Backport RadixTree.h and improve PDB generation

clang/docs/doxygen.cfg.in

@@ -1126,7 +1126,7 @@ HTML_TIMESTAMP         = YES
 # The default value is: NO.
 # This tag requires that the tag GENERATE_HTML is set to YES.
 
-HTML_DYNAMIC_SECTIONS  = NO
+HTML_DYNAMIC_SECTIONS  = YES
 
 # With HTML_INDEX_NUM_ENTRIES one can control the preferred number of entries
 # shown in the various tree structured indices initially; the user can expand
@@ -1153,7 +1153,7 @@ HTML_INDEX_NUM_ENTRIES = 100
 # The default value is: NO.
 # This tag requires that the tag GENERATE_HTML is set to YES.
 
-GENERATE_DOCSET        = NO
+GENERATE_DOCSET        = YES
 
 # This tag determines the name of the docset feed. A documentation feed provides
 # an umbrella under which multiple documentation sets from a single provider
@@ -1169,15 +1169,15 @@ DOCSET_FEEDNAME        = "Doxygen generated docs"
 # The default value is: org.doxygen.Project.
 # This tag requires that the tag GENERATE_DOCSET is set to YES.
 
-DOCSET_BUNDLE_ID       = org.doxygen.Project
+DOCSET_BUNDLE_ID       = clang
 
 # The DOCSET_PUBLISHER_ID tag specifies a string that should uniquely identify
 # the documentation publisher. This should be a reverse domain-name style
 # string, e.g. com.mycompany.MyDocSet.documentation.
 # The default value is: org.doxygen.Publisher.
 # This tag requires that the tag GENERATE_DOCSET is set to YES.
 
-DOCSET_PUBLISHER_ID    = org.doxygen.Publisher
+DOCSET_PUBLISHER_ID    = Publisher
 
 # The DOCSET_PUBLISHER_NAME tag identifies the documentation publisher.
 # The default value is: Publisher.
@@ -1252,7 +1252,7 @@ TOC_EXPAND             = NO
 # The default value is: NO.
 # This tag requires that the tag GENERATE_HTML is set to YES.
 
-GENERATE_QHP           = @clang_doxygen_generate_qhp@
+GENERATE_QHP           = YES
 
 # If the QHG_LOCATION tag is specified, the QCH_FILE tag can be used to specify
 # the file name of the resulting .qch file. The path specified is relative to
@@ -1267,7 +1267,7 @@ QCH_FILE               = @clang_doxygen_qch_filename@
 # The default value is: org.doxygen.Project.
 # This tag requires that the tag GENERATE_QHP is set to YES.
 
-QHP_NAMESPACE          = @clang_doxygen_qhp_namespace@
+QHP_NAMESPACE          = clang
 
 # The QHP_VIRTUAL_FOLDER tag specifies the namespace to use when generating Qt
 # Help Project output. For more information please see Qt Help Project / Virtual
@@ -1337,7 +1337,7 @@ ECLIPSE_DOC_ID         = org.doxygen.Project
 # The default value is: NO.
 # This tag requires that the tag GENERATE_HTML is set to YES.
 
-DISABLE_INDEX          = NO
+DISABLE_INDEX          = YES
 
 # The GENERATE_TREEVIEW tag is used to specify whether a tree-like index
 # structure should be generated to display hierarchical information. If the tag
@@ -1468,7 +1468,7 @@ MATHJAX_CODEFILE       =
 # The default value is: YES.
 # This tag requires that the tag GENERATE_HTML is set to YES.
 
-SEARCHENGINE           = @enable_searchengine@
+SEARCHENGINE           = NO
 
 # When the SERVER_BASED_SEARCH tag is enabled the search engine will be
 # implemented using a web server instead of a web client using Javascript. There
@@ -2090,7 +2090,7 @@ CLASS_GRAPH            = YES
 # The default value is: YES.
 # This tag requires that the tag HAVE_DOT is set to YES.
 
-COLLABORATION_GRAPH    = YES
+COLLABORATION_GRAPH    = NO
 
 # If the GROUP_GRAPHS tag is set to YES then doxygen will generate a graph for
 # groups, showing the direct groups dependencies.

clang/include/clang/AST/OpenMPClause.h

@@ -8851,7 +8851,7 @@ struct TargetOMPContext final : public llvm::omp::OMPContext {
                    const FunctionDecl *CurrentFunctionDecl,
                    ArrayRef<llvm::omp::TraitProperty> ConstructTraits);
 
-  virtual ~TargetOMPContext() = default;
+  virtual ~TargetOMPContext() override = default;
 
   /// See llvm::omp::OMPContext::matchesISATrait
   bool matchesISATrait(StringRef RawString) const override;

clang/include/clang/Basic/CodeGenOptions.h

@@ -398,6 +398,10 @@ class CodeGenOptions : public CodeGenOptionsBase {
   /// by sanitizer coverage pass.
   std::vector<std::string> SanitizeCoverageAllowlistFiles;
 
+  /// Files whose declarations get rich debug info (types retained and
+  /// declaration-only subprograms emitted).
+  std::vector<std::string> DebugInfoAllowedFiles;
+
   /// The guard style used for stack protector to get a initial value, this
   /// value usually be gotten from TLS or get from __stack_chk_guard, or some
   /// other styles we may implement in the future.

clang/include/clang/Basic/FileSystemOptions.h

@@ -24,6 +24,11 @@ class FileSystemOptions {
   /// If set, paths are resolved as if the working directory was
   /// set to the value of WorkingDir.
   std::string WorkingDir;
+
+  /// If set, file lookups that fail are retried with a case-insensitive
+  /// directory scan.  Useful for compiling against SDKs whose header paths
+  /// use a different case than the #include directives.
+  bool CaseInsensitivePaths = false;
 };
 
 } // end namespace clang

clang/include/clang/Driver/Options.td

@@ -1282,6 +1282,11 @@ def fdebug_compilation_dir_EQ : Joined<["-"], "fdebug-compilation-dir=">,
     Group<f_Group>, Flags<[CC1Option, CC1AsOption, CoreOption]>,
     HelpText<"The compilation directory to embed in the debug info">,
     MarshallingInfoString<CodeGenOpts<"DebugCompilationDir">>;
+def fdebug_info_allowed_file_EQ : Joined<["-"], "fdebug-info-allowed-file=">,
+    Group<f_Group>, Flags<[CC1Option, CoreOption]>,
+    HelpText<"Emit rich debug info only for declarations from the given file "
+             "(repeatable)">,
+    MarshallingInfoStringVector<CodeGenOpts<"DebugInfoAllowedFiles">>;
 def fdebug_compilation_dir : Separate<["-"], "fdebug-compilation-dir">,
     Group<f_Group>, Flags<[CC1Option, CC1AsOption, CoreOption]>,
     Alias<fdebug_compilation_dir_EQ>;
@@ -4376,6 +4381,10 @@ def working_directory : Separate<["-"], "working-directory">, Flags<[CC1Option]>
   MarshallingInfoString<FileSystemOpts<"WorkingDir">>;
 def working_directory_EQ : Joined<["-"], "working-directory=">, Flags<[CC1Option]>,
   Alias<working_directory>;
+def fcase_insensitive_paths : Flag<["-"], "fcase-insensitive-paths">,
+  Flags<[CC1Option]>,
+  HelpText<"Retry failed file lookups with a case-insensitive directory scan">,
+  MarshallingInfoFlag<FileSystemOpts<"CaseInsensitivePaths">>;
 
 // Double dash options, which are usually an alias for one of the previous
 // options.

clang/lib/Basic/FileManager.cpp

@@ -577,6 +577,43 @@ FileManager::getBufferForFileImpl(StringRef Filename, int64_t FileSize,
                               isVolatile);
 }
 
+/// Walk a path component-by-component, resolving each level with a
+/// case-insensitive directory scan.  Returns the real-cased path on success.
+static std::optional<std::string>
+findPathCaseInsensitive(StringRef CaseInsensitivePath) {
+  using llvm::sys::fs::directory_iterator;
+  namespace path = llvm::sys::path;
+
+  SmallString<256> Resolved = path::root_path(CaseInsensitivePath);
+  StringRef Relative = path::relative_path(CaseInsensitivePath);
+  if (Relative.empty())
+    return std::nullopt;
+
+  for (StringRef Component :
+       llvm::make_range(path::begin(Relative), path::end(Relative))) {
+    std::error_code EC;
+    std::string Match;
+    unsigned Hits = 0;
+    for (directory_iterator It(Resolved, EC), End; It != End && !EC;
+         It.increment(EC)) {
+      StringRef EntryName = path::filename(It->path());
+      if (EntryName.equals_insensitive(Component)) {
+        if (Hits == 0)
+          Match = EntryName.str();
+        ++Hits;
+      }
+    }
+    if (EC || Hits == 0)
+      return std::nullopt;
+    if (Hits > 1)
+      llvm::errs() << "warning: ambiguous case-insensitive match for '"
+                   << Component << "' in '" << Resolved
+                   << "'; picking '" << Match << "'\n";
+    path::append(Resolved, Match);
+  }
+  return Resolved.str().str();
+}
+
 /// getStatValue - Get the 'stat' information for the specified path,
 /// using the cache to accelerate it if possible.  This returns true
 /// if the path points to a virtual file or does not exist, or returns
@@ -587,15 +624,21 @@ FileManager::getStatValue(StringRef Path, llvm::vfs::Status &Status,
                           bool isFile, std::unique_ptr<llvm::vfs::File> *F) {
   // FIXME: FileSystemOpts shouldn't be passed in here, all paths should be
   // absolute!
-  if (FileSystemOpts.WorkingDir.empty())
-    return FileSystemStatCache::get(Path, Status, isFile, F,
-                                    StatCache.get(), *FS);
-
   SmallString<128> FilePath(Path);
-  FixupRelativePath(FilePath);
-
-  return FileSystemStatCache::get(FilePath.c_str(), Status, isFile, F,
-                                  StatCache.get(), *FS);
+  if (!FileSystemOpts.WorkingDir.empty())
+    FixupRelativePath(FilePath);
+
+  auto stat = [&](StringRef P) {
+    return FileSystemStatCache::get(P, Status, isFile, F, StatCache.get(), *FS);
+  };
+
+  auto EC = stat(FilePath);
+  if (!EC)
+    return EC;
+  if (FileSystemOpts.CaseInsensitivePaths)
+    if (auto Resolved = findPathCaseInsensitive(FilePath))
+      return stat(*Resolved);
+  return EC;
 }
 
 std::error_code

clang/lib/CodeGen/CGDebugInfo.cpp

@@ -4189,6 +4189,7 @@ void CGDebugInfo::EmitFunctionDecl(GlobalDecl GD, SourceLocation Loc,
   llvm::DINode::DIFlags Flags = llvm::DINode::FlagZero;
   llvm::DIFile *Unit = getOrCreateFile(Loc);
   bool IsDeclForCallSite = Fn ? true : false;
+
   llvm::DIScope *FDContext =
       IsDeclForCallSite ? Unit : getDeclContextDescriptor(D);
   llvm::DINodeArray TParamsArray;
@@ -4219,10 +4220,17 @@ void CGDebugInfo::EmitFunctionDecl(GlobalDecl GD, SourceLocation Loc,
 
   llvm::DINodeArray Annotations = CollectBTFDeclTagAnnotations(D);
   llvm::DISubroutineType *STy = getOrCreateFunctionType(D, FnType, Unit);
+  // Only link to a prior declaration when creating a call-site entry.
+  // For standalone declarations the SP *is* the declaration; passing
+  // getFunctionDeclaration() here would let DISubprogram::get() unique
+  // the new node to the existing member SP, producing a self-referencing
+  // declaration link that causes infinite recursion in the DWARF emitter.
+  llvm::DISubprogram *Decl =
+      IsDeclForCallSite ? getFunctionDeclaration(D) : nullptr;
   llvm::DISubprogram *SP =
       DBuilder.createFunction(FDContext, Name, LinkageName, Unit, LineNo, STy,
                               ScopeLine, Flags, SPFlags, TParamsArray.get(),
-                              getFunctionDeclaration(D), nullptr, Annotations);
+                              Decl, nullptr, Annotations);
 
   // Preserve btf_decl_tag attributes for parameters of extern functions
   // for BPF target. The parameters created in this loop are attached as
@@ -4243,6 +4251,8 @@ void CGDebugInfo::EmitFunctionDecl(GlobalDecl GD, SourceLocation Loc,
 
   if (IsDeclForCallSite)
     Fn->setSubprogram(SP);
+  else if (CGM.getCodeGenOpts().hasMaybeUnusedDebugInfo())
+    DBuilder.retainType(SP);
 
   DBuilder.finalizeSubprogram(SP);
 }

clang/lib/CodeGen/CodeGenModule.cpp

@@ -6276,6 +6276,42 @@ void CodeGenModule::EmitDeclContext(const DeclContext *DC) {
   }
 }
 
+/// Returns true if D comes from a file selected for rich debug info.
+/// If -fdebug-info-allowed-file is not set, all declarations pass.
+static bool isDeclInAllowedFiles(const Decl *D, CodeGenModule &CGM) {
+  const auto &Allowed = CGM.getCodeGenOpts().DebugInfoAllowedFiles;
+  if (Allowed.empty())
+    return true;
+  auto &SM = CGM.getContext().getSourceManager();
+  auto Filename = SM.getFilename(SM.getExpansionLoc(D->getLocation()));
+  if (Filename.empty())
+    return false;
+  for (const auto &F : Allowed)
+    if (Filename == F)
+      return true;
+  return false;
+}
+
+/// Emit a DISubprogram for a function declaration so its prototype ends
+/// up in the debug output (LF_FUNC_ID in CodeView) even when the TU
+/// has no definition.
+static void emitFunctionDecl(CodeGenModule &CGM, const FunctionDecl *FD) {
+  // Only emit for declaration-only functions; definitions already get
+  // a DISubprogram through the normal codegen path.
+  if (FD->doesThisDeclarationHaveABody())
+    return;
+  // Skip declarations whose type can't be represented in debug info
+  // (e.g. undeduced auto return types, dependent template types).
+  const auto *FPT = FD->getType()->getAs<FunctionProtoType>();
+  if (!FPT || FPT->isDependentType() || FPT->isUndeducedType())
+    return;
+  if (!isDeclInAllowedFiles(FD, CGM))
+    return;
+  CGDebugInfo *DI = CGM.getModuleDebugInfo();
+  if (DI)
+    DI->EmitFunctionDecl(GlobalDecl(FD), FD->getLocation(), QualType(FPT, 0));
+}
+
 /// EmitTopLevelDecl - Emit code for a single top level declaration.
 void CodeGenModule::EmitTopLevelDecl(Decl *D) {
   // Ignore dependent declarations.
@@ -6295,6 +6331,7 @@ void CodeGenModule::EmitTopLevelDecl(Decl *D) {
     // Always provide some coverage mapping
     // even for the functions that aren't emitted.
     AddDeferredUnusedCoverageMapping(D);
+    emitFunctionDecl(*this, cast<FunctionDecl>(D));
     break;
 
   case Decl::CXXDeductionGuide:
@@ -6331,7 +6368,7 @@ void CodeGenModule::EmitTopLevelDecl(Decl *D) {
   case Decl::CXXRecord: {
     CXXRecordDecl *CRD = cast<CXXRecordDecl>(D);
     if (CGDebugInfo *DI = getModuleDebugInfo()) {
-      if (CRD->hasDefinition())
+      if (CRD->hasDefinition() && isDeclInAllowedFiles(D, *this))
         DI->EmitAndRetainType(getContext().getRecordType(cast<RecordDecl>(D)));
       if (auto *ES = D->getASTContext().getExternalSource())
         if (ES->hasExternalDefinitions(D) == ExternalASTSource::EK_Never)
@@ -6553,19 +6590,22 @@ void CodeGenModule::EmitTopLevelDecl(Decl *D) {
   case Decl::Typedef:
   case Decl::TypeAlias: // using foo = bar; [C++11]
     if (CGDebugInfo *DI = getModuleDebugInfo())
-      DI->EmitAndRetainType(
-          getContext().getTypedefType(cast<TypedefNameDecl>(D)));
+      if (isDeclInAllowedFiles(D, *this))
+        DI->EmitAndRetainType(
+            getContext().getTypedefType(cast<TypedefNameDecl>(D)));
     break;
 
   case Decl::Record:
     if (CGDebugInfo *DI = getModuleDebugInfo())
-      if (cast<RecordDecl>(D)->getDefinition())
+      if (cast<RecordDecl>(D)->getDefinition() &&
+          isDeclInAllowedFiles(D, *this))
         DI->EmitAndRetainType(getContext().getRecordType(cast<RecordDecl>(D)));
     break;
 
   case Decl::Enum:
     if (CGDebugInfo *DI = getModuleDebugInfo())
-      if (cast<EnumDecl>(D)->getDefinition())
+      if (cast<EnumDecl>(D)->getDefinition() &&
+          isDeclInAllowedFiles(D, *this))
         DI->EmitAndRetainType(getContext().getEnumType(cast<EnumDecl>(D)));
     break;
 

clang/lib/Driver/ToolChains/Clang.cpp

@@ -5743,6 +5743,8 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
   CmdArgs.push_back(D.ResourceDir.c_str());
 
   Args.AddLastArg(CmdArgs, options::OPT_working_directory);
+  Args.AddLastArg(CmdArgs, options::OPT_fcase_insensitive_paths);
+  Args.AddAllArgs(CmdArgs, options::OPT_fdebug_info_allowed_file_EQ);
 
   RenderARCMigrateToolOptions(D, Args, CmdArgs);
 

clang/test/CodeGen/debug-info-allowed-file.c

@@ -0,0 +1,26 @@
+// Verify that -fdebug-info-allowed-file restricts rich debug-info emission to
+// declarations from the listed files.
+
+// RUN: rm -rf %t && mkdir -p %t
+// RUN: echo 'struct allowed_struct {}; void allowed_fn(int);' > %t/allowed.h
+// RUN: echo 'struct excluded_struct {}; void excluded_fn(int);' > %t/excluded.h
+
+// With the filter, only decls from allowed.h appear in the debug output.
+// RUN: %clang_cc1 -debug-info-kind=unused-types -emit-llvm -o - %s -I %t \
+// RUN:   -fdebug-info-allowed-file=%t/allowed.h \
+// RUN:   | FileCheck %s --check-prefix=FILTERED
+
+// Without the filter, both decls are emitted.
+// RUN: %clang_cc1 -debug-info-kind=unused-types -emit-llvm -o - %s -I %t \
+// RUN:   | FileCheck %s --check-prefix=UNFILTERED
+
+#include "allowed.h"
+#include "excluded.h"
+
+// FILTERED-DAG: name: "allowed_struct"
+// FILTERED-DAG: name: "allowed_fn"
+// FILTERED-NOT: name: "excluded_struct"
+// FILTERED-NOT: name: "excluded_fn"
+
+// UNFILTERED-DAG: name: "allowed_struct"
+// UNFILTERED-DAG: name: "excluded_struct"

clang/test/CodeGen/debug-info-codeview-emit-decl-subprogram.c

@@ -0,0 +1,16 @@
+// Verify that a declaration-only function gets its DISubprogram retained
+// and lowered to an LF_FUNC_ID record in the CodeView output.
+
+// REQUIRES: x86-registered-target
+
+// RUN: %clang_cc1 -triple x86_64-windows-msvc -gcodeview \
+// RUN:   -debug-info-kind=unused-types -emit-obj -o %t.o %s
+// RUN: llvm-readobj %t.o --codeview | FileCheck %s
+
+void declared_fn(int);
+
+// CHECK:      FuncId ({{.*}}) {
+// CHECK-NEXT:   TypeLeafKind: LF_FUNC_ID
+// CHECK:        FunctionType: void (int)
+// CHECK-NEXT:   Name: declared_fn
+// CHECK-NEXT: }

clang/test/Driver/fcase-insensitive-paths.c

@@ -0,0 +1,3 @@
+// RUN: %clang -### -fcase-insensitive-paths -c %s 2>&1 | FileCheck %s
+
+// CHECK: "-fcase-insensitive-paths"

clang/test/Driver/fdebug-info-allowed-file.c

@@ -0,0 +1,3 @@
+// RUN: %clang -### -fdebug-info-allowed-file=a.h -fdebug-info-allowed-file=b.h -c %s 2>&1 | FileCheck %s
+
+// CHECK: "-fdebug-info-allowed-file=a.h" "-fdebug-info-allowed-file=b.h"