ + SANITIZER_CHECK(ImplicitConversion, implicit_conversion, 0, \ + "Implicit integer conversion overflowed or lost data") \ + SANITIZER_CHECK(InvalidBuiltin, invalid_builtin, 0, \ + "Invalid use of builtin function") \ + SANITIZER_CHECK(InvalidObjCCast, invalid_objc_cast, 0, \ + "Invalid Objective-C cast") \ + SANITIZER_CHECK(LoadInvalidValue, load_invalid_value, 0, \ + "Loaded an invalid or uninitialized value for the type") \ + SANITIZER_CHECK(MissingReturn, missing_return, 0, \ + "Execution reached the end of a value-returning function " \ + "without returning a value") \ + SANITIZER_CHECK(MulOverflow, mul_overflow, 0, \ + "Integer multiplication overflowed") \ + SANITIZER_CHECK(NegateOverflow, negate_overflow, 0, \ + "Integer negation overflowed") \ + SANITIZER_CHECK( \ + NullabilityArg, nullability_arg, 0, \ + "Passing null as an argument which is annotated with _Nonnull") \ + SANITIZER_CHECK(NullabilityReturn, nullability_return, 1, \ + "Returning null from a function with a return type " \ + "annotated with _Nonnull") \ + SANITIZER_CHECK(NonnullArg, nonnull_arg, 0, \ + "Passing null pointer as an argument which is declared to " \ + "never be null") \ + SANITIZER_CHECK(NonnullReturn, nonnull_return, 1, \ + "Returning null pointer from a function which is declared " \ + "to never return null") \ + SANITIZER_CHECK(OutOfBounds, out_of_bounds, 0, "Array index out of bounds") \ + SANITIZER_CHECK(PointerOverflow, pointer_overflow, 0, \ + "Pointer arithmetic overflowed bounds") \ + SANITIZER_CHECK(ShiftOutOfBounds, shift_out_of_bounds, 0, \ + "Shift exponent is too large for the type") \ + SANITIZER_CHECK(SubOverflow, sub_overflow, 0, \ + "Integer subtraction overflowed") \ + SANITIZER_CHECK(TypeMismatch, type_mismatch, 1, \ + "Type mismatch in operation") \ + SANITIZER_CHECK(AlignmentAssumption, alignment_assumption, 0, \ + "Alignment assumption violated") \ + SANITIZER_CHECK( \ + VLABoundNotPositive, vla_bound_not_positive, 0, \ + "Variable length array bound evaluates to non-positive value") \ + SANITIZER_CHECK(BoundsSafety, bounds_safety, 0, \ + "") // BoundsSafety Msg is empty because it is not considered + // part of UBSan; therefore, no trap reason is emitted for + // this case. enum SanitizerHandler { -#define SANITIZER_CHECK(Enum, Name, Version) Enum, +#define SANITIZER_CHECK(Enum, Name, Version, Msg) Enum, LIST_SANITIZER_CHECKS #undef SANITIZER_CHECK }; diff --git a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp index e469a01..70f510a 100644 --- a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp @@ -633,6 +633,41 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Function *F = CGM.getIntrinsic(IID, {LoadTy}); return Builder.CreateCall(F, {Addr}); } + case AMDGPU::BI__builtin_amdgcn_global_load_monitor_b32: + case AMDGPU::BI__builtin_amdgcn_global_load_monitor_b64: + case AMDGPU::BI__builtin_amdgcn_global_load_monitor_b128: + case AMDGPU::BI__builtin_amdgcn_flat_load_monitor_b32: + case AMDGPU::BI__builtin_amdgcn_flat_load_monitor_b64: + case AMDGPU::BI__builtin_amdgcn_flat_load_monitor_b128: { + + Intrinsic::ID IID; + switch (BuiltinID) { + case AMDGPU::BI__builtin_amdgcn_global_load_monitor_b32: + IID = Intrinsic::amdgcn_global_load_monitor_b32; + break; + case AMDGPU::BI__builtin_amdgcn_global_load_monitor_b64: + IID = Intrinsic::amdgcn_global_load_monitor_b64; + break; + case AMDGPU::BI__builtin_amdgcn_global_load_monitor_b128: + IID = Intrinsic::amdgcn_global_load_monitor_b128; + break; + case AMDGPU::BI__builtin_amdgcn_flat_load_monitor_b32: + IID = Intrinsic::amdgcn_flat_load_monitor_b32; + break; + case AMDGPU::BI__builtin_amdgcn_flat_load_monitor_b64: + IID = Intrinsic::amdgcn_flat_load_monitor_b64; + break; + case AMDGPU::BI__builtin_amdgcn_flat_load_monitor_b128: + IID = Intrinsic::amdgcn_flat_load_monitor_b128; + break; + } + + llvm::Type *LoadTy = ConvertType(E->getType()); + llvm::Value *Addr = EmitScalarExpr(E->getArg(0)); + llvm::Value *Val = EmitScalarExpr(E->getArg(1)); + llvm::Function *F = CGM.getIntrinsic(IID, {LoadTy}); + return Builder.CreateCall(F, {Addr, Val}); + } case AMDGPU::BI__builtin_amdgcn_load_to_lds: { // Should this have asan instrumentation? return emitBuiltinWithOneOverloadedType<5>(*this, E, diff --git a/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp b/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp index b7fd70e..33a8d8f 100644 --- a/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp @@ -12,7 +12,10 @@ #include "CGBuiltin.h" #include "clang/Basic/TargetBuiltins.h" +#include "llvm/ADT/APInt.h" +#include "llvm/IR/Constants.h" #include "llvm/IR/IntrinsicsWebAssembly.h" +#include "llvm/Support/ErrorHandling.h" using namespace clang; using namespace CodeGen; @@ -218,6 +221,64 @@ Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID, Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_ref_null_func); return Builder.CreateCall(Callee); } + case WebAssembly::BI__builtin_wasm_test_function_pointer_signature: { + Value *FuncRef = EmitScalarExpr(E->getArg(0)); + + // Get the function type from the argument's static type + QualType ArgType = E->getArg(0)->getType(); + const PointerType *PtrTy = ArgType->getAs<PointerType>(); + assert(PtrTy && "Sema should have ensured this is a function pointer"); + + const FunctionType *FuncTy = PtrTy->getPointeeType()->getAs<FunctionType>(); + assert(FuncTy && "Sema should have ensured this is a function pointer"); + + // In the llvm IR, we won't have access any more to the type of the function + // pointer so we need to insert this type information somehow. The + // @llvm.wasm.ref.test.func takes varargs arguments whose values are unused + // to indicate the type of the function to test for. See the test here: + // llvm/test/CodeGen/WebAssembly/ref-test-func.ll + // + // The format is: first we include the return types (since this is a C + // function pointer, there will be 0 or one of these) then a token type to + // indicate the boundary between return types and param types, then the + // param types. + + llvm::FunctionType *LLVMFuncTy = + cast<llvm::FunctionType>(ConvertType(QualType(FuncTy, 0))); + + unsigned NParams = LLVMFuncTy->getNumParams(); + std::vector<Value *> Args; + Args.reserve(NParams + 3); + // The only real argument is the FuncRef + Args.push_back(FuncRef); + + // Add the type information + auto addType = [this, &Args](llvm::Type *T) { + if (T->isVoidTy()) { + // Do nothing + } else if (T->isFloatingPointTy()) { + Args.push_back(ConstantFP::get(T, 0)); + } else if (T->isIntegerTy()) { + Args.push_back(ConstantInt::get(T, 0)); + } else if (T->isPointerTy()) { + Args.push_back(ConstantPointerNull::get(llvm::PointerType::get( + getLLVMContext(), T->getPointerAddressSpace()))); + } else { + // TODO: Handle reference types. For now, we reject them in Sema. + llvm_unreachable("Unhandled type"); + } + }; + + addType(LLVMFuncTy->getReturnType()); + // The token type indicates the boundary between return types and param + // types. + Args.push_back(PoisonValue::get(llvm::Type::getTokenTy(getLLVMContext()))); + for (unsigned i = 0; i < NParams; i++) { + addType(LLVMFuncTy->getParamType(i)); + } + Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_ref_test_func); + return Builder.CreateCall(Callee, Args); + } case WebAssembly::BI__builtin_wasm_swizzle_i8x16: { Value *Src = EmitScalarExpr(E->getArg(0)); Value *Indices = EmitScalarExpr(E->getArg(1)); diff --git a/clang/lib/Driver/Driver.cpp b/clang/lib/Driver/Driver.cpp index 892049e..853f694 100644 --- a/clang/lib/Driver/Driver.cpp +++ b/clang/lib/Driver/Driver.cpp @@ -209,8 +209,8 @@ Driver::Driver(StringRef ClangExecutable, StringRef TargetTriple, CCLogDiagnostics(false), CCGenDiagnostics(false), CCPrintProcessStats(false), CCPrintInternalStats(false), TargetTriple(TargetTriple), Saver(Alloc), PrependArg(nullptr), - CheckInputsExist(true), ProbePrecompiled(true), - SuppressMissingInputWarning(false) { + PreferredLinker(CLANG_DEFAULT_LINKER), CheckInputsExist(true), + ProbePrecompiled(true), SuppressMissingInputWarning(false) { // Provide a sane fallback if no VFS is specified. if (!this->VFS) this->VFS = llvm::vfs::getRealFileSystem(); diff --git a/clang/lib/Driver/SanitizerArgs.cpp b/clang/lib/Driver/SanitizerArgs.cpp index 21e4cff..98793a5 100644 --- a/clang/lib/Driver/SanitizerArgs.cpp +++ b/clang/lib/Driver/SanitizerArgs.cpp @@ -1382,6 +1382,12 @@ void SanitizerArgs::addArgs(const ToolChain &TC, const llvm::opt::ArgList &Args, CmdArgs.push_back(Args.MakeArgString("-fsanitize-annotate-debug-info=" + toString(AnnotateDebugInfo))); + if (const Arg *A = + Args.getLastArg(options::OPT_fsanitize_debug_trap_reasons, + options::OPT_fno_sanitize_debug_trap_reasons)) { + CmdArgs.push_back(Args.MakeArgString(A->getAsString(Args))); + } + addSpecialCaseListOpt(Args, CmdArgs, "-fsanitize-ignorelist=", UserIgnorelistFiles); addSpecialCaseListOpt(Args, CmdArgs, diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp index 47f93fa1..1d7dad0 100644 --- a/clang/lib/Driver/ToolChain.cpp +++ b/clang/lib/Driver/ToolChain.cpp @@ -1087,7 +1087,7 @@ std::string ToolChain::GetLinkerPath(bool *LinkerIsLLD) const { // Get -fuse-ld= first to prevent -Wunused-command-line-argument. -fuse-ld= is // considered as the linker flavor, e.g. "bfd", "gold", or "lld". const Arg* A = Args.getLastArg(options::OPT_fuse_ld_EQ); - StringRef UseLinker = A ? A->getValue() : CLANG_DEFAULT_LINKER; + StringRef UseLinker = A ? A->getValue() : getDriver().getPreferredLinker(); // --ld-path= takes precedence over -fuse-ld= and specifies the executable // name. -B, COMPILER_PATH and PATH and consulted if the value does not diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp index 7d0c142..9d882db 100644 --- a/clang/lib/Driver/ToolChains/Clang.cpp +++ b/clang/lib/Driver/ToolChains/Clang.cpp @@ -3881,17 +3881,17 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D, const ArgList &Args, const InputInfo &Input, const InputInfo &Output, bool HaveStd20, ArgStringList &CmdArgs) { - bool IsCXX = types::isCXX(Input.getType()); - bool HaveStdCXXModules = IsCXX && HaveStd20; + const bool IsCXX = types::isCXX(Input.getType()); + const bool HaveStdCXXModules = IsCXX && HaveStd20; bool HaveModules = HaveStdCXXModules; // -fmodules enables the use of precompiled modules (off by default). // Users can pass -fno-cxx-modules to turn off modules support for // C++/Objective-C++ programs. + const bool AllowedInCXX = Args.hasFlag(options::OPT_fcxx_modules, + options::OPT_fno_cxx_modules, true); bool HaveClangModules = false; if (Args.hasFlag(options::OPT_fmodules, options::OPT_fno_modules, false)) { - bool AllowedInCXX = Args.hasFlag(options::OPT_fcxx_modules, - options::OPT_fno_cxx_modules, true); if (AllowedInCXX || !IsCXX) { CmdArgs.push_back("-fmodules"); HaveClangModules = true; @@ -3900,6 +3900,9 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D, HaveModules |= HaveClangModules; + if (HaveModules && !AllowedInCXX) + CmdArgs.push_back("-fno-cxx-modules"); + // -fmodule-maps enables implicit reading of module map files. By default, // this is enabled if we are using Clang's flavor of precompiled modules. if (Args.hasFlag(options::OPT_fimplicit_module_maps, diff --git a/clang/lib/Driver/ToolChains/MSVC.cpp b/clang/lib/Driver/ToolChains/MSVC.cpp index 7d31eea..bb469ff 100644 --- a/clang/lib/Driver/ToolChains/MSVC.cpp +++ b/clang/lib/Driver/ToolChains/MSVC.cpp @@ -279,8 +279,8 @@ void visualstudio::Linker::ConstructJob(Compilation &C, const JobAction &JA, AddRunTimeLibs(TC, TC.getDriver(), CmdArgs, Args); } - StringRef Linker = - Args.getLastArgValue(options::OPT_fuse_ld_EQ, CLANG_DEFAULT_LINKER); + StringRef Linker = Args.getLastArgValue(options::OPT_fuse_ld_EQ, + TC.getDriver().getPreferredLinker()); if (Linker.empty()) Linker = "link"; // We need to translate 'lld' into 'lld-link'. diff --git a/clang/lib/Driver/ToolChains/MinGW.cpp b/clang/lib/Driver/ToolChains/MinGW.cpp index b2e36ae..6abd0c0 100644 --- a/clang/lib/Driver/ToolChains/MinGW.cpp +++ b/clang/lib/Driver/ToolChains/MinGW.cpp @@ -548,7 +548,7 @@ toolchains::MinGW::MinGW(const Driver &D, const llvm::Triple &Triple, getFilePaths().push_back(Base + "lib"); NativeLLVMSupport = - Args.getLastArgValue(options::OPT_fuse_ld_EQ, CLANG_DEFAULT_LINKER) + Args.getLastArgValue(options::OPT_fuse_ld_EQ, D.getPreferredLinker()) .equals_insensitive("lld"); } diff --git a/clang/lib/Driver/ToolChains/Solaris.cpp b/clang/lib/Driver/ToolChains/Solaris.cpp index a3574e1..02aa598 100644 --- a/clang/lib/Driver/ToolChains/Solaris.cpp +++ b/clang/lib/Driver/ToolChains/Solaris.cpp @@ -39,7 +39,7 @@ void solaris::Assembler::ConstructJob(Compilation &C, const JobAction &JA, bool solaris::isLinkerGnuLd(const ToolChain &TC, const ArgList &Args) { // Only used if targetting Solaris. const Arg *A = Args.getLastArg(options::OPT_fuse_ld_EQ); - StringRef UseLinker = A ? A->getValue() : CLANG_DEFAULT_LINKER; + StringRef UseLinker = A ? A->getValue() : TC.getDriver().getPreferredLinker(); return UseLinker == "bfd" || UseLinker == "gld"; } @@ -52,7 +52,7 @@ static bool getPIE(const ArgList &Args, const ToolChain &TC) { TC.isPIEDefault(Args)); } -// FIXME: Need to handle CLANG_DEFAULT_LINKER here? +// FIXME: Need to handle PreferredLinker here? std::string solaris::Linker::getLinkerPath(const ArgList &Args) const { const ToolChain &ToolChain = getToolChain(); if (const Arg *A = Args.getLastArg(options::OPT_fuse_ld_EQ)) { @@ -345,7 +345,7 @@ SanitizerMask Solaris::getSupportedSanitizers() const { const char *Solaris::getDefaultLinker() const { // FIXME: Only handle Solaris ld and GNU ld here. - return llvm::StringSwitch<const char *>(CLANG_DEFAULT_LINKER) + return llvm::StringSwitch<const char *>(getDriver().getPreferredLinker()) .Cases("bfd", "gld", "/usr/gnu/bin/ld") .Default("/usr/bin/ld"); } diff --git a/clang/lib/Driver/ToolChains/UEFI.cpp b/clang/lib/Driver/ToolChains/UEFI.cpp index ac6668e..2b41173 100644 --- a/clang/lib/Driver/ToolChains/UEFI.cpp +++ b/clang/lib/Driver/ToolChains/UEFI.cpp @@ -83,8 +83,8 @@ void tools::uefi::Linker::ConstructJob(Compilation &C, const JobAction &JA, // This should ideally be handled by ToolChain::GetLinkerPath but we need // to special case some linker paths. In the case of lld, we need to // translate 'lld' into 'lld-link'. - StringRef Linker = - Args.getLastArgValue(options::OPT_fuse_ld_EQ, CLANG_DEFAULT_LINKER); + StringRef Linker = Args.getLastArgValue(options::OPT_fuse_ld_EQ, + TC.getDriver().getPreferredLinker()); if (Linker.empty() || Linker == "lld") Linker = "lld-link"; diff --git a/clang/lib/Format/Format.cpp b/clang/lib/Format/Format.cpp index 62feb3d..0637807 100644 --- a/clang/lib/Format/Format.cpp +++ b/clang/lib/Format/Format.cpp @@ -731,6 +731,7 @@ template <> struct MappingTraits<FormatStyle::SpaceBeforeParensCustom> { IO.mapOptional("AfterFunctionDeclarationName", Spacing.AfterFunctionDeclarationName); IO.mapOptional("AfterIfMacros", Spacing.AfterIfMacros); + IO.mapOptional("AfterNot", Spacing.AfterNot); IO.mapOptional("AfterOverloadedOperator", Spacing.AfterOverloadedOperator); IO.mapOptional("AfterPlacementOperator", Spacing.AfterPlacementOperator); IO.mapOptional("AfterRequiresInClause", Spacing.AfterRequiresInClause); @@ -1753,7 +1754,6 @@ FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) { GoogleStyle.AttributeMacros.push_back("absl_nullable"); GoogleStyle.AttributeMacros.push_back("absl_nullability_unknown"); GoogleStyle.BreakTemplateDeclarations = FormatStyle::BTDS_Yes; - GoogleStyle.DerivePointerAlignment = true; GoogleStyle.IncludeStyle.IncludeBlocks = tooling::IncludeStyle::IBS_Regroup; GoogleStyle.IncludeStyle.IncludeCategories = {{"^<ext/.*\\.h>", 2, 0, false}, {"^<.*\\.h>", 1, 0, false}, @@ -1862,6 +1862,7 @@ FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) { } else if (Language == FormatStyle::LK_ObjC) { GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; GoogleStyle.ColumnLimit = 100; + GoogleStyle.DerivePointerAlignment = true; // "Regroup" doesn't work well for ObjC yet (main header heuristic, // relationship between ObjC standard library headers and other heades, // #imports, etc.) @@ -2638,31 +2639,44 @@ private: int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) { int AlignmentDiff = 0; + for (const AnnotatedLine *Line : Lines) { AlignmentDiff += countVariableAlignments(Line->Children); - for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) { + + for (const auto *Tok = Line->getFirstNonComment(); Tok; Tok = Tok->Next) { if (Tok->isNot(TT_PointerOrReference)) continue; - // Don't treat space in `void foo() &&` as evidence. - if (const auto *Prev = Tok->getPreviousNonComment()) { - if (Prev->is(tok::r_paren) && Prev->MatchingParen) { - if (const auto *Func = - Prev->MatchingParen->getPreviousNonComment()) { - if (Func->isOneOf(TT_FunctionDeclarationName, TT_StartOfName, - TT_OverloadedOperator)) { - continue; - } - } - } + + const auto *Prev = Tok->Previous; + const bool PrecededByName = Prev && Prev->Tok.getIdentifierInfo(); + const bool SpaceBefore = Tok->hasWhitespaceBefore(); + + // e.g. `int **`, `int*&`, etc. + while (Tok->Next && Tok->Next->is(TT_PointerOrReference)) + Tok = Tok->Next; + + const auto *Next = Tok->Next; + const bool FollowedByName = Next && Next->Tok.getIdentifierInfo(); + const bool SpaceAfter = Next && Next->hasWhitespaceBefore(); + + if ((!PrecededByName && !FollowedByName) || + // e.g. `int * i` or `int*i` + (PrecededByName && FollowedByName && SpaceBefore == SpaceAfter)) { + continue; } - bool SpaceBefore = Tok->hasWhitespaceBefore(); - bool SpaceAfter = Tok->Next->hasWhitespaceBefore(); - if (SpaceBefore && !SpaceAfter) + + if ((PrecededByName && SpaceBefore) || + (FollowedByName && !SpaceAfter)) { + // Right alignment. ++AlignmentDiff; - if (!SpaceBefore && SpaceAfter) + } else if ((PrecededByName && !SpaceBefore) || + (FollowedByName && SpaceAfter)) { + // Left alignment. --AlignmentDiff; + } } } + return AlignmentDiff; } diff --git a/clang/lib/Format/TokenAnnotator.cpp b/clang/lib/Format/TokenAnnotator.cpp index 581bfba..d28d2fd 100644 --- a/clang/lib/Format/TokenAnnotator.cpp +++ b/clang/lib/Format/TokenAnnotator.cpp @@ -5478,7 +5478,8 @@ bool TokenAnnotator::spaceRequiredBefore(const AnnotatedLine &Line, if (Left.TokenText == "!") return Style.SpaceAfterLogicalNot; assert(Left.TokenText == "not"); - return Right.isOneOf(tok::coloncolon, TT_UnaryOperator); + return Right.isOneOf(tok::coloncolon, TT_UnaryOperator) || + (Right.is(tok::l_paren) && Style.SpaceBeforeParensOptions.AfterNot); } // If the next token is a binary operator or a selector name, we have diff --git a/clang/lib/Parse/ParseDeclCXX.cpp b/clang/lib/Parse/ParseDeclCXX.cpp index 31392d1d..bc8841c 100644 --- a/clang/lib/Parse/ParseDeclCXX.cpp +++ b/clang/lib/Parse/ParseDeclCXX.cpp @@ -4940,9 +4940,8 @@ void Parser::ParseHLSLRootSignatureAttributeArgs(ParsedAttributes &Attrs) { // signature string and construct the in-memory elements if (!Found) { // Invoke the root signature parser to construct the in-memory constructs - SmallVector<hlsl::RootSignatureElement> RootElements; - hlsl::RootSignatureParser Parser(getLangOpts().HLSLRootSigVer, RootElements, - Signature, PP); + hlsl::RootSignatureParser Parser(getLangOpts().HLSLRootSigVer, Signature, + PP); if (Parser.parse()) { T.consumeClose(); return; @@ -4950,7 +4949,7 @@ void Parser::ParseHLSLRootSignatureAttributeArgs(ParsedAttributes &Attrs) { // Construct the declaration. Actions.HLSL().ActOnFinishRootSignatureDecl(RootSignatureLoc, DeclIdent, - RootElements); + Parser.getElements()); } // Create the arg for the ParsedAttr diff --git a/clang/lib/Parse/ParseHLSLRootSignature.cpp b/clang/lib/Parse/ParseHLSLRootSignature.cpp index db9ed83..98dc458 100644 --- a/clang/lib/Parse/ParseHLSLRootSignature.cpp +++ b/clang/lib/Parse/ParseHLSLRootSignature.cpp @@ -27,11 +27,10 @@ static const TokenKind RootElementKeywords[] = { }; RootSignatureParser::RootSignatureParser( - llvm::dxbc::RootSignatureVersion Version, - SmallVector<RootSignatureElement> &Elements, StringLiteral *Signature, + llvm::dxbc::RootSignatureVersion Version, StringLiteral *Signature, Preprocessor &PP) - : Version(Version), Elements(Elements), Signature(Signature), - Lexer(Signature->getString()), PP(PP), CurToken(0) {} + : Version(Version), Signature(Signature), Lexer(Signature->getString()), + PP(PP), CurToken(0) {} bool RootSignatureParser::parse() { // Iterate as many RootSignatureElements as possible, until we hit the diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp index 56608e9..d50eeff 100644 --- a/clang/lib/Sema/Sema.cpp +++ b/clang/lib/Sema/Sema.cpp @@ -1616,6 +1616,8 @@ void Sema::ActOnEndOfTranslationUnit() { if (!PP.isIncrementalProcessingEnabled()) TUScope = nullptr; + + checkExposure(Context.getTranslationUnitDecl()); } diff --git a/clang/lib/Sema/SemaModule.cpp b/clang/lib/Sema/SemaModule.cpp index 7c982bc..98ebd70 100644 --- a/clang/lib/Sema/SemaModule.cpp +++ b/clang/lib/Sema/SemaModule.cpp @@ -13,6 +13,7 @@ #include "clang/AST/ASTConsumer.h" #include "clang/AST/ASTMutationListener.h" +#include "clang/AST/RecursiveASTVisitor.h" #include "clang/Lex/HeaderSearch.h" #include "clang/Lex/Preprocessor.h" #include "clang/Sema/ParsedAttr.h" @@ -485,6 +486,7 @@ Sema::ActOnModuleDecl(SourceLocation StartLoc, SourceLocation ModuleLoc, // implementation unit importing its interface). Make this module visible // and return the import decl to be added to the current TU. if (Interface) { + HadImportedNamedModules = true; makeTransitiveImportsVisible(getASTContext(), VisibleModules, Interface, Mod, ModuleLoc, @@ -728,6 +730,8 @@ DeclResult Sema::ActOnModuleImport(SourceLocation StartLoc, getCurrentModule()->Imports.insert(Mod); } + HadImportedNamedModules = true; + return Import; } @@ -1102,3 +1106,471 @@ bool Sema::isCurrentModulePurview() const { return false; } } + +//===----------------------------------------------------------------------===// +// Checking Exposure in modules // +//===----------------------------------------------------------------------===// + +namespace { +class ExposureChecker { +public: + ExposureChecker(Sema &S) : SemaRef(S) {} + + bool checkExposure(const VarDecl *D, bool Diag); + bool checkExposure(const CXXRecordDecl *D, bool Diag); + bool checkExposure(const Stmt *S, bool Diag); + bool checkExposure(const FunctionDecl *D, bool Diag); + bool checkExposure(const NamedDecl *D, bool Diag); + void checkExposureInContext(const DeclContext *DC); + bool isExposureCandidate(const NamedDecl *D); + + bool isTULocal(QualType Ty); + bool isTULocal(const NamedDecl *ND); + bool isTULocal(const Expr *E); + + Sema &SemaRef; + +private: + llvm::DenseSet<const NamedDecl *> ExposureSet; + llvm::DenseSet<const NamedDecl *> KnownNonExposureSet; +}; + +bool ExposureChecker::isTULocal(QualType Ty) { + // [basic.link]p15: + // An entity is TU-local if it is + // - a type, type alias, namespace, namespace alias, function, variable, or + // template that + // -- has internal linkage, or + return Ty->getLinkage() == Linkage::Internal; + + // TODO: + // [basic.link]p15.2: + // a type with no name that is defined outside a class-specifier, function + // body, or initializer or is introduced by a defining-type-specifier that + // is used to declare only TU-local entities, +} + +bool ExposureChecker::isTULocal(const NamedDecl *D) { + if (!D) + return false; + + // [basic.link]p15: + // An entity is TU-local if it is + // - a type, type alias, namespace, namespace alias, function, variable, or + // template that + // -- has internal linkage, or + if (D->getLinkageInternal() == Linkage::Internal) + return true; + + if (D->isInAnonymousNamespace()) + return true; + + // [basic.link]p15.1.2: + // does not have a name with linkage and is declared, or introduced by a + // lambda-expression, within the definition of a TU-local entity, + if (D->getLinkageInternal() == Linkage::None) + if (auto *ND = dyn_cast<NamedDecl>(D->getDeclContext()); + ND && isTULocal(ND)) + return true; + + // [basic.link]p15.3, p15.4: + // - a specialization of a TU-local template, + // - a specialization of a template with any TU-local template argument, or + ArrayRef<TemplateArgument> TemplateArgs; + NamedDecl *PrimaryTemplate = nullptr; + if (auto *CTSD = dyn_cast<ClassTemplateSpecializationDecl>(D)) { + TemplateArgs = CTSD->getTemplateArgs().asArray(); + PrimaryTemplate = CTSD->getSpecializedTemplate(); + if (isTULocal(PrimaryTemplate)) + return true; + } else if (auto *VTSD = dyn_cast<VarTemplateSpecializationDecl>(D)) { + TemplateArgs = VTSD->getTemplateArgs().asArray(); + PrimaryTemplate = VTSD->getSpecializedTemplate(); + if (isTULocal(PrimaryTemplate)) + return true; + } else if (auto *FD = dyn_cast<FunctionDecl>(D)) { + if (auto *TAList = FD->getTemplateSpecializationArgs()) + TemplateArgs = TAList->asArray(); + + PrimaryTemplate = FD->getPrimaryTemplate(); + if (isTULocal(PrimaryTemplate)) + return true; + } + + if (!PrimaryTemplate) + // Following off, we only check for specializations. + return false; + + if (KnownNonExposureSet.count(D)) + return false; + + for (auto &TA : TemplateArgs) { + switch (TA.getKind()) { + case TemplateArgument::Type: + if (isTULocal(TA.getAsType())) + return true; + break; + case TemplateArgument::Declaration: + if (isTULocal(TA.getAsDecl())) + return true; + break; + default: + break; + } + } + + // [basic.link]p15.5 + // - a specialization of a template whose (possibly instantiated) declaration + // is an exposure. + if (checkExposure(PrimaryTemplate, /*Diag=*/false)) + return true; + + // Avoid calling checkExposure again since it is expensive. + KnownNonExposureSet.insert(D); + return false; +} + +bool ExposureChecker::isTULocal(const Expr *E) { + if (!E) + return false; + + // [basic.link]p16: + // A value or object is TU-local if either + // - it is of TU-local type, + if (isTULocal(E->getType())) + return true; + + E = E->IgnoreParenImpCasts(); + // [basic.link]p16.2: + // - it is, or is a pointer to, a TU-local function or the object associated + // with a TU-local variable, + // - it is an object of class or array type and any of its subobjects or any + // of the objects or functions to which its non-static data members of + // reference type refer is TU-local and is usable in constant expressions, or + // FIXME: But how can we know the value of pointers or arrays at compile time? + if (const auto *DRE = dyn_cast<DeclRefExpr>(E)) { + if (auto *FD = dyn_cast_or_null<FunctionDecl>(DRE->getFoundDecl())) + return isTULocal(FD); + else if (auto *VD = dyn_cast_or_null<VarDecl>(DRE->getFoundDecl())) + return isTULocal(VD); + else if (auto *RD = dyn_cast_or_null<CXXRecordDecl>(DRE->getFoundDecl())) + return isTULocal(RD); + } + + // TODO: + // [basic.link]p16.4: + // it is a reflection value that represents... + + return false; +} + +bool ExposureChecker::isExposureCandidate(const NamedDecl *D) { + if (!D) + return false; + + // [basic.link]p17: + // If a (possibly instantiated) declaration of, or a deduction guide for, + // a non-TU-local entity in a module interface unit + // (outside the private-module-fragment, if any) or + // module partition is an exposure, the program is ill-formed. + Module *M = D->getOwningModule(); + if (!M || !M->isInterfaceOrPartition()) + return false; + + if (D->isImplicit()) + return false; + + // [basic.link]p14: + // A declaration is an exposure if it either names a TU-local entity + // (defined below), ignoring: + // ... + // - friend declarations in a class definition + if (D->getFriendObjectKind() && + isa<CXXRecordDecl>(D->getLexicalDeclContext())) + return false; + + return true; +} + +bool ExposureChecker::checkExposure(const NamedDecl *D, bool Diag) { + if (!isExposureCandidate(D)) + return false; + + if (auto *FD = dyn_cast<FunctionDecl>(D)) + return checkExposure(FD, Diag); + if (auto *FTD = dyn_cast<FunctionTemplateDecl>(D)) + return checkExposure(FTD->getTemplatedDecl(), Diag); + + if (auto *VD = dyn_cast<VarDecl>(D)) + return checkExposure(VD, Diag); + if (auto *VTD = dyn_cast<VarTemplateDecl>(D)) + return checkExposure(VTD->getTemplatedDecl(), Diag); + + if (auto *RD = dyn_cast<CXXRecordDecl>(D)) + return checkExposure(RD, Diag); + + if (auto *CTD = dyn_cast<ClassTemplateDecl>(D)) + return checkExposure(CTD->getTemplatedDecl(), Diag); + + return false; +} + +bool ExposureChecker::checkExposure(const FunctionDecl *FD, bool Diag) { + bool IsExposure = false; + if (isTULocal(FD->getReturnType())) { + IsExposure = true; + if (Diag) + SemaRef.Diag(FD->getReturnTypeSourceRange().getBegin(), + diag::warn_exposure) + << FD->getReturnType(); + } + + for (ParmVarDecl *Parms : FD->parameters()) + if (isTULocal(Parms->getType())) { + IsExposure = true; + if (Diag) + SemaRef.Diag(Parms->getLocation(), diag::warn_exposure) + << Parms->getType(); + } + + bool IsImplicitInstantiation = + FD->getTemplateSpecializationKind() == TSK_ImplicitInstantiation; + + // [basic.link]p14: + // A declaration is an exposure if it either names a TU-local entity + // (defined below), ignoring: + // - the function-body for a non-inline function or function template + // (but not the deduced return + // type for a (possibly instantiated) definition of a function with a + // declared return type that uses a placeholder type + // ([dcl.spec.auto])), + Diag &= + (FD->isInlined() || IsImplicitInstantiation) && !FD->isDependentContext(); + + IsExposure |= checkExposure(FD->getBody(), Diag); + if (IsExposure) + ExposureSet.insert(FD); + + return IsExposure; +} + +bool ExposureChecker::checkExposure(const VarDecl *VD, bool Diag) { + bool IsExposure = false; + // [basic.link]p14: + // A declaration is an exposure if it either names a TU-local entity (defined + // below), ignoring: + // ... + // or defines a constexpr variable initialized to a TU-local value (defined + // below). + if (VD->isConstexpr() && isTULocal(VD->getInit())) { + IsExposure = true; + if (Diag) + SemaRef.Diag(VD->getInit()->getExprLoc(), diag::warn_exposure) + << VD->getInit(); + } + + if (isTULocal(VD->getType())) { + IsExposure = true; + if (Diag) + SemaRef.Diag(VD->getLocation(), diag::warn_exposure) << VD->getType(); + } + + // [basic.link]p14: + // ..., ignoring: + // - the initializer for a variable or variable template (but not the + // variable's type), + // + // Note: although the spec says to ignore the initializer for all variable, + // for the code we generated now for inline variables, it is dangerous if the + // initializer of an inline variable is TULocal. + Diag &= !VD->getDeclContext()->isDependentContext() && VD->isInline(); + IsExposure |= checkExposure(VD->getInit(), Diag); + if (IsExposure) + ExposureSet.insert(VD); + + return IsExposure; +} + +bool ExposureChecker::checkExposure(const CXXRecordDecl *RD, bool Diag) { + if (!RD->hasDefinition()) + return false; + + bool IsExposure = false; + for (CXXMethodDecl *Method : RD->methods()) + IsExposure |= checkExposure(Method, Diag); + + for (FieldDecl *FD : RD->fields()) { + if (isTULocal(FD->getType())) { + IsExposure = true; + if (Diag) + SemaRef.Diag(FD->getLocation(), diag::warn_exposure) << FD->getType(); + } + } + + for (const CXXBaseSpecifier &Base : RD->bases()) { + if (isTULocal(Base.getType())) { + IsExposure = true; + if (Diag) + SemaRef.Diag(Base.getBaseTypeLoc(), diag::warn_exposure) + << Base.getType(); + } + } + + if (IsExposure) + ExposureSet.insert(RD); + + return IsExposure; +} + +template <typename CallbackTy> +class ReferenceTULocalChecker + : public clang::RecursiveASTVisitor<ReferenceTULocalChecker<CallbackTy>> { +public: + ReferenceTULocalChecker(ExposureChecker &C, CallbackTy &&Callback) + : Checker(C), Callback(std::move(Callback)) {} + + bool VisitDeclRefExpr(DeclRefExpr *DRE) { + ValueDecl *Referenced = DRE->getDecl(); + if (!Referenced) + return true; + + if (!Checker.isTULocal(Referenced)) + // We don't care if the referenced declaration is not TU-local. + return true; + + Qualifiers Qual = DRE->getType().getQualifiers(); + // [basic.link]p14: + // A declaration is an exposure if it either names a TU-local entity + // (defined below), ignoring: + // ... + // - any reference to a non-volatile const object ... + if (Qual.hasConst() && !Qual.hasVolatile()) + return true; + + // [basic.link]p14: + // ..., ignoring: + // ... + // (p14.4) - ... or reference with internal or no linkage initialized with + // a constant expression that is not an odr-use + ASTContext &Context = Referenced->getASTContext(); + Linkage L = Referenced->getLinkageInternal(); + if (DRE->isNonOdrUse() && (L == Linkage::Internal || L == Linkage::None)) + if (auto *VD = dyn_cast<VarDecl>(Referenced); + VD && VD->getInit() && !VD->getInit()->isValueDependent() && + VD->getInit()->isConstantInitializer(Context, /*IsForRef=*/false)) + return true; + + Callback(DRE, Referenced); + return true; + } + + ExposureChecker &Checker; + CallbackTy Callback; +}; + +template <typename CallbackTy> +ReferenceTULocalChecker(ExposureChecker &, CallbackTy &&) + -> ReferenceTULocalChecker<CallbackTy>; + +bool ExposureChecker::checkExposure(const Stmt *S, bool Diag) { + if (!S) + return false; + + bool HasReferencedTULocals = false; + ReferenceTULocalChecker Checker( + *this, [this, &HasReferencedTULocals, Diag](DeclRefExpr *DRE, + ValueDecl *Referenced) { + if (Diag) { + SemaRef.Diag(DRE->getExprLoc(), diag::warn_exposure) << Referenced; + } + HasReferencedTULocals = true; + }); + Checker.TraverseStmt(const_cast<Stmt *>(S)); + return HasReferencedTULocals; +} + +void ExposureChecker::checkExposureInContext(const DeclContext *DC) { + for (auto *TopD : DC->noload_decls()) { + auto *TopND = dyn_cast<NamedDecl>(TopD); + if (!TopND) + continue; + + if (auto *Namespace = dyn_cast<NamespaceDecl>(TopND)) { + checkExposureInContext(Namespace); + continue; + } + + // [basic.link]p17: + // If a (possibly instantiated) declaration of, or a deduction guide for, + // a non-TU-local entity in a module interface unit + // (outside the private-module-fragment, if any) or + // module partition is an exposure, the program is ill-formed. + if (!TopND->isFromASTFile() && isExposureCandidate(TopND) && + !isTULocal(TopND)) + checkExposure(TopND, /*Diag=*/true); + } +} + +} // namespace + +void Sema::checkExposure(const TranslationUnitDecl *TU) { + if (!TU) + return; + + ExposureChecker Checker(*this); + + Module *M = TU->getOwningModule(); + if (M && M->isInterfaceOrPartition()) + Checker.checkExposureInContext(TU); + + // [basic.link]p18: + // If a declaration that appears in one translation unit names a TU-local + // entity declared in another translation unit that is not a header unit, + // the program is ill-formed. + for (auto FDAndInstantiationLocPair : PendingCheckReferenceForTULocal) { + FunctionDecl *FD = FDAndInstantiationLocPair.first; + SourceLocation PointOfInstantiation = FDAndInstantiationLocPair.second; + + if (!FD->hasBody()) + continue; + + ReferenceTULocalChecker(Checker, [&, this](DeclRefExpr *DRE, + ValueDecl *Referenced) { + // A "defect" in current implementation. Now an implicit instantiation of + // a template, the instantiation is considered to be in the same module + // unit as the template instead of the module unit where the instantiation + // happens. + // + // See test/Modules/Exposre-2.cppm for example. + if (!Referenced->isFromASTFile()) + return; + + if (!Referenced->isInAnotherModuleUnit()) + return; + + // This is not standard conforming. But given there are too many static + // (inline) functions in headers in existing code, it is more user + // friendly to ignore them temporarily now. maybe we can have another flag + // for this. + if (Referenced->getOwningModule()->isExplicitGlobalModule() && + isa<FunctionDecl>(Referenced)) + return; + + Diag(PointOfInstantiation, + diag::warn_reference_tu_local_entity_in_other_tu) + << FD << Referenced + << Referenced->getOwningModule()->getTopLevelModuleName(); + }).TraverseStmt(FD->getBody()); + } +} + +void Sema::checkReferenceToTULocalFromOtherTU( + FunctionDecl *FD, SourceLocation PointOfInstantiation) { + // Checking if a declaration have any reference to TU-local entities in other + // TU is expensive. Try to avoid it as much as possible. + if (!FD || !HadImportedNamedModules) + return; + + PendingCheckReferenceForTULocal.push_back( + std::make_pair(FD, PointOfInstantiation)); +} diff --git a/clang/lib/Sema/SemaOpenACC.cpp b/clang/lib/Sema/SemaOpenACC.cpp index 128a5db..8bfea62 100644 --- a/clang/lib/Sema/SemaOpenACC.cpp +++ b/clang/lib/Sema/SemaOpenACC.cpp @@ -699,11 +699,19 @@ ExprResult SemaOpenACC::ActOnVar(OpenACCDirectiveKind DK, OpenACCClauseKind CK, // OpenACC3.3 2.13: // A 'var' in a 'declare' directive must be a variable or array name. if ((CK == OpenACCClauseKind::UseDevice || - DK == OpenACCDirectiveKind::Declare) && - isa<ArraySectionExpr, ArraySubscriptExpr>(CurVarExpr)) { - Diag(VarExpr->getExprLoc(), diag::err_acc_not_a_var_ref_use_device_declare) - << (DK == OpenACCDirectiveKind::Declare); - return ExprError(); + DK == OpenACCDirectiveKind::Declare)) { + if (isa<ArraySubscriptExpr>(CurVarExpr)) { + Diag(VarExpr->getExprLoc(), + diag::err_acc_not_a_var_ref_use_device_declare) + << (DK == OpenACCDirectiveKind::Declare); + return ExprError(); + } + // As an extension, we allow 'array sections'/'sub-arrays' here, as that is + // effectively defining an array, and are in common use. + if (isa<ArraySectionExpr>(CurVarExpr)) + Diag(VarExpr->getExprLoc(), + diag::ext_acc_array_section_use_device_declare) + << (DK == OpenACCDirectiveKind::Declare); } // Sub-arrays/subscript-exprs are fine as long as the base is a diff --git a/clang/lib/Sema/SemaOpenACCAtomic.cpp b/clang/lib/Sema/SemaOpenACCAtomic.cpp index 9c8c8d1..a9319dc 100644 --- a/clang/lib/Sema/SemaOpenACCAtomic.cpp +++ b/clang/lib/Sema/SemaOpenACCAtomic.cpp @@ -576,6 +576,11 @@ class AtomicOperandChecker { return AssocStmt; } + const Expr *IgnoreBeforeCompare(const Expr *E) { + return E->IgnoreParenImpCasts()->IgnoreParenNoopCasts( + SemaRef.getASTContext()); + } + bool CheckVarRefsSame(IDACInfo::ExprKindTy FirstKind, const Expr *FirstX, IDACInfo::ExprKindTy SecondKind, const Expr *SecondX) { llvm::FoldingSetNodeID First_ID, Second_ID; @@ -648,8 +653,10 @@ class AtomicOperandChecker { if (CheckOperandVariable(AssignRes->RHS, PD)) return getRecoveryExpr(); - if (CheckVarRefsSame(FirstExprResults.ExprKind, FirstExprResults.X_Var, - IDACInfo::SimpleAssign, AssignRes->RHS)) + if (CheckVarRefsSame(FirstExprResults.ExprKind, + IgnoreBeforeCompare(FirstExprResults.X_Var), + IDACInfo::SimpleAssign, + IgnoreBeforeCompare(AssignRes->RHS))) return getRecoveryExpr(); break; } @@ -660,9 +667,10 @@ class AtomicOperandChecker { if (SecondExprResults.Failed) return getRecoveryExpr(); - if (CheckVarRefsSame(FirstExprResults.ExprKind, FirstExprResults.X_Var, + if (CheckVarRefsSame(FirstExprResults.ExprKind, + IgnoreBeforeCompare(FirstExprResults.X_Var), SecondExprResults.ExprKind, - SecondExprResults.X_Var)) + IgnoreBeforeCompare(SecondExprResults.X_Var))) return getRecoveryExpr(); break; } diff --git a/clang/lib/Sema/SemaOpenACCClause.cpp b/clang/lib/Sema/SemaOpenACCClause.cpp index 3f90fe8..b54a012 100644 --- a/clang/lib/Sema/SemaOpenACCClause.cpp +++ b/clang/lib/Sema/SemaOpenACCClause.cpp @@ -1919,6 +1919,14 @@ ExprResult SemaOpenACC::CheckReductionVar(OpenACCDirectiveKind DirectiveKind, << EltTy << /*Sub array base type*/ 1; return ExprError(); } + } else if (VarExpr->getType()->isArrayType()) { + // Arrays are considered an 'aggregate variable' explicitly, so are OK, no + // additional checking required. + // + // Glossary: Aggregate variables – a variable of any non-scalar datatype, + // including array or composite variables. + // + // The next branch (record decl) checks for composite variables. } else if (auto *RD = VarExpr->getType()->getAsRecordDecl()) { if (!RD->isStruct() && !RD->isClass()) { Diag(VarExpr->getExprLoc(), diag::err_acc_reduction_composite_type) @@ -2246,7 +2254,13 @@ bool SemaOpenACC::CheckDeclareClause(SemaOpenACC::OpenACCParsedClause &Clause, continue; } } else { - const auto *DRE = cast<DeclRefExpr>(VarExpr); + + const Expr *VarExprTemp = VarExpr; + + while (const auto *ASE = dyn_cast<ArraySectionExpr>(VarExprTemp)) + VarExprTemp = ASE->getBase()->IgnoreParenImpCasts(); + + const auto *DRE = cast<DeclRefExpr>(VarExprTemp); if (const auto *Var = dyn_cast<VarDecl>(DRE->getDecl())) { CurDecl = Var->getCanonicalDecl(); diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp index e2c3cdc..233bb65 100644 --- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp +++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp @@ -5853,6 +5853,8 @@ void Sema::InstantiateFunctionDefinition(SourceLocation PointOfInstantiation, // context seems wrong. Investigate more. ActOnFinishFunctionBody(Function, Body.get(), /*IsInstantiation=*/true); + checkReferenceToTULocalFromOtherTU(Function, PointOfInstantiation); + PerformDependentDiagnostics(PatternDecl, TemplateArgs); if (auto *Listener = getASTMutationListener()) diff --git a/clang/lib/Sema/SemaWasm.cpp b/clang/lib/Sema/SemaWasm.cpp index 6faea24..8998492 100644 --- a/clang/lib/Sema/SemaWasm.cpp +++ b/clang/lib/Sema/SemaWasm.cpp @@ -227,6 +227,53 @@ bool SemaWasm::BuiltinWasmTableCopy(CallExpr *TheCall) { return false; } +bool SemaWasm::BuiltinWasmTestFunctionPointerSignature(CallExpr *TheCall) { + if (SemaRef.checkArgCount(TheCall, 1)) + return true; + + Expr *FuncPtrArg = TheCall->getArg(0); + QualType ArgType = FuncPtrArg->getType(); + + // Check that the argument is a function pointer + const PointerType *PtrTy = ArgType->getAs<PointerType>(); + if (!PtrTy) { + return Diag(FuncPtrArg->getBeginLoc(), + diag::err_typecheck_expect_function_pointer) + << ArgType << FuncPtrArg->getSourceRange(); + } + + const FunctionProtoType *FuncTy = + PtrTy->getPointeeType()->getAs<FunctionProtoType>(); + if (!FuncTy) { + return Diag(FuncPtrArg->getBeginLoc(), + diag::err_typecheck_expect_function_pointer) + << ArgType << FuncPtrArg->getSourceRange(); + } + + // Check that the function pointer doesn't use reference types + if (FuncTy->getReturnType().isWebAssemblyReferenceType()) { + return Diag( + FuncPtrArg->getBeginLoc(), + diag::err_wasm_builtin_test_fp_sig_cannot_include_reference_type) + << 0 << FuncTy->getReturnType() << FuncPtrArg->getSourceRange(); + } + auto NParams = FuncTy->getNumParams(); + for (unsigned I = 0; I < NParams; I++) { + if (FuncTy->getParamType(I).isWebAssemblyReferenceType()) { + return Diag( + FuncPtrArg->getBeginLoc(), + diag:: + err_wasm_builtin_test_fp_sig_cannot_include_reference_type) + << 1 << FuncPtrArg->getSourceRange(); + } + } + + // Set return type to int (the result of the test) + TheCall->setType(getASTContext().IntTy); + + return false; +} + bool SemaWasm::CheckWebAssemblyBuiltinFunctionCall(const TargetInfo &TI, unsigned BuiltinID, CallExpr *TheCall) { @@ -249,6 +296,8 @@ bool SemaWasm::CheckWebAssemblyBuiltinFunctionCall(const TargetInfo &TI, return BuiltinWasmTableFill(TheCall); case WebAssembly::BI__builtin_wasm_table_copy: return BuiltinWasmTableCopy(TheCall); + case WebAssembly::BI__builtin_wasm_test_function_pointer_signature: + return BuiltinWasmTestFunctionPointerSignature(TheCall); } return false; diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp index 10aedb6..f896f9f1 100644 --- a/clang/lib/Serialization/ASTReader.cpp +++ b/clang/lib/Serialization/ASTReader.cpp @@ -8488,6 +8488,7 @@ bool ASTReader::LoadExternalSpecializationsImpl(SpecLookupTableTy &SpecLookups, bool ASTReader::LoadExternalSpecializations(const Decl *D, bool OnlyPartial) { assert(D); + CompleteRedeclChain(D); bool NewSpecsFound = LoadExternalSpecializationsImpl(PartialSpecializationsLookups, D); if (OnlyPartial) diff --git a/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp index d7eea7e..152129e 100644 --- a/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp +++ b/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp @@ -25,18 +25,22 @@ using namespace clang; using namespace ento; namespace { + +class DerefBugType : public BugType { + StringRef ArrayMsg, FieldMsg; + +public: + DerefBugType(CheckerFrontend *FE, StringRef Desc, const char *AMsg, + const char *FMsg = nullptr) + : BugType(FE, Desc), ArrayMsg(AMsg), FieldMsg(FMsg ? FMsg : AMsg) {} + StringRef getArrayMsg() const { return ArrayMsg; } + StringRef getFieldMsg() const { return FieldMsg; } +}; + class DereferenceChecker - : public Checker< check::Location, - check::Bind, - EventDispatcher<ImplicitNullDerefEvent> > { - enum DerefKind { - NullPointer, - UndefinedPointerValue, - AddressOfLabel, - FixedAddress, - }; - - void reportBug(DerefKind K, ProgramStateRef State, const Stmt *S, + : public CheckerFamily<check::Location, check::Bind, + EventDispatcher<ImplicitNullDerefEvent>> { + void reportBug(const DerefBugType &BT, ProgramStateRef State, const Stmt *S, CheckerContext &C) const; bool suppressReport(CheckerContext &C, const Expr *E) const; @@ -52,13 +56,23 @@ public: const LocationContext *LCtx, bool loadedFrom = false); - bool CheckNullDereference = false; - bool CheckFixedDereference = false; - - std::unique_ptr<BugType> BT_Null; - std::unique_ptr<BugType> BT_Undef; - std::unique_ptr<BugType> BT_Label; - std::unique_ptr<BugType> BT_FixedAddress; + CheckerFrontend NullDerefChecker, FixedDerefChecker; + const DerefBugType NullBug{&NullDerefChecker, "Dereference of null pointer", + "a null pointer dereference", + "a dereference of a null pointer"}; + const DerefBugType UndefBug{&NullDerefChecker, + "Dereference of undefined pointer value", + "an undefined pointer dereference", + "a dereference of an undefined pointer value"}; + const DerefBugType LabelBug{&NullDerefChecker, + "Dereference of the address of a label", + "an undefined pointer dereference", + "a dereference of an address of a label"}; + const DerefBugType FixedAddressBug{&FixedDerefChecker, + "Dereference of a fixed address", + "a dereference of a fixed address"}; + + StringRef getDebugTag() const override { return "DereferenceChecker"; } }; } // end anonymous namespace @@ -158,115 +172,87 @@ static bool isDeclRefExprToReference(const Expr *E) { return false; } -void DereferenceChecker::reportBug(DerefKind K, ProgramStateRef State, - const Stmt *S, CheckerContext &C) const { - const BugType *BT = nullptr; - llvm::StringRef DerefStr1; - llvm::StringRef DerefStr2; - switch (K) { - case DerefKind::NullPointer: - if (!CheckNullDereference) { - C.addSink(); - return; - } - BT = BT_Null.get(); - DerefStr1 = " results in a null pointer dereference"; - DerefStr2 = " results in a dereference of a null pointer"; - break; - case DerefKind::UndefinedPointerValue: - if (!CheckNullDereference) { - C.addSink(); +void DereferenceChecker::reportBug(const DerefBugType &BT, + ProgramStateRef State, const Stmt *S, + CheckerContext &C) const { + if (&BT == &FixedAddressBug) { + if (!FixedDerefChecker.isEnabled()) + // Deliberately don't add a sink node if check is disabled. + // This situation may be valid in special cases. return; - } - BT = BT_Undef.get(); - DerefStr1 = " results in an undefined pointer dereference"; - DerefStr2 = " results in a dereference of an undefined pointer value"; - break; - case DerefKind::AddressOfLabel: - if (!CheckNullDereference) { + } else { + if (!NullDerefChecker.isEnabled()) { C.addSink(); return; } - BT = BT_Label.get(); - DerefStr1 = " results in an undefined pointer dereference"; - DerefStr2 = " results in a dereference of an address of a label"; - break; - case DerefKind::FixedAddress: - // Deliberately don't add a sink node if check is disabled. - // This situation may be valid in special cases. - if (!CheckFixedDereference) - return; - - BT = BT_FixedAddress.get(); - DerefStr1 = " results in a dereference of a fixed address"; - DerefStr2 = " results in a dereference of a fixed address"; - break; - }; + } // Generate an error node. ExplodedNode *N = C.generateErrorNode(State); if (!N) return; - SmallString<100> buf; - llvm::raw_svector_ostream os(buf); + SmallString<100> Buf; + llvm::raw_svector_ostream Out(Buf); SmallVector<SourceRange, 2> Ranges; switch (S->getStmtClass()) { case Stmt::ArraySubscriptExprClass: { - os << "Array access"; + Out << "Array access"; const ArraySubscriptExpr *AE = cast<ArraySubscriptExpr>(S); - AddDerefSource(os, Ranges, AE->getBase()->IgnoreParenCasts(), - State.get(), N->getLocationContext()); - os << DerefStr1; + AddDerefSource(Out, Ranges, AE->getBase()->IgnoreParenCasts(), State.get(), + N->getLocationContext()); + Out << " results in " << BT.getArrayMsg(); break; } case Stmt::ArraySectionExprClass: { - os << "Array access"; + Out << "Array access"; const ArraySectionExpr *AE = cast<ArraySectionExpr>(S); - AddDerefSource(os, Ranges, AE->getBase()->IgnoreParenCasts(), - State.get(), N->getLocationContext()); - os << DerefStr1; + AddDerefSource(Out, Ranges, AE->getBase()->IgnoreParenCasts(), State.get(), + N->getLocationContext()); + Out << " results in " << BT.getArrayMsg(); break; } case Stmt::UnaryOperatorClass: { - os << BT->getDescription(); + Out << BT.getDescription(); const UnaryOperator *U = cast<UnaryOperator>(S); - AddDerefSource(os, Ranges, U->getSubExpr()->IgnoreParens(), - State.get(), N->getLocationContext(), true); + AddDerefSource(Out, Ranges, U->getSubExpr()->IgnoreParens(), State.get(), + N->getLocationContext(), true); break; } case Stmt::MemberExprClass: { const MemberExpr *M = cast<MemberExpr>(S); if (M->isArrow() || isDeclRefExprToReference(M->getBase())) { - os << "Access to field '" << M->getMemberNameInfo() << "'" << DerefStr2; - AddDerefSource(os, Ranges, M->getBase()->IgnoreParenCasts(), - State.get(), N->getLocationContext(), true); + Out << "Access to field '" << M->getMemberNameInfo() << "' results in " + << BT.getFieldMsg(); + AddDerefSource(Out, Ranges, M->getBase()->IgnoreParenCasts(), State.get(), + N->getLocationContext(), true); } break; } case Stmt::ObjCIvarRefExprClass: { const ObjCIvarRefExpr *IV = cast<ObjCIvarRefExpr>(S); - os << "Access to instance variable '" << *IV->getDecl() << "'" << DerefStr2; - AddDerefSource(os, Ranges, IV->getBase()->IgnoreParenCasts(), - State.get(), N->getLocationContext(), true); + Out << "Access to instance variable '" << *IV->getDecl() << "' results in " + << BT.getFieldMsg(); + AddDerefSource(Out, Ranges, IV->getBase()->IgnoreParenCasts(), State.get(), + N->getLocationContext(), true); break; } default: break; } - auto report = std::make_unique<PathSensitiveBugReport>( - *BT, buf.empty() ? BT->getDescription() : buf.str(), N); + auto BR = std::make_unique<PathSensitiveBugReport>( + BT, Buf.empty() ? BT.getDescription() : Buf.str(), N); - bugreporter::trackExpressionValue(N, bugreporter::getDerefExpr(S), *report); + bugreporter::trackExpressionValue(N, bugreporter::getDerefExpr(S), *BR); for (SmallVectorImpl<SourceRange>::iterator I = Ranges.begin(), E = Ranges.end(); I!=E; ++I) - report->addRange(*I); + BR->addRange(*I); - C.emitReport(std::move(report)); + C.emitReport(std::move(BR)); } void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S, @@ -275,7 +261,7 @@ void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S, if (l.isUndef()) { const Expr *DerefExpr = getDereferenceExpr(S); if (!suppressReport(C, DerefExpr)) - reportBug(DerefKind::UndefinedPointerValue, C.getState(), DerefExpr, C); + reportBug(UndefBug, C.getState(), DerefExpr, C); return; } @@ -296,7 +282,7 @@ void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S, // we call an "explicit" null dereference. const Expr *expr = getDereferenceExpr(S); if (!suppressReport(C, expr)) { - reportBug(DerefKind::NullPointer, nullState, expr, C); + reportBug(NullBug, nullState, expr, C); return; } } @@ -314,7 +300,7 @@ void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S, if (location.isConstant()) { const Expr *DerefExpr = getDereferenceExpr(S, isLoad); if (!suppressReport(C, DerefExpr)) - reportBug(DerefKind::FixedAddress, notNullState, DerefExpr, C); + reportBug(FixedAddressBug, notNullState, DerefExpr, C); return; } @@ -330,7 +316,7 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S, // One should never write to label addresses. if (auto Label = L.getAs<loc::GotoLabel>()) { - reportBug(DerefKind::AddressOfLabel, C.getState(), S, C); + reportBug(LabelBug, C.getState(), S, C); return; } @@ -351,7 +337,7 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S, if (!StNonNull) { const Expr *expr = getDereferenceExpr(S, /*IsBind=*/true); if (!suppressReport(C, expr)) { - reportBug(DerefKind::NullPointer, StNull, expr, C); + reportBug(NullBug, StNull, expr, C); return; } } @@ -369,7 +355,7 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S, if (V.isConstant()) { const Expr *DerefExpr = getDereferenceExpr(S, true); if (!suppressReport(C, DerefExpr)) - reportBug(DerefKind::FixedAddress, State, DerefExpr, C); + reportBug(FixedAddressBug, State, DerefExpr, C); return; } @@ -392,26 +378,8 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S, C.addTransition(State, this); } -void ento::registerDereferenceModeling(CheckerManager &Mgr) { - Mgr.registerChecker<DereferenceChecker>(); -} - -bool ento::shouldRegisterDereferenceModeling(const CheckerManager &) { - return true; -} - void ento::registerNullDereferenceChecker(CheckerManager &Mgr) { - auto *Chk = Mgr.getChecker<DereferenceChecker>(); - Chk->CheckNullDereference = true; - Chk->BT_Null.reset(new BugType(Mgr.getCurrentCheckerName(), - "Dereference of null pointer", - categories::LogicError)); - Chk->BT_Undef.reset(new BugType(Mgr.getCurrentCheckerName(), - "Dereference of undefined pointer value", - categories::LogicError)); - Chk->BT_Label.reset(new BugType(Mgr.getCurrentCheckerName(), - "Dereference of the address of a label", - categories::LogicError)); + Mgr.getChecker<DereferenceChecker>()->NullDerefChecker.enable(Mgr); } bool ento::shouldRegisterNullDereferenceChecker(const CheckerManager &) { @@ -419,11 +387,7 @@ bool ento::shouldRegisterNullDereferenceChecker(const CheckerManager &) { } void ento::registerFixedAddressDereferenceChecker(CheckerManager &Mgr) { - auto *Chk = Mgr.getChecker<DereferenceChecker>(); - Chk->CheckFixedDereference = true; - Chk->BT_FixedAddress.reset(new BugType(Mgr.getCurrentCheckerName(), - "Dereference of a fixed address", - categories::LogicError)); + Mgr.getChecker<DereferenceChecker>()->FixedDerefChecker.enable(Mgr); } bool ento::shouldRegisterFixedAddressDereferenceChecker( diff --git a/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp index 68efdba..a7704da 100644 --- a/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp +++ b/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp @@ -3730,13 +3730,15 @@ PathDiagnosticPieceRef MallocBugVisitor::VisitNode(const ExplodedNode *N, return nullptr; } - // Save the first destructor/function as release point. - assert(!ReleaseFunctionLC && "There should be only one release point"); + // Record the stack frame that is _responsible_ for this memory release + // event. This will be used by the false positive suppression heuristics + // that recognize the release points of reference-counted objects. + // + // Usually (e.g. in C) we say that the _responsible_ stack frame is the + // current innermost stack frame: ReleaseFunctionLC = CurrentLC->getStackFrame(); - - // See if we're releasing memory while inlining a destructor that - // decrement reference counters (or one of its callees). - // This turns on various common false positive suppressions. + // ...but if the stack contains a destructor call, then we say that the + // outermost destructor stack frame is the _responsible_ one: for (const LocationContext *LC = CurrentLC; LC; LC = LC->getParent()) { if (const auto *DD = dyn_cast<CXXDestructorDecl>(LC->getDecl())) { if (isReferenceCountingPointerDestructor(DD)) { |