[Explicit Module Builds] Ensure IRGen uses Swift compiler's target triple when '-clang-target' is set.

The Clang importer's Clang instance may be configured with a different (higher)
OS version than the compilation target itself in order to be able to load
pre-compiled Clang modules that are aligned with the broader SDK, and match the
SDK deployment target against which Swift modules are also built. In this case,
we must use the Swift compiler's OS version triple in order to generate the
binary as-requested.

This change makes 'ClangImporter' 'Implementation' keep track of a distinct
'TargetInfo' and 'CodeGenOpts' containers that are meant to be used by clients
in IRGen. When '-clang-target' is not set, they are defined to be copies of the
'ClangImporter's built-in module-loading Clang instance. When '-clang-target' is
set, they are configured with the Swift compilation's target triple and OS
version (but otherwise identical) instead. To distinguish IRGen clients from
module loading clients, 'getModuleAvailabilityTarget' is added for module
loading clients of 'ClangImporter'.

The notion of using a different triple for loading Clang modules arises for the
following reason:
- Swift is able to load Swift modules built against a different target triple
  than the source module that is being compiled. Swift relies on availability
  annotations on the API within the loaded modules to ensure that compilation
  for the current target only uses appropriately-available API from its
  dependencies.
- Clang, in contrast, requires that compilation only ever load modules (.pcm)
  that are precisely aligned to the current source compilation. Because the
  target triple (OS version in particular) between Swift source compilation and
  Swift dependency module compilation may differ, this would otherwise result in
  builtin multiple copies of the same Clang module, against different OS
  versions, once for each different triple in the build graph.
Instead, with Explicitly-Built Modules, Swift sets a '-clang-target' argument
that ensures that all Clang modules participating in the build are built against
the SDK deployment target, matching the Swift modules in the SDK, which allows
them to expose a maximally-available API surface as required by
potentially-depending Swift modules' target OS version.
--------------------------------------------
For example:
Suppose we are building a source module 'Foo', targeting 'macosx10.0', using an
SDK with a deployment target of 'macosx12.0'. Swift modules in said SDK will be
built for 'macosx12.0' (as hard-coded in their textual interfaces), meaning they
may reference symbols expected to be present in dependency Clang modules at that
target OS version.

Suppose the source module 'Foo' depends on Swift module 'Bar', which then
depends on Clang module `Baz`. 'Bar' must be built targeting 'macosx12.0'
(SDK-matching deployment target is hard-coded into its textual interface). Which
means that 'Bar' expects 'Baz' to expose symbols that may only be available when
targeting at least 'macosx12.0'. e.g. 'Baz' may have symbols guarded with
'__MAC_OS_X_VERSION_MIN_REQUIRED >= __MAC_12_0'. For this reason, we use
'-clang-target' to ensure 'Baz' is built targeting 'macosx12.0', and can be
loaded by both 'Foo' and 'Bar'.

As a result, we cannot direclty use the Clang instance's target triple here and
must check if we need to instead use the triple of the Swift compiler instance.

Resolves rdar://109228963

Author: artemcm

include/swift/AST/ClangModuleLoader.h

@@ -124,7 +124,21 @@ class ClangModuleLoader : public ModuleLoader {
   using ModuleLoader::ModuleLoader;
 
 public:
-  virtual clang::TargetInfo &getTargetInfo() const = 0;
+  /// This module loader's Clang instance may be configured with a different
+  /// (higher) OS version than the compilation target itself in order to be able
+  /// to load pre-compiled Clang modules that are aligned with the broader SDK,
+  /// and match the SDK deployment target against which Swift modules are also
+  /// built.
+  ///
+  /// In this case, we must use the Swift compiler's OS version triple when
+  /// performing codegen, and the importer's Clang instance OS version triple
+  /// during module loading. `getModuleAvailabilityTarget` is for module-loading
+  /// clients only, and uses the latter.
+  ///
+  /// (The implementing `ClangImporter` class maintains separate Target info
+  /// for use by IRGen/CodeGen clients)
+  virtual clang::TargetInfo &getModuleAvailabilityTarget() const = 0;
+
   virtual clang::ASTContext &getClangASTContext() const = 0;
   virtual clang::Preprocessor &getClangPreprocessor() const = 0;
   virtual clang::Sema &getClangSema() const = 0;

include/swift/ClangImporter/ClangImporter.h

@@ -50,6 +50,7 @@ namespace clang {
   class VisibleDeclConsumer;
   class DeclarationName;
   class CompilerInvocation;
+  class TargetOptions;
 namespace tooling {
 namespace dependencies {
   struct ModuleDeps;
@@ -173,7 +174,7 @@ class ClangImporter final : public ClangModuleLoader {
          DWARFImporterDelegate *dwarfImporterDelegate = nullptr);
 
   static std::vector<std::string>
-  getClangArguments(ASTContext &ctx);
+  getClangArguments(ASTContext &ctx, bool ignoreClangTarget = false);
 
   static std::unique_ptr<clang::CompilerInvocation>
   createClangInvocation(ClangImporter *importer,
@@ -445,13 +446,33 @@ class ClangImporter final : public ClangModuleLoader {
       StringRef moduleName,
       ModuleDependencyKind moduleKind,
       ModuleDependenciesCache &cache);
-
-  clang::TargetInfo &getTargetInfo() const override;
+  clang::TargetInfo &getModuleAvailabilityTarget() const override;
   clang::ASTContext &getClangASTContext() const override;
   clang::Preprocessor &getClangPreprocessor() const override;
   clang::Sema &getClangSema() const override;
   const clang::CompilerInstance &getClangInstance() const override;
-  clang::CodeGenOptions &getClangCodeGenOpts() const;
+
+  /// ClangImporter's Clang instance may be configured with a different
+  /// (higher) OS version than the compilation target itself in order to be able
+  /// to load pre-compiled Clang modules that are aligned with the broader SDK,
+  /// and match the SDK deployment target against which Swift modules are also
+  /// built.
+  ///
+  /// In this case, we must use the Swift compiler's OS version triple when
+  /// performing codegen, and the importer's Clang instance OS version triple
+  /// during module loading.
+  ///
+  /// `ClangImporter`'s `Implementation` keeps track of a distinct `TargetInfo`
+  /// and `CodeGenOpts` containers that are meant to be used by clients in
+  /// IRGen. When a separate `-clang-target` is not set, they are defined to be
+  /// copies of the `ClangImporter`'s built-in module-loading Clang instance.
+  /// When `-clang-target` is set, they are configured with the Swift
+  /// compilation's target triple and OS version (but otherwise identical)
+  /// instead. To distinguish IRGen clients from module loading clients,
+  /// `getModuleAvailabilityTarget` should be used instead by module-loading
+  /// clients.
+  clang::TargetInfo &getTargetInfo() const;
+  clang::CodeGenOptions &getCodeGenOpts() const;
 
   std::string getClangModuleHash() const;
 

lib/ClangImporter/ClangImporter.cpp

@@ -718,11 +718,11 @@ getEmbedBitcodeInvocationArguments(std::vector<std::string> &invocationArgStrs,
 void
 importer::addCommonInvocationArguments(
     std::vector<std::string> &invocationArgStrs,
-    ASTContext &ctx) {
+    ASTContext &ctx, bool ignoreClangTarget) {
   using ImporterImpl = ClangImporter::Implementation;
   llvm::Triple triple = ctx.LangOpts.Target;
   // Use clang specific target triple if given.
-  if (ctx.LangOpts.ClangTarget.has_value()) {
+  if (ctx.LangOpts.ClangTarget.has_value() && !ignoreClangTarget) {
     triple = ctx.LangOpts.ClangTarget.value();
   }
   SearchPathOptions &searchPathOpts = ctx.SearchPathOpts;
@@ -970,7 +970,7 @@ Optional<std::string> ClangImporter::getOrCreatePCH(
 }
 
 std::vector<std::string>
-ClangImporter::getClangArguments(ASTContext &ctx) {
+ClangImporter::getClangArguments(ASTContext &ctx, bool ignoreClangTarget) {
   std::vector<std::string> invocationArgStrs;
   // Clang expects this to be like an actual command line. So we need to pass in
   // "clang" for argv[0]
@@ -991,7 +991,7 @@ ClangImporter::getClangArguments(ASTContext &ctx) {
     getEmbedBitcodeInvocationArguments(invocationArgStrs, ctx);
     break;
   }
-  addCommonInvocationArguments(invocationArgStrs, ctx);
+  addCommonInvocationArguments(invocationArgStrs, ctx, ignoreClangTarget);
   return invocationArgStrs;
 }
 
@@ -1094,15 +1094,6 @@ ClangImporter::create(ASTContext &ctx,
   std::unique_ptr<ClangImporter> importer{
       new ClangImporter(ctx, tracker, dwarfImporterDelegate)};
   auto &importerOpts = ctx.ClangImporterOpts;
-  importer->Impl.ClangArgs = getClangArguments(ctx);
-  ArrayRef<std::string> invocationArgStrs = importer->Impl.ClangArgs;
-  if (importerOpts.DumpClangDiagnostics) {
-    llvm::errs() << "'";
-    llvm::interleave(
-                     invocationArgStrs, [](StringRef arg) { llvm::errs() << arg; },
-                     [] { llvm::errs() << "' '"; });
-    llvm::errs() << "'\n";
-  }
 
   if (isPCHFilenameExtension(importerOpts.BridgingHeader)) {
     importer->Impl.setSinglePCHImport(importerOpts.BridgingHeader);
@@ -1142,6 +1133,15 @@ ClangImporter::create(ASTContext &ctx,
 
   // Create a new Clang compiler invocation.
   {
+    importer->Impl.ClangArgs = getClangArguments(ctx);
+    ArrayRef<std::string> invocationArgStrs = importer->Impl.ClangArgs;
+    if (importerOpts.DumpClangDiagnostics) {
+      llvm::errs() << "'";
+      llvm::interleave(
+                       invocationArgStrs, [](StringRef arg) { llvm::errs() << arg; },
+                       [] { llvm::errs() << "' '"; });
+      llvm::errs() << "'\n";
+    }
     importer->Impl.Invocation = createClangInvocation(
         importer.get(), importerOpts, VFS, invocationArgStrs);
     if (!importer->Impl.Invocation)
@@ -1217,6 +1217,27 @@ ClangImporter::create(ASTContext &ctx,
                          clang::SourceLocation());
   clangDiags.setFatalsAsError(ctx.Diags.getShowDiagnosticsAfterFatalError());
 
+  // Use Clang to configure/save options for Swift IRGen/CodeGen
+  if (ctx.LangOpts.ClangTarget.has_value()) {
+    // If '-clang-target' is set, create a mock invocation with the Swift triple
+    // to configure CodeGen and Target options for Swift compilation.
+    auto swiftTargetClangArgs = getClangArguments(ctx, true);
+    ArrayRef<std::string> invocationArgStrs = swiftTargetClangArgs;
+    auto swiftTargetClangInvocation = createClangInvocation(
+        importer.get(), importerOpts, VFS, invocationArgStrs);
+    if (!swiftTargetClangInvocation)
+      return nullptr;
+    importer->Impl.setSwiftTargetInfo(clang::TargetInfo::CreateTargetInfo(
+        clangDiags, swiftTargetClangInvocation->TargetOpts));
+    importer->Impl.setSwiftCodeGenOptions(new clang::CodeGenOptions(
+        swiftTargetClangInvocation->getCodeGenOpts()));
+  } else {
+    // Just use the existing Invocation's directly
+    importer->Impl.setSwiftTargetInfo(clang::TargetInfo::CreateTargetInfo(
+        clangDiags, importer->Impl.Invocation->TargetOpts));
+    importer->Impl.setSwiftCodeGenOptions(
+        new clang::CodeGenOptions(importer->Impl.Invocation->getCodeGenOpts()));
+  }
 
   // Create the associated action.
   importer->Impl.Action.reset(new ParsingAction(ctx, *importer,
@@ -1872,7 +1893,7 @@ bool ClangImporter::canImportModule(ImportPath::Module modulePath,
   clang::Module *m;
   auto &ctx = Impl.getClangASTContext();
   auto &lo = ctx.getLangOpts();
-  auto &ti = getTargetInfo();
+  auto &ti = getModuleAvailabilityTarget();
 
   auto available = clangModule->isAvailable(lo, ti, r, mh, m);
   if (!available)
@@ -3687,10 +3708,14 @@ StringRef ClangModuleUnit::getLoadedFilename() const {
   return StringRef();
 }
 
-clang::TargetInfo &ClangImporter::getTargetInfo() const {
+clang::TargetInfo &ClangImporter::getModuleAvailabilityTarget() const {
   return Impl.Instance->getTarget();
 }
 
+clang::TargetInfo &ClangImporter::getTargetInfo() const {
+  return *Impl.getSwiftTargetInfo();
+}
+
 clang::ASTContext &ClangImporter::getClangASTContext() const {
   return Impl.getClangASTContext();
 }
@@ -3720,8 +3745,8 @@ clang::Sema &ClangImporter::getClangSema() const {
   return Impl.getClangSema();
 }
 
-clang::CodeGenOptions &ClangImporter::getClangCodeGenOpts() const {
-  return Impl.getClangCodeGenOpts();
+clang::CodeGenOptions &ClangImporter::getCodeGenOpts() const {
+  return *Impl.getSwiftCodeGenOptions();
 }
 
 std::string ClangImporter::getClangModuleHash() const {

lib/ClangImporter/ImporterImpl.h

@@ -39,6 +39,7 @@
 #include "clang/AST/DeclVisitor.h"
 #include "clang/AST/RecursiveASTVisitor.h"
 #include "clang/Basic/IdentifierTable.h"
+#include "clang/Basic/TargetInfo.h"
 #include "clang/Frontend/CompilerInstance.h"
 #include "clang/Serialization/ModuleFileExtension.h"
 #include "llvm/ADT/APSInt.h"
@@ -588,6 +589,29 @@ class LLVM_LIBRARY_VISIBILITY ClangImporter::Implementation
     return Instance.get();
   }
 
+private:
+  /// The Importer may be configured to load modules of a different OS Version
+  /// than the underlying Swift compilation. This is the `TargetOptions`
+  /// corresponding to the instantiating Swift compilation's triple. These are
+  /// to be used by all IRGen/CodeGen clients of `ClangImporter`.
+  std::unique_ptr<clang::TargetInfo> CodeGenTargetInfo;
+  std::unique_ptr<clang::CodeGenOptions> CodeGenOpts;
+
+public:
+  void setSwiftTargetInfo(clang::TargetInfo *SwiftTargetInfo) {
+    CodeGenTargetInfo.reset(SwiftTargetInfo);
+  }
+  clang::TargetInfo *getSwiftTargetInfo() const {
+    return CodeGenTargetInfo.get();
+  }
+
+  void setSwiftCodeGenOptions(clang::CodeGenOptions *SwiftCodeGenOpts) {
+    CodeGenOpts.reset(SwiftCodeGenOpts);
+  }
+  clang::CodeGenOptions *getSwiftCodeGenOptions() const {
+    return CodeGenOpts.get();
+  }
+
 private:
   /// Generation number that is used for crude versioning.
   ///
@@ -835,7 +859,7 @@ class LLVM_LIBRARY_VISIBILITY ClangImporter::Implementation
     return Instance->getPreprocessor();
   }
 
-  clang::CodeGenOptions &getClangCodeGenOpts() const {
+  clang::CodeGenOptions &getCodeGenOpts() const {
     return Instance->getCodeGenOpts();
   }
 
@@ -1814,7 +1838,8 @@ void getNormalInvocationArguments(std::vector<std::string> &invocationArgStrs,
 
 /// Add command-line arguments common to all imports of Clang code.
 void addCommonInvocationArguments(std::vector<std::string> &invocationArgStrs,
-                                  ASTContext &ctx);
+                                  ASTContext &ctx,
+                                  bool ignoreClangTarget);
 
 /// Finds a particular kind of nominal by looking through typealiases.
 template <typename T>

lib/IRGen/IRGenModule.cpp

@@ -86,7 +86,7 @@ static llvm::StructType *createStructType(IRGenModule &IGM,
 }
 
 static clang::CodeGenerator *createClangCodeGenerator(ASTContext &Context,
-                                                 llvm::LLVMContext &LLVMContext,
+                                                      llvm::LLVMContext &LLVMContext,
                                                       const IRGenOptions &Opts,
                                                       StringRef ModuleName,
                                                       StringRef PD) {
@@ -95,7 +95,7 @@ static clang::CodeGenerator *createClangCodeGenerator(ASTContext &Context,
   assert(Importer && "No clang module loader!");
   auto &ClangContext = Importer->getClangASTContext();
 
-  auto &CGO = Importer->getClangCodeGenOpts();
+  auto &CGO = Importer->getCodeGenOpts();
   if (CGO.OpaquePointers) {
     LLVMContext.setOpaquePointers(true);
   } else {