diff options
Diffstat (limited to 'lib/Serialization/PCHReader.cpp')
| -rw-r--r-- | lib/Serialization/PCHReader.cpp | 3857 |
1 files changed, 3857 insertions, 0 deletions
diff --git a/lib/Serialization/PCHReader.cpp b/lib/Serialization/PCHReader.cpp new file mode 100644 index 0000000000..852bbaf21a --- /dev/null +++ b/lib/Serialization/PCHReader.cpp @@ -0,0 +1,3857 @@ +//===--- PCHReader.cpp - Precompiled Headers Reader -------------*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This file defines the PCHReader class, which reads a precompiled header. +// +//===----------------------------------------------------------------------===// + +#include "clang/Serialization/PCHReader.h" +#include "clang/Serialization/PCHDeserializationListener.h" +#include "clang/Frontend/FrontendDiagnostic.h" +#include "clang/Frontend/Utils.h" +#include "clang/Sema/Sema.h" +#include "clang/AST/ASTConsumer.h" +#include "clang/AST/ASTContext.h" +#include "clang/AST/Expr.h" +#include "clang/AST/Type.h" +#include "clang/AST/TypeLocVisitor.h" +#include "clang/Lex/MacroInfo.h" +#include "clang/Lex/PreprocessingRecord.h" +#include "clang/Lex/Preprocessor.h" +#include "clang/Lex/HeaderSearch.h" +#include "clang/Basic/OnDiskHashTable.h" +#include "clang/Basic/SourceManager.h" +#include "clang/Basic/SourceManagerInternals.h" +#include "clang/Basic/FileManager.h" +#include "clang/Basic/TargetInfo.h" +#include "clang/Basic/Version.h" +#include "llvm/ADT/StringExtras.h" +#include "llvm/Bitcode/BitstreamReader.h" +#include "llvm/Support/MemoryBuffer.h" +#include "llvm/Support/ErrorHandling.h" +#include "llvm/System/Path.h" +#include <algorithm> +#include <iterator> +#include <cstdio> +#include <sys/stat.h> +using namespace clang; + +//===----------------------------------------------------------------------===// +// PCH reader validator implementation +//===----------------------------------------------------------------------===// + +PCHReaderListener::~PCHReaderListener() {} + +bool +PCHValidator::ReadLanguageOptions(const LangOptions &LangOpts) { + const LangOptions &PPLangOpts = PP.getLangOptions(); +#define PARSE_LANGOPT_BENIGN(Option) +#define PARSE_LANGOPT_IMPORTANT(Option, DiagID) \ + if (PPLangOpts.Option != LangOpts.Option) { \ + Reader.Diag(DiagID) << LangOpts.Option << PPLangOpts.Option; \ + return true; \ + } + + PARSE_LANGOPT_BENIGN(Trigraphs); + PARSE_LANGOPT_BENIGN(BCPLComment); + PARSE_LANGOPT_BENIGN(DollarIdents); + PARSE_LANGOPT_BENIGN(AsmPreprocessor); + PARSE_LANGOPT_IMPORTANT(GNUMode, diag::warn_pch_gnu_extensions); + PARSE_LANGOPT_IMPORTANT(GNUKeywords, diag::warn_pch_gnu_keywords); + PARSE_LANGOPT_BENIGN(ImplicitInt); + PARSE_LANGOPT_BENIGN(Digraphs); + PARSE_LANGOPT_BENIGN(HexFloats); + PARSE_LANGOPT_IMPORTANT(C99, diag::warn_pch_c99); + PARSE_LANGOPT_IMPORTANT(Microsoft, diag::warn_pch_microsoft_extensions); + PARSE_LANGOPT_IMPORTANT(CPlusPlus, diag::warn_pch_cplusplus); + PARSE_LANGOPT_IMPORTANT(CPlusPlus0x, diag::warn_pch_cplusplus0x); + PARSE_LANGOPT_BENIGN(CXXOperatorName); + PARSE_LANGOPT_IMPORTANT(ObjC1, diag::warn_pch_objective_c); + PARSE_LANGOPT_IMPORTANT(ObjC2, diag::warn_pch_objective_c2); + PARSE_LANGOPT_IMPORTANT(ObjCNonFragileABI, diag::warn_pch_nonfragile_abi); + PARSE_LANGOPT_IMPORTANT(ObjCNonFragileABI2, diag::warn_pch_nonfragile_abi2); + PARSE_LANGOPT_IMPORTANT(NoConstantCFStrings, + diag::warn_pch_no_constant_cfstrings); + PARSE_LANGOPT_BENIGN(PascalStrings); + PARSE_LANGOPT_BENIGN(WritableStrings); + PARSE_LANGOPT_IMPORTANT(LaxVectorConversions, + diag::warn_pch_lax_vector_conversions); + PARSE_LANGOPT_IMPORTANT(AltiVec, diag::warn_pch_altivec); + PARSE_LANGOPT_IMPORTANT(Exceptions, diag::warn_pch_exceptions); + PARSE_LANGOPT_IMPORTANT(SjLjExceptions, diag::warn_pch_sjlj_exceptions); + PARSE_LANGOPT_IMPORTANT(NeXTRuntime, diag::warn_pch_objc_runtime); + PARSE_LANGOPT_IMPORTANT(Freestanding, diag::warn_pch_freestanding); + PARSE_LANGOPT_IMPORTANT(NoBuiltin, diag::warn_pch_builtins); + PARSE_LANGOPT_IMPORTANT(ThreadsafeStatics, + diag::warn_pch_thread_safe_statics); + PARSE_LANGOPT_IMPORTANT(POSIXThreads, diag::warn_pch_posix_threads); + PARSE_LANGOPT_IMPORTANT(Blocks, diag::warn_pch_blocks); + PARSE_LANGOPT_BENIGN(EmitAllDecls); + PARSE_LANGOPT_IMPORTANT(MathErrno, diag::warn_pch_math_errno); + PARSE_LANGOPT_BENIGN(getSignedOverflowBehavior()); + PARSE_LANGOPT_IMPORTANT(HeinousExtensions, + diag::warn_pch_heinous_extensions); + // FIXME: Most of the options below are benign if the macro wasn't + // used. Unfortunately, this means that a PCH compiled without + // optimization can't be used with optimization turned on, even + // though the only thing that changes is whether __OPTIMIZE__ was + // defined... but if __OPTIMIZE__ never showed up in the header, it + // doesn't matter. We could consider making this some special kind + // of check. + PARSE_LANGOPT_IMPORTANT(Optimize, diag::warn_pch_optimize); + PARSE_LANGOPT_IMPORTANT(OptimizeSize, diag::warn_pch_optimize_size); + PARSE_LANGOPT_IMPORTANT(Static, diag::warn_pch_static); + PARSE_LANGOPT_IMPORTANT(PICLevel, diag::warn_pch_pic_level); + PARSE_LANGOPT_IMPORTANT(GNUInline, diag::warn_pch_gnu_inline); + PARSE_LANGOPT_IMPORTANT(NoInline, diag::warn_pch_no_inline); + PARSE_LANGOPT_IMPORTANT(AccessControl, diag::warn_pch_access_control); + PARSE_LANGOPT_IMPORTANT(CharIsSigned, diag::warn_pch_char_signed); + PARSE_LANGOPT_IMPORTANT(ShortWChar, diag::warn_pch_short_wchar); + if ((PPLangOpts.getGCMode() != 0) != (LangOpts.getGCMode() != 0)) { + Reader.Diag(diag::warn_pch_gc_mode) + << LangOpts.getGCMode() << PPLangOpts.getGCMode(); + return true; + } + PARSE_LANGOPT_BENIGN(getVisibilityMode()); + PARSE_LANGOPT_IMPORTANT(getStackProtectorMode(), + diag::warn_pch_stack_protector); + PARSE_LANGOPT_BENIGN(InstantiationDepth); + PARSE_LANGOPT_IMPORTANT(OpenCL, diag::warn_pch_opencl); + PARSE_LANGOPT_BENIGN(CatchUndefined); + PARSE_LANGOPT_IMPORTANT(ElideConstructors, diag::warn_pch_elide_constructors); + PARSE_LANGOPT_BENIGN(SpellChecking); +#undef PARSE_LANGOPT_IMPORTANT +#undef PARSE_LANGOPT_BENIGN + + return false; +} + +bool PCHValidator::ReadTargetTriple(llvm::StringRef Triple) { + if (Triple == PP.getTargetInfo().getTriple().str()) + return false; + + Reader.Diag(diag::warn_pch_target_triple) + << Triple << PP.getTargetInfo().getTriple().str(); + return true; +} + +struct EmptyStringRef { + bool operator ()(llvm::StringRef r) const { return r.empty(); } +}; +struct EmptyBlock { + bool operator ()(const PCHPredefinesBlock &r) const { return r.Data.empty(); } +}; + +static bool EqualConcatenations(llvm::SmallVector<llvm::StringRef, 2> L, + PCHPredefinesBlocks R) { + // First, sum up the lengths. + unsigned LL = 0, RL = 0; + for (unsigned I = 0, N = L.size(); I != N; ++I) { + LL += L[I].size(); + } + for (unsigned I = 0, N = R.size(); I != N; ++I) { + RL += R[I].Data.size(); + } + if (LL != RL) + return false; + if (LL == 0 && RL == 0) + return true; + + // Kick out empty parts, they confuse the algorithm below. + L.erase(std::remove_if(L.begin(), L.end(), EmptyStringRef()), L.end()); + R.erase(std::remove_if(R.begin(), R.end(), EmptyBlock()), R.end()); + + // Do it the hard way. At this point, both vectors must be non-empty. + llvm::StringRef LR = L[0], RR = R[0].Data; + unsigned LI = 0, RI = 0, LN = L.size(), RN = R.size(); + (void) RN; + for (;;) { + // Compare the current pieces. + if (LR.size() == RR.size()) { + // If they're the same length, it's pretty easy. + if (LR != RR) + return false; + // Both pieces are done, advance. + ++LI; + ++RI; + // If either string is done, they're both done, since they're the same + // length. + if (LI == LN) { + assert(RI == RN && "Strings not the same length after all?"); + return true; + } + LR = L[LI]; + RR = R[RI].Data; + } else if (LR.size() < RR.size()) { + // Right piece is longer. + if (!RR.startswith(LR)) + return false; + ++LI; + assert(LI != LN && "Strings not the same length after all?"); + RR = RR.substr(LR.size()); + LR = L[LI]; + } else { + // Left piece is longer. + if (!LR.startswith(RR)) + return false; + ++RI; + assert(RI != RN && "Strings not the same length after all?"); + LR = LR.substr(RR.size()); + RR = R[RI].Data; + } + } +} + +static std::pair<FileID, llvm::StringRef::size_type> +FindMacro(const PCHPredefinesBlocks &Buffers, llvm::StringRef MacroDef) { + std::pair<FileID, llvm::StringRef::size_type> Res; + for (unsigned I = 0, N = Buffers.size(); I != N; ++I) { + Res.second = Buffers[I].Data.find(MacroDef); + if (Res.second != llvm::StringRef::npos) { + Res.first = Buffers[I].BufferID; + break; + } + } + return Res; +} + +bool PCHValidator::ReadPredefinesBuffer(const PCHPredefinesBlocks &Buffers, + llvm::StringRef OriginalFileName, + std::string &SuggestedPredefines) { + // We are in the context of an implicit include, so the predefines buffer will + // have a #include entry for the PCH file itself (as normalized by the + // preprocessor initialization). Find it and skip over it in the checking + // below. + llvm::SmallString<256> PCHInclude; + PCHInclude += "#include \""; + PCHInclude += NormalizeDashIncludePath(OriginalFileName); + PCHInclude += "\"\n"; + std::pair<llvm::StringRef,llvm::StringRef> Split = + llvm::StringRef(PP.getPredefines()).split(PCHInclude.str()); + llvm::StringRef Left = Split.first, Right = Split.second; + if (Left == PP.getPredefines()) { + Error("Missing PCH include entry!"); + return true; + } + + // If the concatenation of all the PCH buffers is equal to the adjusted + // command line, we're done. + // We build a SmallVector of the command line here, because we'll eventually + // need to support an arbitrary amount of pieces anyway (when we have chained + // PCH reading). + llvm::SmallVector<llvm::StringRef, 2> CommandLine; + CommandLine.push_back(Left); + CommandLine.push_back(Right); + if (EqualConcatenations(CommandLine, Buffers)) + return false; + + SourceManager &SourceMgr = PP.getSourceManager(); + + // The predefines buffers are different. Determine what the differences are, + // and whether they require us to reject the PCH file. + llvm::SmallVector<llvm::StringRef, 8> PCHLines; + for (unsigned I = 0, N = Buffers.size(); I != N; ++I) + Buffers[I].Data.split(PCHLines, "\n", /*MaxSplit=*/-1, /*KeepEmpty=*/false); + + llvm::SmallVector<llvm::StringRef, 8> CmdLineLines; + Left.split(CmdLineLines, "\n", /*MaxSplit=*/-1, /*KeepEmpty=*/false); + Right.split(CmdLineLines, "\n", /*MaxSplit=*/-1, /*KeepEmpty=*/false); + + // Sort both sets of predefined buffer lines, since we allow some extra + // definitions and they may appear at any point in the output. + std::sort(CmdLineLines.begin(), CmdLineLines.end()); + std::sort(PCHLines.begin(), PCHLines.end()); + + // Determine which predefines that were used to build the PCH file are missing + // from the command line. + std::vector<llvm::StringRef> MissingPredefines; + std::set_difference(PCHLines.begin(), PCHLines.end(), + CmdLineLines.begin(), CmdLineLines.end(), + std::back_inserter(MissingPredefines)); + + bool MissingDefines = false; + bool ConflictingDefines = false; + for (unsigned I = 0, N = MissingPredefines.size(); I != N; ++I) { + llvm::StringRef Missing = MissingPredefines[I]; + if (!Missing.startswith("#define ")) { + Reader.Diag(diag::warn_pch_compiler_options_mismatch); + return true; + } + + // This is a macro definition. Determine the name of the macro we're + // defining. + std::string::size_type StartOfMacroName = strlen("#define "); + std::string::size_type EndOfMacroName + = Missing.find_first_of("( \n\r", StartOfMacroName); + assert(EndOfMacroName != std::string::npos && + "Couldn't find the end of the macro name"); + llvm::StringRef MacroName = Missing.slice(StartOfMacroName, EndOfMacroName); + + // Determine whether this macro was given a different definition on the + // command line. + std::string MacroDefStart = "#define " + MacroName.str(); + std::string::size_type MacroDefLen = MacroDefStart.size(); + llvm::SmallVector<llvm::StringRef, 8>::iterator ConflictPos + = std::lower_bound(CmdLineLines.begin(), CmdLineLines.end(), + MacroDefStart); + for (; ConflictPos != CmdLineLines.end(); ++ConflictPos) { + if (!ConflictPos->startswith(MacroDefStart)) { + // Different macro; we're done. + ConflictPos = CmdLineLines.end(); + break; + } + + assert(ConflictPos->size() > MacroDefLen && + "Invalid #define in predefines buffer?"); + if ((*ConflictPos)[MacroDefLen] != ' ' && + (*ConflictPos)[MacroDefLen] != '(') + continue; // Longer macro name; keep trying. + + // We found a conflicting macro definition. + break; + } + + if (ConflictPos != CmdLineLines.end()) { + Reader.Diag(diag::warn_cmdline_conflicting_macro_def) + << MacroName; + + // Show the definition of this macro within the PCH file. + std::pair<FileID, llvm::StringRef::size_type> MacroLoc = + FindMacro(Buffers, Missing); + assert(MacroLoc.second!=llvm::StringRef::npos && "Unable to find macro!"); + SourceLocation PCHMissingLoc = + SourceMgr.getLocForStartOfFile(MacroLoc.first) + .getFileLocWithOffset(MacroLoc.second); + Reader.Diag(PCHMissingLoc, diag::note_pch_macro_defined_as) << MacroName; + + ConflictingDefines = true; + continue; + } + + // If the macro doesn't conflict, then we'll just pick up the macro + // definition from the PCH file. Warn the user that they made a mistake. + if (ConflictingDefines) + continue; // Don't complain if there are already conflicting defs + + if (!MissingDefines) { + Reader.Diag(diag::warn_cmdline_missing_macro_defs); + MissingDefines = true; + } + + // Show the definition of this macro within the PCH file. + std::pair<FileID, llvm::StringRef::size_type> MacroLoc = + FindMacro(Buffers, Missing); + assert(MacroLoc.second!=llvm::StringRef::npos && "Unable to find macro!"); + SourceLocation PCHMissingLoc = + SourceMgr.getLocForStartOfFile(MacroLoc.first) + .getFileLocWithOffset(MacroLoc.second); + Reader.Diag(PCHMissingLoc, diag::note_using_macro_def_from_pch); + } + + if (ConflictingDefines) + return true; + + // Determine what predefines were introduced based on command-line + // parameters that were not present when building the PCH + // file. Extra #defines are okay, so long as the identifiers being + // defined were not used within the precompiled header. + std::vector<llvm::StringRef> ExtraPredefines; + std::set_difference(CmdLineLines.begin(), CmdLineLines.end(), + PCHLines.begin(), PCHLines.end(), + std::back_inserter(ExtraPredefines)); + for (unsigned I = 0, N = ExtraPredefines.size(); I != N; ++I) { + llvm::StringRef &Extra = ExtraPredefines[I]; + if (!Extra.startswith("#define ")) { + Reader.Diag(diag::warn_pch_compiler_options_mismatch); + return true; + } + + // This is an extra macro definition. Determine the name of the + // macro we're defining. + std::string::size_type StartOfMacroName = strlen("#define "); + std::string::size_type EndOfMacroName + = Extra.find_first_of("( \n\r", StartOfMacroName); + assert(EndOfMacroName != std::string::npos && + "Couldn't find the end of the macro name"); + llvm::StringRef MacroName = Extra.slice(StartOfMacroName, EndOfMacroName); + + // Check whether this name was used somewhere in the PCH file. If + // so, defining it as a macro could change behavior, so we reject + // the PCH file. + if (IdentifierInfo *II = Reader.get(MacroName)) { + Reader.Diag(diag::warn_macro_name_used_in_pch) << II; + return true; + } + + // Add this definition to the suggested predefines buffer. + SuggestedPredefines += Extra; + SuggestedPredefines += '\n'; + } + + // If we get here, it's because the predefines buffer had compatible + // contents. Accept the PCH file. + return false; +} + +void PCHValidator::ReadHeaderFileInfo(const HeaderFileInfo &HFI, + unsigned ID) { + PP.getHeaderSearchInfo().setHeaderFileInfoForUID(HFI, ID); + ++NumHeaderInfos; +} + +void PCHValidator::ReadCounter(unsigned Value) { + PP.setCounterValue(Value); +} + +//===----------------------------------------------------------------------===// +// PCH reader implementation +//===----------------------------------------------------------------------===// + +PCHReader::PCHReader(Preprocessor &PP, ASTContext *Context, + const char *isysroot, bool DisableValidation) + : Listener(new PCHValidator(PP, *this)), DeserializationListener(0), + SourceMgr(PP.getSourceManager()), FileMgr(PP.getFileManager()), + Diags(PP.getDiagnostics()), SemaObj(0), PP(&PP), Context(Context), + Consumer(0), isysroot(isysroot), DisableValidation(DisableValidation), + NumStatHits(0), NumStatMisses(0), NumSLocEntriesRead(0), + TotalNumSLocEntries(0), NumStatementsRead(0), TotalNumStatements(0), + NumMacrosRead(0), TotalNumMacros(0), NumSelectorsRead(0), + NumMethodPoolEntriesRead(0), NumMethodPoolMisses(0), + TotalNumMethodPoolEntries(0), NumLexicalDeclContextsRead(0), + TotalLexicalDeclContexts(0), NumVisibleDeclContextsRead(0), + TotalVisibleDeclContexts(0), NumCurrentElementsDeserializing(0) { + RelocatablePCH = false; +} + +PCHReader::PCHReader(SourceManager &SourceMgr, FileManager &FileMgr, + Diagnostic &Diags, const char *isysroot, + bool DisableValidation) + : DeserializationListener(0), SourceMgr(SourceMgr), FileMgr(FileMgr), + Diags(Diags), SemaObj(0), PP(0), Context(0), Consumer(0), + isysroot(isysroot), DisableValidation(DisableValidation), NumStatHits(0), + NumStatMisses(0), NumSLocEntriesRead(0), TotalNumSLocEntries(0), + NumStatementsRead(0), TotalNumStatements(0), NumMacrosRead(0), + TotalNumMacros(0), NumSelectorsRead(0), NumMethodPoolEntriesRead(0), + NumMethodPoolMisses(0), TotalNumMethodPoolEntries(0), + NumLexicalDeclContextsRead(0), TotalLexicalDeclContexts(0), + NumVisibleDeclContextsRead(0), TotalVisibleDeclContexts(0), + NumCurrentElementsDeserializing(0) { + RelocatablePCH = false; +} + +PCHReader::~PCHReader() { + for (unsigned i = 0, e = Chain.size(); i != e; ++i) + delete Chain[e - i - 1]; +} + +PCHReader::PerFileData::PerFileData() + : StatCache(0), LocalNumSLocEntries(0), LocalNumTypes(0), TypeOffsets(0), + LocalNumDecls(0), DeclOffsets(0), LocalNumIdentifiers(0), + IdentifierOffsets(0), IdentifierTableData(0), IdentifierLookupTable(0), + LocalNumMacroDefinitions(0), MacroDefinitionOffsets(0), + NumPreallocatedPreprocessingEntities(0), SelectorLookupTable(0), + SelectorLookupTableData(0), SelectorOffsets(0), LocalNumSelectors(0) +{} + +void +PCHReader::setDeserializationListener(PCHDeserializationListener *Listener) { + DeserializationListener = Listener; + if (DeserializationListener) + DeserializationListener->SetReader(this); +} + + +namespace { +class PCHSelectorLookupTrait { + PCHReader &Reader; + +public: + struct data_type { + pch::SelectorID ID; + ObjCMethodList Instance, Factory; + }; + + typedef Selector external_key_type; + typedef external_key_type internal_key_type; + + explicit PCHSelectorLookupTrait(PCHReader &Reader) : Reader(Reader) { } + + static bool EqualKey(const internal_key_type& a, + const internal_key_type& b) { + return a == b; + } + + static unsigned ComputeHash(Selector Sel) { + unsigned N = Sel.getNumArgs(); + if (N == 0) + ++N; + unsigned R = 5381; + for (unsigned I = 0; I != N; ++I) + if (IdentifierInfo *II = Sel.getIdentifierInfoForSlot(I)) + R = llvm::HashString(II->getName(), R); + return R; + } + + // This hopefully will just get inlined and removed by the optimizer. + static const internal_key_type& + GetInternalKey(const external_key_type& x) { return x; } + + static std::pair<unsigned, unsigned> + ReadKeyDataLength(const unsigned char*& d) { + using namespace clang::io; + unsigned KeyLen = ReadUnalignedLE16(d); + unsigned DataLen = ReadUnalignedLE16(d); + return std::make_pair(KeyLen, DataLen); + } + + internal_key_type ReadKey(const unsigned char* d, unsigned) { + using namespace clang::io; + SelectorTable &SelTable = Reader.getContext()->Selectors; + unsigned N = ReadUnalignedLE16(d); + IdentifierInfo *FirstII + = Reader.DecodeIdentifierInfo(ReadUnalignedLE32(d)); + if (N == 0) + return SelTable.getNullarySelector(FirstII); + else if (N == 1) + return SelTable.getUnarySelector(FirstII); + + llvm::SmallVector<IdentifierInfo *, 16> Args; + Args.push_back(FirstII); + for (unsigned I = 1; I != N; ++I) + Args.push_back(Reader.DecodeIdentifierInfo(ReadUnalignedLE32(d))); + + return SelTable.getSelector(N, Args.data()); + } + + data_type ReadData(Selector, const unsigned char* d, unsigned DataLen) { + using namespace clang::io; + + data_type Result; + + Result.ID = ReadUnalignedLE32(d); + unsigned NumInstanceMethods = ReadUnalignedLE16(d); + unsigned NumFactoryMethods = ReadUnalignedLE16(d); + + // Load instance methods + ObjCMethodList *Prev = 0; + for (unsigned I = 0; I != NumInstanceMethods; ++I) { + ObjCMethodDecl *Method + = cast<ObjCMethodDecl>(Reader.GetDecl(ReadUnalignedLE32(d))); + if (!Result.Instance.Method) { + // This is the first method, which is the easy case. + Result.Instance.Method = Method; + Prev = &Result.Instance; + continue; + } + + ObjCMethodList *Mem = + Reader.getSema()->BumpAlloc.Allocate<ObjCMethodList>(); + Prev->Next = new (Mem) ObjCMethodList(Method, 0); + Prev = Prev->Next; + } + + // Load factory methods + Prev = 0; + for (unsigned I = 0; I != NumFactoryMethods; ++I) { + ObjCMethodDecl *Method + = cast<ObjCMethodDecl>(Reader.GetDecl(ReadUnalignedLE32(d))); + if (!Result.Factory.Method) { + // This is the first method, which is the easy case. + Result.Factory.Method = Method; + Prev = &Result.Factory; + continue; + } + + ObjCMethodList *Mem = + Reader.getSema()->BumpAlloc.Allocate<ObjCMethodList>(); + Prev->Next = new (Mem) ObjCMethodList(Method, 0); + Prev = Prev->Next; + } + + return Result; + } +}; + +} // end anonymous namespace + +/// \brief The on-disk hash table used for the global method pool. +typedef OnDiskChainedHashTable<PCHSelectorLookupTrait> + PCHSelectorLookupTable; + +namespace { +class PCHIdentifierLookupTrait { + PCHReader &Reader; + llvm::BitstreamCursor &Stream; + + // If we know the IdentifierInfo in advance, it is here and we will + // not build a new one. Used when deserializing information about an + // identifier that was constructed before the PCH file was read. + IdentifierInfo *KnownII; + +public: + typedef IdentifierInfo * data_type; + + typedef const std::pair<const char*, unsigned> external_key_type; + + typedef external_key_type internal_key_type; + + PCHIdentifierLookupTrait(PCHReader &Reader, llvm::BitstreamCursor &Stream, + IdentifierInfo *II = 0) + : Reader(Reader), Stream(Stream), KnownII(II) { } + + static bool EqualKey(const internal_key_type& a, + const internal_key_type& b) { + return (a.second == b.second) ? memcmp(a.first, b.first, a.second) == 0 + : false; + } + + static unsigned ComputeHash(const internal_key_type& a) { + return llvm::HashString(llvm::StringRef(a.first, a.second)); + } + + // This hopefully will just get inlined and removed by the optimizer. + static const internal_key_type& + GetInternalKey(const external_key_type& x) { return x; } + + static std::pair<unsigned, unsigned> + ReadKeyDataLength(const unsigned char*& d) { + using namespace clang::io; + unsigned DataLen = ReadUnalignedLE16(d); + unsigned KeyLen = ReadUnalignedLE16(d); + return std::make_pair(KeyLen, DataLen); + } + + static std::pair<const char*, unsigned> + ReadKey(const unsigned char* d, unsigned n) { + assert(n >= 2 && d[n-1] == '\0'); + return std::make_pair((const char*) d, n-1); + } + + IdentifierInfo *ReadData(const internal_key_type& k, + const unsigned char* d, + unsigned DataLen) { + using namespace clang::io; + pch::IdentID ID = ReadUnalignedLE32(d); + bool IsInteresting = ID & 0x01; + + // Wipe out the "is interesting" bit. + ID = ID >> 1; + + if (!IsInteresting) { + // For uninteresting identifiers, just build the IdentifierInfo + // and associate it with the persistent ID. + IdentifierInfo *II = KnownII; + if (!II) + II = &Reader.getIdentifierTable().getOwn(k.first, k.first + k.second); + Reader.SetIdentifierInfo(ID, II); + II->setIsFromPCH(); + return II; + } + + unsigned Bits = ReadUnalignedLE16(d); + bool CPlusPlusOperatorKeyword = Bits & 0x01; + Bits >>= 1; + bool HasRevertedTokenIDToIdentifier = Bits & 0x01; + Bits >>= 1; + bool Poisoned = Bits & 0x01; + Bits >>= 1; + bool ExtensionToken = Bits & 0x01; + Bits >>= 1; + bool hasMacroDefinition = Bits & 0x01; + Bits >>= 1; + unsigned ObjCOrBuiltinID = Bits & 0x3FF; + Bits >>= 10; + + assert(Bits == 0 && "Extra bits in the identifier?"); + DataLen -= 6; + + // Build the IdentifierInfo itself and link the identifier ID with + // the new IdentifierInfo. + IdentifierInfo *II = KnownII; + if (!II) + II = &Reader.getIdentifierTable().getOwn(k.first, k.first + k.second); + Reader.SetIdentifierInfo(ID, II); + + // Set or check the various bits in the IdentifierInfo structure. + // Token IDs are read-only. + if (HasRevertedTokenIDToIdentifier) + II->RevertTokenIDToIdentifier(); + II->setObjCOrBuiltinID(ObjCOrBuiltinID); + assert(II->isExtensionToken() == ExtensionToken && + "Incorrect extension token flag"); + (void)ExtensionToken; + II->setIsPoisoned(Poisoned); + assert(II->isCPlusPlusOperatorKeyword() == CPlusPlusOperatorKeyword && + "Incorrect C++ operator keyword flag"); + (void)CPlusPlusOperatorKeyword; + + // If this identifier is a macro, deserialize the macro + // definition. + if (hasMacroDefinition) { + uint32_t Offset = ReadUnalignedLE32(d); + Reader.ReadMacroRecord(Stream, Offset); + DataLen -= 4; + } + + // Read all of the declarations visible at global scope with this + // name. + if (Reader.getContext() == 0) return II; + if (DataLen > 0) { + llvm::SmallVector<uint32_t, 4> DeclIDs; + for (; DataLen > 0; DataLen -= 4) + DeclIDs.push_back(ReadUnalignedLE32(d)); + Reader.SetGloballyVisibleDecls(II, DeclIDs); + } + + II->setIsFromPCH(); + return II; + } +}; + +} // end anonymous namespace + +/// \brief The on-disk hash table used to contain information about +/// all of the identifiers in the program. +typedef OnDiskChainedHashTable<PCHIdentifierLookupTrait> + PCHIdentifierLookupTable; + +void PCHReader::Error(const char *Msg) { + Diag(diag::err_fe_pch_malformed) << Msg; +} + +/// \brief Check the contents of the concatenation of all predefines buffers in +/// the PCH chain against the contents of the predefines buffer of the current +/// compiler invocation. +/// +/// The contents should be the same. If not, then some command-line option +/// changed the preprocessor state and we must probably reject the PCH file. +/// +/// \returns true if there was a mismatch (in which case the PCH file +/// should be ignored), or false otherwise. +bool PCHReader::CheckPredefinesBuffers() { + if (Listener) + return Listener->ReadPredefinesBuffer(PCHPredefinesBuffers, + ActualOriginalFileName, + SuggestedPredefines); + return false; +} + +//===----------------------------------------------------------------------===// +// Source Manager Deserialization +//===----------------------------------------------------------------------===// + +/// \brief Read the line table in the source manager block. +/// \returns true if ther was an error. +bool PCHReader::ParseLineTable(llvm::SmallVectorImpl<uint64_t> &Record) { + unsigned Idx = 0; + LineTableInfo &LineTable = SourceMgr.getLineTable(); + + // Parse the file names + std::map<int, int> FileIDs; + for (int I = 0, N = Record[Idx++]; I != N; ++I) { + // Extract the file name + unsigned FilenameLen = Record[Idx++]; + std::string Filename(&Record[Idx], &Record[Idx] + FilenameLen); + Idx += FilenameLen; + MaybeAddSystemRootToFilename(Filename); + FileIDs[I] = LineTable.getLineTableFilenameID(Filename.c_str(), + Filename.size()); + } + + // Parse the line entries + std::vector<LineEntry> Entries; + while (Idx < Record.size()) { + int FID = Record[Idx++]; + + // Extract the line entries + unsigned NumEntries = Record[Idx++]; + assert(NumEntries && "Numentries is 00000"); + Entries.clear(); + Entries.reserve(NumEntries); + for (unsigned I = 0; I != NumEntries; ++I) { + unsigned FileOffset = Record[Idx++]; + unsigned LineNo = Record[Idx++]; + int FilenameID = FileIDs[Record[Idx++]]; + SrcMgr::CharacteristicKind FileKind + = (SrcMgr::CharacteristicKind)Record[Idx++]; + unsigned IncludeOffset = Record[Idx++]; + Entries.push_back(LineEntry::get(FileOffset, LineNo, FilenameID, + FileKind, IncludeOffset)); + } + LineTable.AddEntry(FID, Entries); + } + + return false; +} + +namespace { + +class PCHStatData { +public: + const bool hasStat; + const ino_t ino; + const dev_t dev; + const mode_t mode; + const time_t mtime; + const off_t size; + + PCHStatData(ino_t i, dev_t d, mode_t mo, time_t m, off_t s) + : hasStat(true), ino(i), dev(d), mode(mo), mtime(m), size(s) {} + + PCHStatData() + : hasStat(false), ino(0), dev(0), mode(0), mtime(0), size(0) {} +}; + +class PCHStatLookupTrait { + public: + typedef const char *external_key_type; + typedef const char *internal_key_type; + + typedef PCHStatData data_type; + + static unsigned ComputeHash(const char *path) { + return llvm::HashString(path); + } + + static internal_key_type GetInternalKey(const char *path) { return path; } + + static bool EqualKey(internal_key_type a, internal_key_type b) { + return strcmp(a, b) == 0; + } + + static std::pair<unsigned, unsigned> + ReadKeyDataLength(const unsigned char*& d) { + unsigned KeyLen = (unsigned) clang::io::ReadUnalignedLE16(d); + unsigned DataLen = (unsigned) *d++; + return std::make_pair(KeyLen + 1, DataLen); + } + + static internal_key_type ReadKey(const unsigned char *d, unsigned) { + return (const char *)d; + } + + static data_type ReadData(const internal_key_type, const unsigned char *d, + unsigned /*DataLen*/) { + using namespace clang::io; + + if (*d++ == 1) + return data_type(); + + ino_t ino = (ino_t) ReadUnalignedLE32(d); + dev_t dev = (dev_t) ReadUnalignedLE32(d); + mode_t mode = (mode_t) ReadUnalignedLE16(d); + time_t mtime = (time_t) ReadUnalignedLE64(d); + off_t size = (off_t) ReadUnalignedLE64(d); + return data_type(ino, dev, mode, mtime, size); + } +}; + +/// \brief stat() cache for precompiled headers. +/// +/// This cache is very similar to the stat cache used by pretokenized +/// headers. +class PCHStatCache : public StatSysCallCache { + typedef OnDiskChainedHashTable<PCHStatLookupTrait> CacheTy; + CacheTy *Cache; + + unsigned &NumStatHits, &NumStatMisses; +public: + PCHStatCache(const unsigned char *Buckets, + const unsigned char *Base, + unsigned &NumStatHits, + unsigned &NumStatMisses) + : Cache(0), NumStatHits(NumStatHits), NumStatMisses(NumStatMisses) { + Cache = CacheTy::Create(Buckets, Base); + } + + ~PCHStatCache() { delete Cache; } + + int stat(const char *path, struct stat *buf) { + // Do the lookup for the file's data in the PCH file. + CacheTy::iterator I = Cache->find(path); + + // If we don't get a hit in the PCH file just forward to 'stat'. + if (I == Cache->end()) { + ++NumStatMisses; + return StatSysCallCache::stat(path, buf); + } + + ++NumStatHits; + PCHStatData Data = *I; + + if (!Data.hasStat) + return 1; + + buf->st_ino = Data.ino; + buf->st_dev = Data.dev; + buf->st_mtime = Data.mtime; + buf->st_mode = Data.mode; + buf->st_size = Data.size; + return 0; + } +}; +} // end anonymous namespace + + +/// \brief Read a source manager block +PCHReader::PCHReadResult PCHReader::ReadSourceManagerBlock(PerFileData &F) { + using namespace SrcMgr; + + llvm::BitstreamCursor &SLocEntryCursor = F.SLocEntryCursor; + + // Set the source-location entry cursor to the current position in + // the stream. This cursor will be used to read the contents of the + // source manager block initially, and then lazily read + // source-location entries as needed. + SLocEntryCursor = F.Stream; + + // The stream itself is going to skip over the source manager block. + if (F.Stream.SkipBlock()) { + Error("malformed block record in PCH file"); + return Failure; + } + + // Enter the source manager block. + if (SLocEntryCursor.EnterSubBlock(pch::SOURCE_MANAGER_BLOCK_ID)) { + Error("malformed source manager block record in PCH file"); + return Failure; + } + + RecordData Record; + while (true) { + unsigned Code = SLocEntryCursor.ReadCode(); + if (Code == llvm::bitc::END_BLOCK) { + if (SLocEntryCursor.ReadBlockEnd()) { + Error("error at end of Source Manager block in PCH file"); + return Failure; + } + return Success; + } + + if (Code == llvm::bitc::ENTER_SUBBLOCK) { + // No known subblocks, always skip them. + SLocEntryCursor.ReadSubBlockID(); + if (SLocEntryCursor.SkipBlock()) { + Error("malformed block record in PCH file"); + return Failure; + } + continue; + } + + if (Code == llvm::bitc::DEFINE_ABBREV) { + SLocEntryCursor.ReadAbbrevRecord(); + continue; + } + + // Read a record. + const char *BlobStart; + unsigned BlobLen; + Record.clear(); + switch (SLocEntryCursor.ReadRecord(Code, Record, &BlobStart, &BlobLen)) { + default: // Default behavior: ignore. + break; + + case pch::SM_LINE_TABLE: + if (ParseLineTable(Record)) + return Failure; + break; + + case pch::SM_SLOC_FILE_ENTRY: + case pch::SM_SLOC_BUFFER_ENTRY: + case pch::SM_SLOC_INSTANTIATION_ENTRY: + // Once we hit one of the source location entries, we're done. + return Success; + } + } +} + +/// \brief Get a cursor that's correctly positioned for reading the source +/// location entry with the given ID. +llvm::BitstreamCursor &PCHReader::SLocCursorForID(unsigned ID) { + assert(ID != 0 && ID <= TotalNumSLocEntries && + "SLocCursorForID should only be called for real IDs."); + + ID -= 1; + PerFileData *F = 0; + for (unsigned I = 0, N = Chain.size(); I != N; ++I) { + F = Chain[N - I - 1]; + if (ID < F->LocalNumSLocEntries) + break; + ID -= F->LocalNumSLocEntries; + } + assert(F && F->LocalNumSLocEntries > ID && "Chain corrupted"); + + F->SLocEntryCursor.JumpToBit(F->SLocOffsets[ID]); + return F->SLocEntryCursor; +} + +/// \brief Read in the source location entry with the given ID. +PCHReader::PCHReadResult PCHReader::ReadSL |