Files
swift-mirror/lib/Serialization/ModuleFile.cpp
Jordan Rose dbf41543f5 [serialization] Add support for TupleTypes.
These are implementing using trailing TUPLE_TYPE_ELT records after the
initial TUPLE_TYPE record. This is the next step towards function decls.

Element initializers are silently ignored for now.

Also, do serialize a record for IdentifierTypes, even though we're not
including the components yet, so that we don't serialize the underlying
type more than once.

Swift SVN r5683
2013-06-19 18:19:41 +00:00

622 lines
19 KiB
C++

//===--- ModuleFile.cpp - Loading a serialized module -----------*- c++ -*-===//
//
// This source file is part of the Swift.org open source project
//
// Copyright (c) 2014 - 2015 Apple Inc. and the Swift project authors
// Licensed under Apache License v2.0 with Runtime Library Exception
//
// See http://swift.org/LICENSE.txt for license information
// See http://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
//
//===----------------------------------------------------------------------===//
#include "ModuleFile.h"
#include "ModuleFormat.h"
#include "swift/AST/AST.h"
#include "swift/Serialization/BCReadingExtras.h"
#include "llvm/Support/MemoryBuffer.h"
using namespace swift;
using namespace swift::serialization;
static ModuleStatus
validateControlBlock(llvm::BitstreamCursor &cursor,
llvm::SmallVectorImpl<uint64_t> &scratch) {
// The control block is malformed until we've at least read a major version
// number.
ModuleStatus result = ModuleStatus::Malformed;
auto next = cursor.advance();
while (next.Kind != llvm::BitstreamEntry::EndBlock) {
if (next.Kind == llvm::BitstreamEntry::Error)
return ModuleStatus::Malformed;
if (next.Kind == llvm::BitstreamEntry::SubBlock) {
// Unknown metadata sub-block, possibly for use by a future version of the
// module format.
if (cursor.SkipBlock())
return ModuleStatus::Malformed;
next = cursor.advance();
continue;
}
scratch.clear();
StringRef blobData;
unsigned kind = cursor.readRecord(next.ID, scratch, &blobData);
switch (kind) {
case control_block::METADATA: {
uint16_t versionMajor = scratch[0];
if (versionMajor > VERSION_MAJOR)
return ModuleStatus::FormatTooNew;
result = ModuleStatus::Valid;
break;
}
default:
// Unknown metadata record, possibly for use by a future version of the
// module format.
break;
}
next = cursor.advance();
}
return result;
}
Identifier ModuleFile::getIdentifier(IdentifierID IID) {
if (IID == 0)
return Identifier();
assert(IID <= Identifiers.size() && "invalid identifier ID");
auto identRecord = Identifiers[IID-1];
if (identRecord.Offset == 0)
return identRecord.Ident;
assert(!IdentifierData.empty() && "no identifier data in module");
StringRef rawStrPtr = IdentifierData.substr(identRecord.Offset);
size_t terminatorOffset = rawStrPtr.find('\0');
assert(terminatorOffset != StringRef::npos &&
"unterminated identifier string data");
return ModuleContext->Ctx.getIdentifier(rawStrPtr.slice(0, terminatorOffset));
}
DeclContext *ModuleFile::getDeclContext(DeclID DID) {
if (DID == 0)
return ModuleContext;
Decl *D = getDecl(DID);
if (auto ND = dyn_cast<NominalTypeDecl>(D))
return ND;
if (auto ED = dyn_cast<ExtensionDecl>(D))
return ED;
if (auto CD = dyn_cast<ConstructorDecl>(D))
return CD;
if (auto DD = dyn_cast<DestructorDecl>(D))
return DD;
if (auto FD = dyn_cast<FuncDecl>(D))
return FD->getBody();
llvm_unreachable("unknown DeclContext kind");
}
Decl *ModuleFile::getDecl(DeclID DID) {
if (DID == 0)
return nullptr;
assert(DID <= Decls.size() && "invalid decl ID");
auto &declOrOffset = Decls[DID-1];
if (declOrOffset.is<Decl *>())
return declOrOffset.get<Decl *>();
DeclTypeCursor.JumpToBit(declOrOffset.get<BitOffset>());
auto entry = DeclTypeCursor.advance();
if (entry.Kind != llvm::BitstreamEntry::Record) {
// We don't know how to serialize decls represented by sub-blocks.
error();
return nullptr;
}
ASTContext &ctx = ModuleContext->Ctx;
SmallVector<uint64_t, 64> scratch;
StringRef blobData;
unsigned recordID = DeclTypeCursor.readRecord(entry.ID, scratch, &blobData);
switch (recordID) {
case decls_block::TYPE_ALIAS_DECL: {
IdentifierID nameID;
DeclID contextID;
TypeID underlyingTypeID;
bool isGeneric;
bool isImplicit;
ArrayRef<uint64_t> inheritedIDs;
decls_block::TypeAliasLayout::readRecord(scratch, nameID, contextID,
underlyingTypeID,
isGeneric, isImplicit,
inheritedIDs);
auto inherited =
MutableArrayRef<TypeLoc>(ctx.Allocate<TypeLoc>(inheritedIDs.size()),
inheritedIDs.size());
TypeLoc *nextInheritedType = inherited.data();
for (TypeID TID : inheritedIDs) {
auto type = getType(TID);
new (nextInheritedType) TypeLoc(TypeLoc::withoutLoc(type));
++nextInheritedType;
}
TypeLoc underlyingType = TypeLoc::withoutLoc(getType(underlyingTypeID));
auto alias = new (ctx) TypeAliasDecl(SourceLoc(),
getIdentifier(nameID),
SourceLoc(),
underlyingType,
getDeclContext(contextID),
inherited);
declOrOffset = alias;
if (isImplicit)
alias->setImplicit();
if (isGeneric)
alias->setGenericParameter();
break;
}
case decls_block::STRUCT_DECL: {
IdentifierID nameID;
DeclID contextID;
bool isImplicit;
ArrayRef<uint64_t> inheritedIDs;
decls_block::StructLayout::readRecord(scratch, nameID, contextID,
isImplicit, inheritedIDs);
auto inherited =
MutableArrayRef<TypeLoc>(ctx.Allocate<TypeLoc>(inheritedIDs.size()),
inheritedIDs.size());
TypeLoc *nextInheritedType = inherited.data();
for (TypeID TID : inheritedIDs) {
auto type = getType(TID);
new (nextInheritedType) TypeLoc(TypeLoc::withoutLoc(type));
++nextInheritedType;
}
ArrayRef<uint64_t> memberIDs;
auto entry = DeclTypeCursor.advance();
if (entry.Kind != llvm::BitstreamEntry::Record)
return {};
StringRef blobData;
SmallVector<uint64_t, 16> memberIDBuffer;
unsigned recordID = DeclTypeCursor.readRecord(entry.ID, memberIDBuffer,
&blobData);
if (recordID != decls_block::DECL_CONTEXT) {
error();
return nullptr;
}
decls_block::DeclContextLayout::readRecord(memberIDBuffer, memberIDs);
auto theStruct = new (ctx) StructDecl(SourceLoc(), getIdentifier(nameID),
SourceLoc(), inherited,
/*generic params=*/nullptr,
getDeclContext(contextID));
declOrOffset = theStruct;
if (isImplicit)
theStruct->setImplicit();
if (!memberIDs.empty()) {
MutableArrayRef<Decl *> members(ctx.Allocate<Decl *>(memberIDs.size()),
memberIDs.size());
auto nextMember = members.begin();
for (auto rawID : memberIDBuffer) {
*nextMember = getDecl(rawID);
++nextMember;
}
theStruct->setMembers(members, SourceRange());
}
break;
}
case decls_block::CONSTRUCTOR_DECL: {
DeclID parentID;
bool isImplicit;
DeclID implicitThisID;
decls_block::ConstructorLayout::readRecord(scratch, parentID, isImplicit,
implicitThisID);
auto thisDecl = cast<VarDecl>(getDecl(implicitThisID));
auto parent = cast<NominalTypeDecl>(getDeclContext(parentID));
auto emptyArgs = TuplePattern::create(ctx, SourceLoc(), {}, SourceLoc());
auto ctor = new (ctx) ConstructorDecl(ctx.getIdentifier("constructor"),
SourceLoc(), emptyArgs, thisDecl,
/*generic params=*/nullptr,
parent);
declOrOffset = ctor;
// FIXME: Actually serialize the type instead of reconstructing it here.
ctor->setType(FunctionType::get(MetaTypeType::get(thisDecl->getType(),
ctx),
FunctionType::get(ctx.TheEmptyTupleType,
thisDecl->getType(),
ctx),
ctx));
if (isImplicit)
ctor->setImplicit();
break;
}
case decls_block::VAR_DECL: {
IdentifierID nameID;
DeclID contextID;
bool isImplicit, isNeverLValue;
TypeID typeID;
DeclID getterID, setterID;
DeclID overriddenID;
decls_block::VarLayout::readRecord(scratch, nameID, contextID, isImplicit,
isNeverLValue, typeID, getterID,
setterID, overriddenID);
auto var = new (ctx) VarDecl(SourceLoc(), getIdentifier(nameID),
getType(typeID), nullptr);
declOrOffset = var;
// Explicitly set the DeclContext /after/ creating the VarDecl.
// Sometimes the context has to reference this decl.
var->setDeclContext(getDeclContext(contextID));
var->setNeverUsedAsLValue(isNeverLValue);
if (isImplicit)
var->setImplicit();
if (getterID || setterID) {
var->setProperty(ctx, SourceLoc(),
cast_or_null<FuncDecl>(getDecl(getterID)),
cast_or_null<FuncDecl>(getDecl(setterID)),
SourceLoc());
}
var->setOverriddenDecl(cast_or_null<VarDecl>(getDecl(overriddenID)));
break;
}
default:
// We don't know how to deserialize this kind of decl.
error();
return nullptr;
}
return declOrOffset.get<Decl *>();
}
Type ModuleFile::getType(TypeID TID) {
if (TID == 0)
return Type();
assert(TID <= Types.size() && "invalid decl ID");
auto &typeOrOffset = Types[TID-1];
if (typeOrOffset.is<Type>())
return typeOrOffset.get<Type>();
DeclTypeCursor.JumpToBit(typeOrOffset.get<BitOffset>());
auto entry = DeclTypeCursor.advance();
if (entry.Kind != llvm::BitstreamEntry::Record) {
// We don't know how to serialize types represented by sub-blocks.
error();
return nullptr;
}
SmallVector<uint64_t, 64> scratch;
StringRef blobData;
unsigned recordID = DeclTypeCursor.readRecord(entry.ID, scratch, &blobData);
switch (recordID) {
case decls_block::BUILTIN_TYPE: {
assert(!blobData.empty() && "missing name in BUILTIN_TYPE record");
SmallVector<ValueDecl *, 1> lookupResult;
ASTContext &ctx = ModuleContext->Ctx;
ctx.TheBuiltinModule->lookupValue({}, ctx.getIdentifier(blobData),
NLKind::QualifiedLookup, lookupResult);
if (lookupResult.empty()) {
// This builtin is not supported.
error();
return nullptr;
}
assert(lookupResult.size() == 1 && "multiple types for the same name");
typeOrOffset = cast<TypeDecl>(lookupResult.front())->getDeclaredType();
break;
}
case decls_block::NAME_ALIAS_TYPE: {
DeclID underlyingID;
decls_block::NameAliasTypeLayout::readRecord(scratch, underlyingID);
auto alias = dyn_cast_or_null<TypeAliasDecl>(getDecl(underlyingID));
if (!alias) {
error();
return nullptr;
}
typeOrOffset = alias->getDeclaredType();
break;
}
case decls_block::STRUCT_TYPE: {
DeclID structID;
TypeID parentID;
decls_block::StructTypeLayout::readRecord(scratch, structID, parentID);
typeOrOffset = StructType::get(cast<StructDecl>(getDecl(structID)),
getType(parentID), ModuleContext->Ctx);
break;
}
case decls_block::PAREN_TYPE: {
TypeID underlyingID;
decls_block::ParenTypeLayout::readRecord(scratch, underlyingID);
typeOrOffset = ParenType::get(ModuleContext->Ctx, getType(underlyingID));
break;
}
case decls_block::TUPLE_TYPE: {
// The tuple record itself is empty. Read all trailing elements.
SmallVector<TupleTypeElt, 8> elements;
while (true) {
auto entry = DeclTypeCursor.advance();
if (entry.Kind != llvm::BitstreamEntry::Record)
break;
scratch.clear();
unsigned recordID = DeclTypeCursor.readRecord(entry.ID, scratch,
&blobData);
if (recordID != decls_block::TUPLE_TYPE_ELT)
break;
IdentifierID nameID;
TypeID typeID;
TypeID varargBaseID;
decls_block::TupleTypeEltLayout::readRecord(scratch, nameID, typeID,
varargBaseID);
{
BCOffsetRAII restoreOffset(DeclTypeCursor);
elements.push_back({getType(typeID), getIdentifier(nameID),
/*initializer=*/nullptr, getType(varargBaseID)});
}
}
typeOrOffset = TupleType::get(elements, ModuleContext->Ctx);
break;
}
case decls_block::IDENTIFIER_TYPE: {
TypeID mappedID;
decls_block::IdentifierTypeLayout::readRecord(scratch, mappedID);
// FIXME: Actually recreate the IdentifierType instead of just aliasing the
// underlying mapped type.
typeOrOffset = getType(mappedID);
break;
}
default:
// We don't know how to deserialize this kind of type.
error();
return nullptr;
}
return typeOrOffset.get<Type>();
}
ModuleFile::ModuleFile(llvm::OwningPtr<llvm::MemoryBuffer> &&input)
: ModuleContext(nullptr),
InputFile(std::move(input)),
InputReader(reinterpret_cast<const uint8_t *>(InputFile->getBufferStart()),
reinterpret_cast<const uint8_t *>(InputFile->getBufferEnd())),
Status(ModuleStatus::Valid) {
llvm::BitstreamCursor cursor{InputReader};
for (unsigned char byte : SIGNATURE) {
if (cursor.AtEndOfStream() || cursor.Read(8) != byte)
return error();
}
// Future-proofing: make sure we validate the control block before we try to
// read any other blocks.
bool hasValidControlBlock = false;
SmallVector<uint64_t, 64> scratch;
auto topLevelEntry = cursor.advance();
while (topLevelEntry.Kind == llvm::BitstreamEntry::SubBlock) {
switch (topLevelEntry.ID) {
case llvm::bitc::BLOCKINFO_BLOCK_ID:
if (cursor.ReadBlockInfoBlock())
return error();
break;
case CONTROL_BLOCK_ID: {
cursor.EnterSubBlock(CONTROL_BLOCK_ID);
ModuleStatus err = validateControlBlock(cursor, scratch);
if (err != ModuleStatus::Valid)
return error(err);
hasValidControlBlock = true;
break;
}
case INPUT_BLOCK_ID: {
if (!hasValidControlBlock)
return error();
cursor.EnterSubBlock(INPUT_BLOCK_ID);
auto next = cursor.advance();
while (next.Kind == llvm::BitstreamEntry::Record) {
scratch.clear();
StringRef blobData;
unsigned kind = cursor.readRecord(next.ID, scratch, &blobData);
switch (kind) {
case input_block::SOURCE_FILE:
assert(scratch.empty());
SourcePaths.push_back(blobData);
break;
default:
// Unknown input kind, possibly for use by a future version of the
// module format.
// FIXME: Should we warn about this?
break;
}
next = cursor.advance();
}
if (next.Kind != llvm::BitstreamEntry::EndBlock)
return error();
break;
}
case DECLS_AND_TYPES_BLOCK_ID: {
if (!hasValidControlBlock)
return error();
// The decls-and-types block is lazily loaded. Save the cursor and load
// any abbrev records at the start of the block.
DeclTypeCursor = cursor;
DeclTypeCursor.EnterSubBlock(DECLS_AND_TYPES_BLOCK_ID);
if (DeclTypeCursor.advance().Kind == llvm::BitstreamEntry::Error)
return error();
// With the main cursor, skip over the block and continue.
if (cursor.SkipBlock())
return error();
break;
}
case IDENTIFIER_DATA_BLOCK_ID: {
if (!hasValidControlBlock)
return error();
cursor.EnterSubBlock(IDENTIFIER_DATA_BLOCK_ID);
auto next = cursor.advanceSkippingSubblocks();
while (next.Kind == llvm::BitstreamEntry::Record) {
scratch.clear();
StringRef blobData;
unsigned kind = cursor.readRecord(next.ID, scratch, &blobData);
switch (kind) {
case identifier_block::IDENTIFIER_DATA:
assert(scratch.empty());
IdentifierData = blobData;
break;
default:
// Unknown identifier data, which this version of the compiler won't
// use.
break;
}
next = cursor.advanceSkippingSubblocks();
}
if (next.Kind != llvm::BitstreamEntry::EndBlock)
return error();
break;
}
case INDEX_BLOCK_ID: {
if (!hasValidControlBlock)
return error();
cursor.EnterSubBlock(INDEX_BLOCK_ID);
auto next = cursor.advanceSkippingSubblocks();
while (next.Kind == llvm::BitstreamEntry::Record) {
scratch.clear();
StringRef blobData;
unsigned kind = cursor.readRecord(next.ID, scratch, &blobData);
switch (kind) {
case index_block::DECL_OFFSETS:
assert(blobData.empty());
Decls.assign(scratch.begin(), scratch.end());
break;
case index_block::TYPE_OFFSETS:
assert(blobData.empty());
Types.assign(scratch.begin(), scratch.end());
break;
case index_block::IDENTIFIER_OFFSETS:
assert(blobData.empty());
Identifiers.assign(scratch.begin(), scratch.end());
break;
case index_block::TOP_LEVEL_DECLS:
assert(blobData.empty());
RawTopLevelIDs.assign(scratch.begin(), scratch.end());
break;
default:
// Unknown index kind, which this version of the compiler won't use.
break;
}
next = cursor.advanceSkippingSubblocks();
}
if (next.Kind != llvm::BitstreamEntry::EndBlock)
return error();
break;
}
case FALL_BACK_TO_TRANSLATION_UNIT_ID:
// This is a bring-up hack and will eventually go away.
Status = ModuleStatus::FallBackToTranslationUnit;
break;
default:
// Unknown top-level block, possibly for use by a future version of the
// module format.
if (cursor.SkipBlock())
return error();
break;
}
topLevelEntry = cursor.advance(llvm::BitstreamCursor::AF_DontPopBlockAtEnd);
}
if (topLevelEntry.Kind != llvm::BitstreamEntry::EndBlock)
return error();
}
void ModuleFile::buildTopLevelDeclMap() {
// FIXME: be more lazy about deserialization by encoding this some other way.
for (DeclID ID : RawTopLevelIDs) {
auto value = cast<ValueDecl>(getDecl(ID));
TopLevelIDs[value->getName()] = ID;
}
RawTopLevelIDs.clear();
}
void ModuleFile::lookupValue(Identifier name,
SmallVectorImpl<ValueDecl*> &results) {
if (!RawTopLevelIDs.empty())
buildTopLevelDeclMap();
if (DeclID ID = TopLevelIDs.lookup(name))
results.push_back(cast<ValueDecl>(getDecl(ID)));
}