[clang][modules] Serialize CodeGenOptions (#146422)

Some `LangOptions` duplicate their `CodeGenOptions` counterparts. My
understanding is that this was done solely because some infrastructure
(like preprocessor initialization, serialization, module compatibility
checks, etc.) were only possible/convenient for `LangOptions`. This PR
implements the missing support for `CodeGenOptions`, which makes it
possible to remove some duplicate `LangOptions` fields and simplify the
logic. Motivated by llvm/llvm-project#146342.

Author: jansvoboda11

clang-tools-extra/clangd/ModulesBuilder.cpp

@@ -219,8 +219,9 @@ bool IsModuleFileUpToDate(PathRef ModuleFilePath,
 
   IntrusiveRefCntPtr<ModuleCache> ModCache = createCrossProcessModuleCache();
   PCHContainerOperations PCHOperations;
+  CodeGenOptions CodeGenOpts;
   ASTReader Reader(PP, *ModCache, /*ASTContext=*/nullptr,
-                   PCHOperations.getRawReader(), {});
+                   PCHOperations.getRawReader(), CodeGenOpts, {});
 
   // We don't need any listener here. By default it will use a validator
   // listener.

clang/include/clang/Basic/CodeGenOptions.def

@@ -209,17 +209,17 @@ CODEGENOPT(ObjCAvoidHeapifyLocalBlocks, 1, 0, Benign)
 
 
 // The optimization options affect frontend options, which in turn do affect the AST.
-VALUE_CODEGENOPT(OptimizationLevel, 2, 0, Affecting) ///< The -O[0-3] option specified.
-VALUE_CODEGENOPT(OptimizeSize, 2, 0, Affecting) ///< If -Os (==1, Benign) or -Oz (==2, Benign) is specified.
+VALUE_CODEGENOPT(OptimizationLevel, 2, 0, Compatible) ///< The -O[0-3] option specified.
+VALUE_CODEGENOPT(OptimizeSize, 2, 0, Compatible) ///< If -Os (==1, Benign) or -Oz (==2, Benign) is specified.
 
 CODEGENOPT(AtomicProfileUpdate , 1, 0, Benign) ///< Set -fprofile-update=atomic
 CODEGENOPT(ContinuousProfileSync, 1, 0, Benign) ///< Enable continuous instrumentation profiling
 /// Choose profile instrumenation kind or no instrumentation.
 
-ENUM_CODEGENOPT(ProfileInstr, llvm::driver::ProfileInstrKind, 4, llvm::driver::ProfileInstrKind::ProfileNone, Benign)
+ENUM_CODEGENOPT(ProfileInstr, ProfileInstrKind, 4, ProfileInstrKind::ProfileNone, Benign)
 
 /// Choose profile kind for PGO use compilation.
-ENUM_CODEGENOPT(ProfileUse, llvm::driver::ProfileInstrKind, 2, llvm::driver::ProfileInstrKind::ProfileNone, Benign)
+ENUM_CODEGENOPT(ProfileUse, ProfileInstrKind, 2, ProfileInstrKind::ProfileNone, Benign)
 /// Partition functions into N groups and select only functions in group i to be
 /// instrumented. Selected group numbers can be 0 to N-1 inclusive.
 VALUE_CODEGENOPT(ProfileTotalFunctionGroups, 32, 1, Benign)
@@ -244,8 +244,8 @@ CODEGENOPT(SaveTempLabels    , 1, 0, Benign) ///< Save temporary labels.
 CODEGENOPT(SanitizeAddressUseAfterScope , 1, 0, Benign) ///< Enable use-after-scope detection
                                                         ///< in AddressSanitizer
 ENUM_CODEGENOPT(SanitizeAddressUseAfterReturn,
-                llvm::AsanDetectStackUseAfterReturnMode, 2,
-                llvm::AsanDetectStackUseAfterReturnMode::Runtime,
+                AsanDetectStackUseAfterReturnMode, 2,
+                AsanDetectStackUseAfterReturnMode::Runtime,
                 Benign
                 ) ///< Set detection mode for stack-use-after-return.
 CODEGENOPT(SanitizeAddressPoisonCustomArrayCookie, 1, 0, Benign) ///< Enable poisoning operator new[] which is not a replaceable
@@ -255,9 +255,9 @@ CODEGENOPT(SanitizeAddressGlobalsDeadStripping, 1, 0, Benign) ///< Enable linker
 CODEGENOPT(SanitizeAddressUseOdrIndicator, 1, 0, Benign) ///< Enable ODR indicator globals
 CODEGENOPT(SanitizeMemoryTrackOrigins, 2, 0, Benign) ///< Enable tracking origins in
                                                      ///< MemorySanitizer
-ENUM_CODEGENOPT(SanitizeAddressDtor, llvm::AsanDtorKind, 2,
-                llvm::AsanDtorKind::Global, Benign)  ///< Set how ASan global
-                                                     ///< destructors are emitted.
+ENUM_CODEGENOPT(SanitizeAddressDtor, AsanDtorKind, 2,
+                AsanDtorKind::Global, Benign)  ///< Set how ASan global
+                                               ///< destructors are emitted.
 CODEGENOPT(SanitizeMemoryParamRetval, 1, 0, Benign) ///< Enable detection of uninitialized
                                                     ///< parameters and return values
                                                     ///< in MemorySanitizer
@@ -375,13 +375,13 @@ VALUE_CODEGENOPT(SmallDataLimit, 32, 0, Benign)
 VALUE_CODEGENOPT(SSPBufferSize, 32, 0, Benign)
 
 /// The kind of inlining to perform.
-ENUM_CODEGENOPT(Inlining, InliningMethod, 2, NormalInlining, Benign)
+ENUM_CODEGENOPT(Inlining, InliningMethod, 2, NormalInlining, Compatible)
 
 /// The maximum stack size a function can have to be considered for inlining.
 VALUE_CODEGENOPT(InlineMaxStackSize, 32, UINT_MAX, Benign)
 
 // Vector functions library to use.
-ENUM_CODEGENOPT(VecLib, llvm::driver::VectorLibrary, 4, llvm::driver::VectorLibrary::NoLibrary, Benign)
+ENUM_CODEGENOPT(VecLib, VectorLibrary, 4, VectorLibrary::NoLibrary, Benign)
 
 /// The default TLS model to use.
 ENUM_CODEGENOPT(DefaultTLSModel, TLSModel, 2, GeneralDynamicTLSModel, Benign)
@@ -457,8 +457,8 @@ ENUM_CODEGENOPT(SwiftAsyncFramePointer, SwiftAsyncFramePointerKind, 2,
 CODEGENOPT(SkipRaxSetup, 1, 0, Benign)
 
 /// Whether to zero out caller-used registers before returning.
-ENUM_CODEGENOPT(ZeroCallUsedRegs, llvm::ZeroCallUsedRegs::ZeroCallUsedRegsKind,
-                5, llvm::ZeroCallUsedRegs::ZeroCallUsedRegsKind::Skip, Benign)
+ENUM_CODEGENOPT(ZeroCallUsedRegs, ZeroCallUsedRegsKind,
+                5, ZeroCallUsedRegsKind::Skip, Benign)
 
 /// Modify C++ ABI to returning `this` pointer from constructors and
 /// non-deleting destructors. (No effect on Microsoft ABI.)
@@ -477,8 +477,8 @@ CODEGENOPT(ResMayAlias, 1, 0, Benign)
 
 /// Controls how unwind v2 (epilog) information should be generated for x64
 /// Windows.
-ENUM_CODEGENOPT(WinX64EHUnwindV2, llvm::WinX64EHUnwindV2Mode,
-                2, llvm::WinX64EHUnwindV2Mode::Disabled, Benign)
+ENUM_CODEGENOPT(WinX64EHUnwindV2, WinX64EHUnwindV2Mode,
+                2, WinX64EHUnwindV2Mode::Disabled, Benign)
 
 /// FIXME: Make DebugOptions its own top-level .def file.
 #include "DebugOptions.def"

clang/include/clang/Basic/CodeGenOptions.h

@@ -46,13 +46,32 @@ class CodeGenOptionsBase {
   enum class CompatibilityKind {
     /// Does affect the construction of the AST in a way that does prevent
     /// module interoperability.
-    Affecting,
+    NotCompatible,
+    /// Does affect the construction of the AST in a way that doesn't prevent
+    /// interoperability (that is, the value can be different between an
+    /// explicit module and the user of that module).
+    Compatible,
     /// Does not affect the construction of the AST in any way (that is, the
     /// value can be different between an implicit module and the user of that
     /// module).
     Benign,
   };
 
+  using CFBranchLabelSchemeKind = CFBranchLabelSchemeKind;
+  using ProfileInstrKind = llvm::driver::ProfileInstrKind;
+  using AsanDetectStackUseAfterReturnMode =
+      llvm::AsanDetectStackUseAfterReturnMode;
+  using AsanDtorKind = llvm::AsanDtorKind;
+  using VectorLibrary = llvm::driver::VectorLibrary;
+  using ZeroCallUsedRegsKind = llvm::ZeroCallUsedRegs::ZeroCallUsedRegsKind;
+  using WinX64EHUnwindV2Mode = llvm::WinX64EHUnwindV2Mode;
+
+  using DebugCompressionType = llvm::DebugCompressionType;
+  using EmitDwarfUnwindType = llvm::EmitDwarfUnwindType;
+  using DebugTemplateNamesKind = llvm::codegenoptions::DebugTemplateNamesKind;
+  using DebugInfoKind = llvm::codegenoptions::DebugInfoKind;
+  using DebuggerKind = llvm::DebuggerKind;
+
 #define CODEGENOPT(Name, Bits, Default, Compatibility) unsigned Name : Bits;
 #define ENUM_CODEGENOPT(Name, Type, Bits, Default, Compatibility)
 #include "clang/Basic/CodeGenOptions.def"

clang/include/clang/Basic/DebugOptions.def

@@ -28,9 +28,9 @@ VALUE_CODEGENOPT(Name, Bits, Default, Compatibility)
 ENUM_CODEGENOPT(Name, Type, Bits, Default, Compatibility)
 #endif
 
-ENUM_DEBUGOPT(CompressDebugSections, llvm::DebugCompressionType, 2,
-                     llvm::DebugCompressionType::None, Benign)
-DEBUGOPT(Dwarf64, 1, 0, Affecting) ///< -gdwarf64.
+ENUM_DEBUGOPT(CompressDebugSections, DebugCompressionType, 2,
+                     DebugCompressionType::None, Benign)
+DEBUGOPT(Dwarf64, 1, 0, Compatible) ///< -gdwarf64.
 DEBUGOPT(EnableDIPreservationVerify, 1, 0, Benign) ///< Enable di preservation
                                                    ///< verify each (it means
                                                    ///< check the original debug
@@ -40,17 +40,17 @@ DEBUGOPT(ForceDwarfFrameSection , 1, 0, Benign) ///< Set when -fforce-dwarf-fram
                                                 ///< is enabled.
 
 ///< Set when -femit-dwarf-unwind is passed.
-ENUM_DEBUGOPT(EmitDwarfUnwind, llvm::EmitDwarfUnwindType, 2,
-              llvm::EmitDwarfUnwindType::Default, Benign)
+ENUM_DEBUGOPT(EmitDwarfUnwind, EmitDwarfUnwindType, 2,
+              EmitDwarfUnwindType::Default, Benign)
 
 DEBUGOPT(NoDwarfDirectoryAsm , 1, 0, Benign) ///< Set when -fno-dwarf-directory-asm
                                              ///< is enabled.
 
 DEBUGOPT(NoInlineLineTables, 1, 0, Benign) ///< Whether debug info should contain
                                            ///< inline line tables.
 
-DEBUGOPT(DebugStrictDwarf, 1, 1, Affecting) ///< Whether or not to use strict DWARF info.
-DEBUGOPT(DebugOmitUnreferencedMethods, 1, 0, Affecting) ///< Omit unreferenced member
+DEBUGOPT(DebugStrictDwarf, 1, 1, Compatible) ///< Whether or not to use strict DWARF info.
+DEBUGOPT(DebugOmitUnreferencedMethods, 1, 0, Compatible) ///< Omit unreferenced member
                                                         ///< functions in type debug info.
 
 /// Control the Assignment Tracking debug info feature.
@@ -60,73 +60,73 @@ ENUM_DEBUGOPT(AssignmentTrackingMode, AssignmentTrackingOpts, 2,
 /// Whether or not to use Key Instructions to determine breakpoint locations.
 DEBUGOPT(DebugKeyInstructions, 1, 0, Benign)
 
-DEBUGOPT(DebugColumnInfo, 1, 0, Affecting) ///< Whether or not to use column information
+DEBUGOPT(DebugColumnInfo, 1, 0, Compatible) ///< Whether or not to use column information
                                            ///< in debug info.
 
-DEBUGOPT(DebugTypeExtRefs, 1, 0, Affecting) ///< Whether or not debug info should contain
+DEBUGOPT(DebugTypeExtRefs, 1, 0, Compatible) ///< Whether or not debug info should contain
                                             ///< external references to a PCH or module.
 
-DEBUGOPT(DebugExplicitImport, 1, 0, Affecting)  ///< Whether or not debug info should
+DEBUGOPT(DebugExplicitImport, 1, 0, Compatible)  ///< Whether or not debug info should
                                                 ///< contain explicit imports for
                                                 ///< anonymous namespaces
 
 /// Set debug info source file hashing algorithm.
-ENUM_DEBUGOPT(DebugSrcHash, DebugSrcHashKind, 2, DSH_MD5, Affecting)
+ENUM_DEBUGOPT(DebugSrcHash, DebugSrcHashKind, 2, DSH_MD5, Compatible)
 
-DEBUGOPT(SplitDwarfInlining, 1, 1, Affecting) ///< Whether to include inlining info in the
+DEBUGOPT(SplitDwarfInlining, 1, 1, Compatible) ///< Whether to include inlining info in the
                                               ///< skeleton CU to allow for symbolication
                                               ///< of inline stack frames without .dwo files.
-DEBUGOPT(DebugFwdTemplateParams, 1, 0, Affecting) ///< Whether to emit complete
+DEBUGOPT(DebugFwdTemplateParams, 1, 0, Compatible) ///< Whether to emit complete
                                                   ///< template parameter descriptions in
                                                   ///< forward declarations (versus just
                                                   ///< including them in the name).
 ENUM_DEBUGOPT(DebugSimpleTemplateNames,
-              llvm::codegenoptions::DebugTemplateNamesKind, 2,
-              llvm::codegenoptions::DebugTemplateNamesKind::Full, Affecting)
+              DebugTemplateNamesKind, 2,
+              DebugTemplateNamesKind::Full, Compatible)
               ///< Whether to emit template parameters in the textual names of
               ///< template specializations.
               ///< Implies DebugFwdTemplateNames to allow decorated names to be
               ///< reconstructed when needed.
 
 /// The kind of generated debug info.
-ENUM_DEBUGOPT(DebugInfo, llvm::codegenoptions::DebugInfoKind, 4,
-              llvm::codegenoptions::NoDebugInfo, Affecting)
+ENUM_DEBUGOPT(DebugInfo, DebugInfoKind, 4,
+              DebugInfoKind::NoDebugInfo, Compatible)
 
 /// Whether to generate macro debug info.
-DEBUGOPT(MacroDebugInfo, 1, 0, Affecting)
+DEBUGOPT(MacroDebugInfo, 1, 0, Compatible)
 
 /// Tune the debug info for this debugger.
-ENUM_DEBUGOPT(DebuggerTuning, llvm::DebuggerKind, 3,
-              llvm::DebuggerKind::Default, Affecting)
+ENUM_DEBUGOPT(DebuggerTuning, DebuggerKind, 3,
+              DebuggerKind::Default, Compatible)
 
 /// Dwarf version. Version zero indicates to LLVM that no DWARF should be
 /// emitted.
-VALUE_DEBUGOPT(DwarfVersion, 3, 0, Affecting)
+VALUE_DEBUGOPT(DwarfVersion, 3, 0, Compatible)
 
 /// Whether we should emit CodeView debug information. It's possible to emit
 /// CodeView and DWARF into the same object.
-DEBUGOPT(EmitCodeView, 1, 0, Affecting)
+DEBUGOPT(EmitCodeView, 1, 0, Compatible)
 
 /// Whether to emit the .debug$H section containing hashes of CodeView types.
-DEBUGOPT(CodeViewGHash, 1, 0, Affecting)
+DEBUGOPT(CodeViewGHash, 1, 0, Compatible)
 
 /// Whether to emit the compiler path and command line into the CodeView debug information.
-DEBUGOPT(CodeViewCommandLine, 1, 0, Affecting)
+DEBUGOPT(CodeViewCommandLine, 1, 0, Compatible)
 
 /// Whether emit extra debug info for sample pgo profile collection.
-DEBUGOPT(DebugInfoForProfiling, 1, 0, Affecting)
+DEBUGOPT(DebugInfoForProfiling, 1, 0, Compatible)
 
 /// Whether to emit DW_TAG_template_alias for template aliases.
-DEBUGOPT(DebugTemplateAlias, 1, 0, Affecting)
+DEBUGOPT(DebugTemplateAlias, 1, 0, Compatible)
 
 /// Whether to emit .debug_gnu_pubnames section instead of .debug_pubnames.
-DEBUGOPT(DebugNameTable, 2, 0, Affecting)
+DEBUGOPT(DebugNameTable, 2, 0, Compatible)
 
 /// Whether to use DWARF base address specifiers in .debug_ranges.
-DEBUGOPT(DebugRangesBaseAddress, 1, 0, Affecting)
+DEBUGOPT(DebugRangesBaseAddress, 1, 0, Compatible)
 
 /// Whether to embed source in DWARF debug line section.
-DEBUGOPT(EmbedSource, 1, 0, Affecting)
+DEBUGOPT(EmbedSource, 1, 0, Compatible)
 
 #undef DEBUGOPT
 #undef ENUM_DEBUGOPT

clang/include/clang/Basic/DiagnosticSerializationKinds.td

@@ -41,6 +41,10 @@ def err_ast_file_langopt_mismatch : Error<"%0 was %select{disabled|enabled}1 in
     "precompiled file '%3' but is currently %select{disabled|enabled}2">;
 def err_ast_file_langopt_value_mismatch : Error<
   "%0 differs in precompiled file '%1' vs. current file">;
+def err_ast_file_codegenopt_mismatch : Error<"%0 was %select{disabled|enabled}1 in "
+  "precompiled file '%3' but is currently %select{disabled|enabled}2">;
+def err_ast_file_codegenopt_value_mismatch
+    : Error<"%0 differs in precompiled file '%1' vs. current file">;
 def err_ast_file_diagopt_mismatch : Error<"%0 is currently enabled, but was not in "
   "the precompiled file '%1'">;
 def err_ast_file_modulecache_mismatch : Error<"precompiled file '%2' was compiled with module cache "

clang/include/clang/Basic/LangOptions.def

@@ -166,8 +166,6 @@ LANGOPT(ModulesValidateTextualHeaderIncludes, 1, 1, Compatible, "validation of t
 LANGOPT(ModulesErrorRecovery, 1, 1, Benign, "automatically importing modules as needed when performing error recovery")
 LANGOPT(ImplicitModules, 1, 1, Benign, "building modules that are not specified via -fmodule-file")
 LANGOPT(ModulesLocalVisibility, 1, 0, Compatible, "local submodule visibility")
-LANGOPT(Optimize          , 1, 0, Compatible, "__OPTIMIZE__ predefined macro")
-LANGOPT(OptimizeSize      , 1, 0, Compatible, "__OPTIMIZE_SIZE__ predefined macro")
 LANGOPT(Static            , 1, 0, Compatible, "__STATIC__ predefined macro (as opposed to __DYNAMIC__)")
 VALUE_LANGOPT(PackStruct  , 32, 0, NotCompatible,
               "default struct packing maximum alignment")
@@ -184,7 +182,6 @@ VALUE_LANGOPT(PIE                    , 1, 0, Compatible, "is pie")
 LANGOPT(ROPI                         , 1, 0, NotCompatible, "Read-only position independence")
 LANGOPT(RWPI                         , 1, 0, NotCompatible, "Read-write position independence")
 LANGOPT(GNUInline                    , 1, 0, Compatible, "GNU inline semantics")
-LANGOPT(NoInlineDefine               , 1, 0, Compatible, "__NO_INLINE__ predefined macro")
 LANGOPT(Deprecated                   , 1, 0, Compatible, "__DEPRECATED predefined macro")
 LANGOPT(FastMath                     , 1, 0, Compatible, "fast FP math optimizations, and __FAST_MATH__ predefined macro")
 LANGOPT(UnsafeFPMath                 , 1, 0, Compatible, "Unsafe Floating Point Math")

clang/include/clang/Frontend/ASTUnit.h

@@ -62,6 +62,7 @@ class ASTContext;
 class ASTDeserializationListener;
 class ASTMutationListener;
 class ASTReader;
+class CodeGenOptions;
 class CompilerInstance;
 class CompilerInvocation;
 class Decl;
@@ -107,6 +108,7 @@ class ASTUnit {
 
 private:
   std::unique_ptr<LangOptions> LangOpts;
+  std::unique_ptr<CodeGenOptions> CodeGenOpts;
   // FIXME: The documentation on \c LoadFrom* member functions states that the
   // DiagnosticsEngine (and therefore DiagnosticOptions) must outlive the
   // returned ASTUnit. This is not the case. Enfore it by storing non-owning

clang/include/clang/Frontend/CompilerInstance.h

@@ -720,6 +720,7 @@ class CompilerInstance : public ModuleLoader {
       DisableValidationForModuleKind DisableValidation,
       bool AllowPCHWithCompilerErrors, Preprocessor &PP, ModuleCache &ModCache,
       ASTContext &Context, const PCHContainerReader &PCHContainerRdr,
+      const CodeGenOptions &CodeGenOpts,
       ArrayRef<std::shared_ptr<ModuleFileExtension>> Extensions,
       ArrayRef<std::shared_ptr<DependencyCollector>> DependencyCollectors,
       void *DeserializationListener, bool OwnDeserializationListener,

clang/include/clang/Serialization/ASTBitCodes.h

@@ -44,7 +44,7 @@ namespace serialization {
 /// Version 4 of AST files also requires that the version control branch and
 /// revision match exactly, since there is no backward compatibility of
 /// AST files at this time.
-const unsigned VERSION_MAJOR = 34;
+const unsigned VERSION_MAJOR = 35;
 
 /// AST file minor version number supported by this version of
 /// Clang.
@@ -399,6 +399,9 @@ enum OptionsRecordTypes {
 
   /// Record code for the preprocessor options table.
   PREPROCESSOR_OPTIONS,
+
+  /// Record code for the codegen options table.
+  CODEGEN_OPTIONS,
 };
 
 /// Record codes for the unhashed control block.