1 //===-- SymbolFilePDB.cpp ---------------------------------------*- C++ -*-===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 
10 #include "SymbolFilePDB.h"
11 
12 #include "clang/Lex/Lexer.h"
13 
14 #include "lldb/Core/Module.h"
15 #include "lldb/Core/PluginManager.h"
16 #include "lldb/Symbol/ClangASTContext.h"
17 #include "lldb/Symbol/CompileUnit.h"
18 #include "lldb/Symbol/LineTable.h"
19 #include "lldb/Symbol/ObjectFile.h"
20 #include "lldb/Symbol/SymbolContext.h"
21 #include "lldb/Symbol/SymbolVendor.h"
22 #include "lldb/Symbol/TypeList.h"
23 #include "lldb/Symbol/TypeMap.h"
24 #include "lldb/Symbol/Variable.h"
25 #include "lldb/Utility/RegularExpression.h"
26 
27 #include "llvm/DebugInfo/PDB/GenericError.h"
28 #include "llvm/DebugInfo/PDB/IPDBDataStream.h"
29 #include "llvm/DebugInfo/PDB/IPDBEnumChildren.h"
30 #include "llvm/DebugInfo/PDB/IPDBLineNumber.h"
31 #include "llvm/DebugInfo/PDB/IPDBSectionContrib.h"
32 #include "llvm/DebugInfo/PDB/IPDBSourceFile.h"
33 #include "llvm/DebugInfo/PDB/IPDBTable.h"
34 #include "llvm/DebugInfo/PDB/PDBSymbol.h"
35 #include "llvm/DebugInfo/PDB/PDBSymbolBlock.h"
36 #include "llvm/DebugInfo/PDB/PDBSymbolCompiland.h"
37 #include "llvm/DebugInfo/PDB/PDBSymbolCompilandDetails.h"
38 #include "llvm/DebugInfo/PDB/PDBSymbolData.h"
39 #include "llvm/DebugInfo/PDB/PDBSymbolExe.h"
40 #include "llvm/DebugInfo/PDB/PDBSymbolFunc.h"
41 #include "llvm/DebugInfo/PDB/PDBSymbolFuncDebugEnd.h"
42 #include "llvm/DebugInfo/PDB/PDBSymbolFuncDebugStart.h"
43 #include "llvm/DebugInfo/PDB/PDBSymbolPublicSymbol.h"
44 #include "llvm/DebugInfo/PDB/PDBSymbolTypeEnum.h"
45 #include "llvm/DebugInfo/PDB/PDBSymbolTypeTypedef.h"
46 #include "llvm/DebugInfo/PDB/PDBSymbolTypeUDT.h"
47 
48 #include "Plugins/Language/CPlusPlus/CPlusPlusLanguage.h"
49 #include "Plugins/SymbolFile/PDB/PDBASTParser.h"
50 
51 #include <regex>
52 
53 using namespace lldb;
54 using namespace lldb_private;
55 using namespace llvm::pdb;
56 
57 namespace {
58 lldb::LanguageType TranslateLanguage(PDB_Lang lang) {
59   switch (lang) {
60   case PDB_Lang::Cpp:
61     return lldb::LanguageType::eLanguageTypeC_plus_plus;
62   case PDB_Lang::C:
63     return lldb::LanguageType::eLanguageTypeC;
64   default:
65     return lldb::LanguageType::eLanguageTypeUnknown;
66   }
67 }
68 
69 bool ShouldAddLine(uint32_t requested_line, uint32_t actual_line,
70                    uint32_t addr_length) {
71   return ((requested_line == 0 || actual_line == requested_line) &&
72           addr_length > 0);
73 }
74 } // namespace
75 
76 void SymbolFilePDB::Initialize() {
77   PluginManager::RegisterPlugin(GetPluginNameStatic(),
78                                 GetPluginDescriptionStatic(), CreateInstance,
79                                 DebuggerInitialize);
80 }
81 
82 void SymbolFilePDB::Terminate() {
83   PluginManager::UnregisterPlugin(CreateInstance);
84 }
85 
86 void SymbolFilePDB::DebuggerInitialize(lldb_private::Debugger &debugger) {}
87 
88 lldb_private::ConstString SymbolFilePDB::GetPluginNameStatic() {
89   static ConstString g_name("pdb");
90   return g_name;
91 }
92 
93 const char *SymbolFilePDB::GetPluginDescriptionStatic() {
94   return "Microsoft PDB debug symbol file reader.";
95 }
96 
97 lldb_private::SymbolFile *
98 SymbolFilePDB::CreateInstance(lldb_private::ObjectFile *obj_file) {
99   return new SymbolFilePDB(obj_file);
100 }
101 
102 SymbolFilePDB::SymbolFilePDB(lldb_private::ObjectFile *object_file)
103     : SymbolFile(object_file), m_session_up(), m_global_scope_up(),
104       m_cached_compile_unit_count(0), m_tu_decl_ctx_up() {}
105 
106 SymbolFilePDB::~SymbolFilePDB() {}
107 
108 uint32_t SymbolFilePDB::CalculateAbilities() {
109   uint32_t abilities = 0;
110   if (!m_obj_file)
111     return 0;
112 
113   if (!m_session_up) {
114     // Lazily load and match the PDB file, but only do this once.
115     std::string exePath = m_obj_file->GetFileSpec().GetPath();
116     auto error = loadDataForEXE(PDB_ReaderType::DIA, llvm::StringRef(exePath),
117                                 m_session_up);
118     if (error) {
119       llvm::consumeError(std::move(error));
120       auto module_sp = m_obj_file->GetModule();
121       if (!module_sp)
122         return 0;
123       // See if any symbol file is specified through `--symfile` option.
124       FileSpec symfile = module_sp->GetSymbolFileFileSpec();
125       if (!symfile)
126         return 0;
127       error = loadDataForPDB(PDB_ReaderType::DIA,
128                              llvm::StringRef(symfile.GetPath()), m_session_up);
129       if (error) {
130         llvm::consumeError(std::move(error));
131         return 0;
132       }
133     }
134   }
135   if (!m_session_up)
136     return 0;
137 
138   auto enum_tables_up = m_session_up->getEnumTables();
139   if (!enum_tables_up)
140     return 0;
141   while (auto table_up = enum_tables_up->getNext()) {
142     if (table_up->getItemCount() == 0)
143       continue;
144     auto type = table_up->getTableType();
145     switch (type) {
146     case PDB_TableType::Symbols:
147       // This table represents a store of symbols with types listed in
148       // PDBSym_Type
149       abilities |= (CompileUnits | Functions | Blocks | GlobalVariables |
150                     LocalVariables | VariableTypes);
151       break;
152     case PDB_TableType::LineNumbers:
153       abilities |= LineTables;
154       break;
155     default:
156       break;
157     }
158   }
159   return abilities;
160 }
161 
162 void SymbolFilePDB::InitializeObject() {
163   lldb::addr_t obj_load_address = m_obj_file->GetFileOffset();
164   lldbassert(obj_load_address && obj_load_address != LLDB_INVALID_ADDRESS);
165   m_session_up->setLoadAddress(obj_load_address);
166   if (!m_global_scope_up)
167     m_global_scope_up = m_session_up->getGlobalScope();
168   lldbassert(m_global_scope_up.get());
169 
170   TypeSystem *type_system =
171       GetTypeSystemForLanguage(lldb::eLanguageTypeC_plus_plus);
172   ClangASTContext *clang_type_system =
173       llvm::dyn_cast_or_null<ClangASTContext>(type_system);
174   lldbassert(clang_type_system);
175   m_tu_decl_ctx_up = llvm::make_unique<CompilerDeclContext>(
176       type_system, clang_type_system->GetTranslationUnitDecl());
177 }
178 
179 uint32_t SymbolFilePDB::GetNumCompileUnits() {
180   if (m_cached_compile_unit_count == 0) {
181     auto compilands = m_global_scope_up->findAllChildren<PDBSymbolCompiland>();
182     if (!compilands)
183       return 0;
184 
185     // The linker could link *.dll (compiland language = LINK), or import
186     // *.dll. For example, a compiland with name `Import:KERNEL32.dll` could be
187     // found as a child of the global scope (PDB executable). Usually, such
188     // compilands contain `thunk` symbols in which we are not interested for
189     // now. However we still count them in the compiland list. If we perform
190     // any compiland related activity, like finding symbols through
191     // llvm::pdb::IPDBSession methods, such compilands will all be searched
192     // automatically no matter whether we include them or not.
193     m_cached_compile_unit_count = compilands->getChildCount();
194 
195     // The linker can inject an additional "dummy" compilation unit into the
196     // PDB. Ignore this special compile unit for our purposes, if it is there.
197     // It is always the last one.
198     auto last_compiland_up =
199         compilands->getChildAtIndex(m_cached_compile_unit_count - 1);
200     lldbassert(last_compiland_up.get());
201     std::string name = last_compiland_up->getName();
202     if (name == "* Linker *")
203       --m_cached_compile_unit_count;
204   }
205   return m_cached_compile_unit_count;
206 }
207 
208 void SymbolFilePDB::GetCompileUnitIndex(
209     const llvm::pdb::PDBSymbolCompiland &pdb_compiland, uint32_t &index) {
210   auto results_up = m_global_scope_up->findAllChildren<PDBSymbolCompiland>();
211   if (!results_up)
212     return;
213   auto uid = pdb_compiland.getSymIndexId();
214   for (uint32_t cu_idx = 0; cu_idx < GetNumCompileUnits(); ++cu_idx) {
215     auto compiland_up = results_up->getChildAtIndex(cu_idx);
216     if (!compiland_up)
217       continue;
218     if (compiland_up->getSymIndexId() == uid) {
219       index = cu_idx;
220       return;
221     }
222   }
223   index = UINT32_MAX;
224   return;
225 }
226 
227 std::unique_ptr<llvm::pdb::PDBSymbolCompiland>
228 SymbolFilePDB::GetPDBCompilandByUID(uint32_t uid) {
229   return m_session_up->getConcreteSymbolById<PDBSymbolCompiland>(uid);
230 }
231 
232 lldb::CompUnitSP SymbolFilePDB::ParseCompileUnitAtIndex(uint32_t index) {
233   if (index >= GetNumCompileUnits())
234     return CompUnitSP();
235 
236   // Assuming we always retrieve same compilands listed in same order through
237   // `PDBSymbolExe::findAllChildren` method, otherwise using `index` to get a
238   // compile unit makes no sense.
239   auto results = m_global_scope_up->findAllChildren<PDBSymbolCompiland>();
240   if (!results)
241     return CompUnitSP();
242   auto compiland_up = results->getChildAtIndex(index);
243   if (!compiland_up)
244     return CompUnitSP();
245   return ParseCompileUnitForUID(compiland_up->getSymIndexId(), index);
246 }
247 
248 lldb::LanguageType
249 SymbolFilePDB::ParseCompileUnitLanguage(const lldb_private::SymbolContext &sc) {
250   // What fields should I expect to be filled out on the SymbolContext?  Is it
251   // safe to assume that `sc.comp_unit` is valid?
252   if (!sc.comp_unit)
253     return lldb::eLanguageTypeUnknown;
254 
255   auto compiland_up = GetPDBCompilandByUID(sc.comp_unit->GetID());
256   if (!compiland_up)
257     return lldb::eLanguageTypeUnknown;
258   auto details = compiland_up->findOneChild<PDBSymbolCompilandDetails>();
259   if (!details)
260     return lldb::eLanguageTypeUnknown;
261   return TranslateLanguage(details->getLanguage());
262 }
263 
264 lldb_private::Function *SymbolFilePDB::ParseCompileUnitFunctionForPDBFunc(
265     const PDBSymbolFunc &pdb_func, const lldb_private::SymbolContext &sc) {
266   lldbassert(sc.comp_unit && sc.module_sp.get());
267 
268   auto file_vm_addr = pdb_func.getVirtualAddress();
269   if (file_vm_addr == LLDB_INVALID_ADDRESS || file_vm_addr == 0)
270     return nullptr;
271 
272   auto func_length = pdb_func.getLength();
273   AddressRange func_range =
274       AddressRange(file_vm_addr, func_length, sc.module_sp->GetSectionList());
275   if (!func_range.GetBaseAddress().IsValid())
276     return nullptr;
277 
278   lldb_private::Type *func_type = ResolveTypeUID(pdb_func.getSymIndexId());
279   if (!func_type)
280     return nullptr;
281 
282   user_id_t func_type_uid = pdb_func.getSignatureId();
283 
284   Mangled mangled = GetMangledForPDBFunc(pdb_func);
285 
286   FunctionSP func_sp =
287       std::make_shared<Function>(sc.comp_unit, pdb_func.getSymIndexId(),
288                                  func_type_uid, mangled, func_type, func_range);
289 
290   sc.comp_unit->AddFunction(func_sp);
291   return func_sp.get();
292 }
293 
294 size_t SymbolFilePDB::ParseCompileUnitFunctions(
295     const lldb_private::SymbolContext &sc) {
296   lldbassert(sc.comp_unit);
297   size_t func_added = 0;
298   auto compiland_up = GetPDBCompilandByUID(sc.comp_unit->GetID());
299   if (!compiland_up)
300     return 0;
301   auto results_up = compiland_up->findAllChildren<PDBSymbolFunc>();
302   if (!results_up)
303     return 0;
304   while (auto pdb_func_up = results_up->getNext()) {
305     auto func_sp =
306         sc.comp_unit->FindFunctionByUID(pdb_func_up->getSymIndexId());
307     if (!func_sp) {
308       if (ParseCompileUnitFunctionForPDBFunc(*pdb_func_up, sc))
309         ++func_added;
310     }
311   }
312   return func_added;
313 }
314 
315 bool SymbolFilePDB::ParseCompileUnitLineTable(
316     const lldb_private::SymbolContext &sc) {
317   lldbassert(sc.comp_unit);
318   if (sc.comp_unit->GetLineTable())
319     return true;
320   return ParseCompileUnitLineTable(sc, 0);
321 }
322 
323 bool SymbolFilePDB::ParseCompileUnitDebugMacros(
324     const lldb_private::SymbolContext &sc) {
325   // PDB doesn't contain information about macros
326   return false;
327 }
328 
329 bool SymbolFilePDB::ParseCompileUnitSupportFiles(
330     const lldb_private::SymbolContext &sc,
331     lldb_private::FileSpecList &support_files) {
332   lldbassert(sc.comp_unit);
333 
334   // In theory this is unnecessary work for us, because all of this information
335   // is easily (and quickly) accessible from DebugInfoPDB, so caching it a
336   // second time seems like a waste.  Unfortunately, there's no good way around
337   // this short of a moderate refactor since SymbolVendor depends on being able
338   // to cache this list.
339   auto compiland_up = GetPDBCompilandByUID(sc.comp_unit->GetID());
340   if (!compiland_up)
341     return false;
342   auto files = m_session_up->getSourceFilesForCompiland(*compiland_up);
343   if (!files || files->getChildCount() == 0)
344     return false;
345 
346   while (auto file = files->getNext()) {
347     FileSpec spec(file->getFileName(), false, FileSpec::Style::windows);
348     support_files.AppendIfUnique(spec);
349   }
350 
351   // LLDB uses the DWARF-like file numeration (one based),
352   // the zeroth file is the compile unit itself
353   support_files.Insert(0, *sc.comp_unit);
354 
355   return true;
356 }
357 
358 bool SymbolFilePDB::ParseImportedModules(
359     const lldb_private::SymbolContext &sc,
360     std::vector<lldb_private::ConstString> &imported_modules) {
361   // PDB does not yet support module debug info
362   return false;
363 }
364 
365 static size_t ParseFunctionBlocksForPDBSymbol(
366     const lldb_private::SymbolContext &sc, uint64_t func_file_vm_addr,
367     const llvm::pdb::PDBSymbol *pdb_symbol, lldb_private::Block *parent_block,
368     bool is_top_parent) {
369   assert(pdb_symbol && parent_block);
370 
371   size_t num_added = 0;
372   switch (pdb_symbol->getSymTag()) {
373   case PDB_SymType::Block:
374   case PDB_SymType::Function: {
375     Block *block = nullptr;
376     auto &raw_sym = pdb_symbol->getRawSymbol();
377     if (auto *pdb_func = llvm::dyn_cast<PDBSymbolFunc>(pdb_symbol)) {
378       if (pdb_func->hasNoInlineAttribute())
379         break;
380       if (is_top_parent)
381         block = parent_block;
382       else
383         break;
384     } else if (llvm::dyn_cast<PDBSymbolBlock>(pdb_symbol)) {
385       auto uid = pdb_symbol->getSymIndexId();
386       if (parent_block->FindBlockByID(uid))
387         break;
388       if (raw_sym.getVirtualAddress() < func_file_vm_addr)
389         break;
390 
391       auto block_sp = std::make_shared<Block>(pdb_symbol->getSymIndexId());
392       parent_block->AddChild(block_sp);
393       block = block_sp.get();
394     } else
395       llvm_unreachable("Unexpected PDB symbol!");
396 
397     block->AddRange(Block::Range(
398         raw_sym.getVirtualAddress() - func_file_vm_addr, raw_sym.getLength()));
399     block->FinalizeRanges();
400     ++num_added;
401 
402     auto results_up = pdb_symbol->findAllChildren();
403     if (!results_up)
404       break;
405     while (auto symbol_up = results_up->getNext()) {
406       num_added += ParseFunctionBlocksForPDBSymbol(
407           sc, func_file_vm_addr, symbol_up.get(), block, false);
408     }
409   } break;
410   default:
411     break;
412   }
413   return num_added;
414 }
415 
416 size_t
417 SymbolFilePDB::ParseFunctionBlocks(const lldb_private::SymbolContext &sc) {
418   lldbassert(sc.comp_unit && sc.function);
419   size_t num_added = 0;
420   auto uid = sc.function->GetID();
421   auto pdb_func_up = m_session_up->getConcreteSymbolById<PDBSymbolFunc>(uid);
422   if (!pdb_func_up)
423     return 0;
424   Block &parent_block = sc.function->GetBlock(false);
425   num_added =
426       ParseFunctionBlocksForPDBSymbol(sc, pdb_func_up->getVirtualAddress(),
427                                       pdb_func_up.get(), &parent_block, true);
428   return num_added;
429 }
430 
431 size_t SymbolFilePDB::ParseTypes(const lldb_private::SymbolContext &sc) {
432   lldbassert(sc.module_sp.get());
433   if (!sc.comp_unit)
434     return 0;
435 
436   size_t num_added = 0;
437   auto compiland = GetPDBCompilandByUID(sc.comp_unit->GetID());
438   if (!compiland)
439     return 0;
440 
441   auto ParseTypesByTagFn = [&num_added, this](const PDBSymbol &raw_sym) {
442     std::unique_ptr<IPDBEnumSymbols> results;
443     PDB_SymType tags_to_search[] = {PDB_SymType::Enum, PDB_SymType::Typedef,
444                                     PDB_SymType::UDT};
445     for (auto tag : tags_to_search) {
446       results = raw_sym.findAllChildren(tag);
447       if (!results || results->getChildCount() == 0)
448         continue;
449       while (auto symbol = results->getNext()) {
450         switch (symbol->getSymTag()) {
451         case PDB_SymType::Enum:
452         case PDB_SymType::UDT:
453         case PDB_SymType::Typedef:
454           break;
455         default:
456           continue;
457         }
458 
459         // This should cause the type to get cached and stored in the `m_types`
460         // lookup.
461         if (!ResolveTypeUID(symbol->getSymIndexId()))
462           continue;
463 
464         ++num_added;
465       }
466     }
467   };
468 
469   if (sc.function) {
470     auto pdb_func = m_session_up->getConcreteSymbolById<PDBSymbolFunc>(
471         sc.function->GetID());
472     if (!pdb_func)
473       return 0;
474     ParseTypesByTagFn(*pdb_func);
475   } else {
476     ParseTypesByTagFn(*compiland);
477 
478     // Also parse global types particularly coming from this compiland.
479     // Unfortunately, PDB has no compiland information for each global type. We
480     // have to parse them all. But ensure we only do this once.
481     static bool parse_all_global_types = false;
482     if (!parse_all_global_types) {
483       ParseTypesByTagFn(*m_global_scope_up);
484       parse_all_global_types = true;
485     }
486   }
487   return num_added;
488 }
489 
490 size_t
491 SymbolFilePDB::ParseVariablesForContext(const lldb_private::SymbolContext &sc) {
492   if (!sc.comp_unit)
493     return 0;
494 
495   size_t num_added = 0;
496   if (sc.function) {
497     auto pdb_func = m_session_up->getConcreteSymbolById<PDBSymbolFunc>(
498         sc.function->GetID());
499     if (!pdb_func)
500       return 0;
501 
502     num_added += ParseVariables(sc, *pdb_func);
503     sc.function->GetBlock(false).SetDidParseVariables(true, true);
504   } else if (sc.comp_unit) {
505     auto compiland = GetPDBCompilandByUID(sc.comp_unit->GetID());
506     if (!compiland)
507       return 0;
508 
509     if (sc.comp_unit->GetVariableList(false))
510       return 0;
511 
512     auto results = m_global_scope_up->findAllChildren<PDBSymbolData>();
513     if (results && results->getChildCount()) {
514       while (auto result = results->getNext()) {
515         auto cu_id = result->getCompilandId();
516         // FIXME: We are not able to determine variable's compile unit.
517         if (cu_id == 0)
518           continue;
519 
520         if (cu_id == sc.comp_unit->GetID())
521           num_added += ParseVariables(sc, *result);
522       }
523     }
524 
525     // FIXME: A `file static` or `global constant` variable appears both in
526     // compiland's children and global scope's children with unexpectedly
527     // different symbol's Id making it ambiguous.
528 
529     // FIXME: 'local constant', for example, const char var[] = "abc", declared
530     // in a function scope, can't be found in PDB.
531 
532     // Parse variables in this compiland.
533     num_added += ParseVariables(sc, *compiland);
534   }
535 
536   return num_added;
537 }
538 
539 lldb_private::Type *SymbolFilePDB::ResolveTypeUID(lldb::user_id_t type_uid) {
540   auto find_result = m_types.find(type_uid);
541   if (find_result != m_types.end())
542     return find_result->second.get();
543 
544   TypeSystem *type_system =
545       GetTypeSystemForLanguage(lldb::eLanguageTypeC_plus_plus);
546   ClangASTContext *clang_type_system =
547       llvm::dyn_cast_or_null<ClangASTContext>(type_system);
548   if (!clang_type_system)
549     return nullptr;
550   PDBASTParser *pdb =
551       llvm::dyn_cast<PDBASTParser>(clang_type_system->GetPDBParser());
552   if (!pdb)
553     return nullptr;
554 
555   auto pdb_type = m_session_up->getSymbolById(type_uid);
556   if (pdb_type == nullptr)
557     return nullptr;
558 
559   lldb::TypeSP result = pdb->CreateLLDBTypeFromPDBType(*pdb_type);
560   if (result) {
561     m_types.insert(std::make_pair(type_uid, result));
562     auto type_list = GetTypeList();
563     if (type_list)
564       type_list->Insert(result);
565   }
566   return result.get();
567 }
568 
569 bool SymbolFilePDB::CompleteType(lldb_private::CompilerType &compiler_type) {
570   // TODO: Implement this
571   return false;
572 }
573 
574 lldb_private::CompilerDecl SymbolFilePDB::GetDeclForUID(lldb::user_id_t uid) {
575   return lldb_private::CompilerDecl();
576 }
577 
578 lldb_private::CompilerDeclContext
579 SymbolFilePDB::GetDeclContextForUID(lldb::user_id_t uid) {
580   // PDB always uses the translation unit decl context for everything.  We can
581   // improve this later but it's not easy because PDB doesn't provide a high
582   // enough level of type fidelity in this area.
583   return *m_tu_decl_ctx_up;
584 }
585 
586 lldb_private::CompilerDeclContext
587 SymbolFilePDB::GetDeclContextContainingUID(lldb::user_id_t uid) {
588   return *m_tu_decl_ctx_up;
589 }
590 
591 void SymbolFilePDB::ParseDeclsForContext(
592     lldb_private::CompilerDeclContext decl_ctx) {}
593 
594 uint32_t
595 SymbolFilePDB::ResolveSymbolContext(const lldb_private::Address &so_addr,
596                                     uint32_t resolve_scope,
597                                     lldb_private::SymbolContext &sc) {
598   uint32_t resolved_flags = 0;
599   if (resolve_scope & eSymbolContextCompUnit ||
600       resolve_scope & eSymbolContextVariable ||
601       resolve_scope & eSymbolContextFunction ||
602       resolve_scope & eSymbolContextBlock ||
603       resolve_scope & eSymbolContextLineEntry) {
604     auto cu_sp = GetCompileUnitContainsAddress(so_addr);
605     if (!cu_sp) {
606       if (resolved_flags | eSymbolContextVariable) {
607         // TODO: Resolve variables
608       }
609       return 0;
610     }
611     sc.comp_unit = cu_sp.get();
612     resolved_flags |= eSymbolContextCompUnit;
613     lldbassert(sc.module_sp == cu_sp->GetModule());
614   }
615 
616   if (resolve_scope & eSymbolContextFunction) {
617     addr_t file_vm_addr = so_addr.GetFileAddress();
618     auto symbol_up =
619         m_session_up->findSymbolByAddress(file_vm_addr, PDB_SymType::Function);
620     if (symbol_up) {
621       auto *pdb_func = llvm::dyn_cast<PDBSymbolFunc>(symbol_up.get());
622       assert(pdb_func);
623       auto func_uid = pdb_func->getSymIndexId();
624       sc.function = sc.comp_unit->FindFunctionByUID(func_uid).get();
625       if (sc.function == nullptr)
626         sc.function = ParseCompileUnitFunctionForPDBFunc(*pdb_func, sc);
627       if (sc.function) {
628         resolved_flags |= eSymbolContextFunction;
629         if (resolve_scope & eSymbolContextBlock) {
630           Block &block = sc.function->GetBlock(true);
631           sc.block = block.FindBlockByID(sc.function->GetID());
632           if (sc.block)
633             resolved_flags |= eSymbolContextBlock;
634         }
635       }
636     }
637   }
638 
639   if (resolve_scope & eSymbolContextLineEntry) {
640     if (auto *line_table = sc.comp_unit->GetLineTable()) {
641       Address addr(so_addr);
642       if (line_table->FindLineEntryByAddress(addr, sc.line_entry))
643         resolved_flags |= eSymbolContextLineEntry;
644     }
645   }
646 
647   return resolved_flags;
648 }
649 
650 uint32_t SymbolFilePDB::ResolveSymbolContext(
651     const lldb_private::FileSpec &file_spec, uint32_t line, bool check_inlines,
652     uint32_t resolve_scope, lldb_private::SymbolContextList &sc_list) {
653   const size_t old_size = sc_list.GetSize();
654   if (resolve_scope & lldb::eSymbolContextCompUnit) {
655     // Locate all compilation units with line numbers referencing the specified
656     // file.  For example, if `file_spec` is <vector>, then this should return
657     // all source files and header files that reference <vector>, either
658     // directly or indirectly.
659     auto compilands = m_session_up->findCompilandsForSourceFile(
660         file_spec.GetPath(), PDB_NameSearchFlags::NS_CaseInsensitive);
661 
662     if (!compilands)
663       return 0;
664 
665     // For each one, either find its previously parsed data or parse it afresh
666     // and add it to the symbol context list.
667     while (auto compiland = compilands->getNext()) {
668       // If we're not checking inlines, then don't add line information for
669       // this file unless the FileSpec matches. For inline functions, we don't
670       // have to match the FileSpec since they could be defined in headers
671       // other than file specified in FileSpec.
672       if (!check_inlines) {
673         std::string source_file = compiland->getSourceFileFullPath();
674         if (source_file.empty())
675           continue;
676         FileSpec this_spec(source_file, false, FileSpec::Style::windows);
677         bool need_full_match = !file_spec.GetDirectory().IsEmpty();
678         if (FileSpec::Compare(file_spec, this_spec, need_full_match) != 0)
679           continue;
680       }
681 
682       SymbolContext sc;
683       auto cu = ParseCompileUnitForUID(compiland->getSymIndexId());
684       if (!cu)
685         continue;
686       sc.comp_unit = cu.get();
687       sc.module_sp = cu->GetModule();
688 
689       // If we were asked to resolve line entries, add all entries to the line
690       // table that match the requested line (or all lines if `line` == 0).
691       if (resolve_scope & (eSymbolContextFunction | eSymbolContextBlock |
692                            eSymbolContextLineEntry)) {
693         bool has_line_table = ParseCompileUnitLineTable(sc, line);
694 
695         if ((resolve_scope & eSymbolContextLineEntry) && !has_line_table) {
696           // The query asks for line entries, but we can't get them for the
697           // compile unit. This is not normal for `line` = 0. So just assert
698           // it.
699           assert(line && "Couldn't get all line entries!\n");
700 
701           // Current compiland does not have the requested line. Search next.
702           continue;
703         }
704 
705         if (resolve_scope & (eSymbolContextFunction | eSymbolContextBlock)) {
706           if (!has_line_table)
707             continue;
708 
709           auto *line_table = sc.comp_unit->GetLineTable();
710           lldbassert(line_table);
711 
712           uint32_t num_line_entries = line_table->GetSize();
713           // Skip the terminal line entry.
714           --num_line_entries;
715 
716           // If `line `!= 0, see if we can resolve function for each line entry
717           // in the line table.
718           for (uint32_t line_idx = 0; line && line_idx < num_line_entries;
719                ++line_idx) {
720             if (!line_table->GetLineEntryAtIndex(line_idx, sc.line_entry))
721               continue;
722 
723             auto file_vm_addr =
724                 sc.line_entry.range.GetBaseAddress().GetFileAddress();
725             if (file_vm_addr == LLDB_INVALID_ADDRESS || file_vm_addr == 0)
726               continue;
727 
728             auto symbol_up = m_session_up->findSymbolByAddress(
729                 file_vm_addr, PDB_SymType::Function);
730             if (symbol_up) {
731               auto func_uid = symbol_up->getSymIndexId();
732               sc.function = sc.comp_unit->FindFunctionByUID(func_uid).get();
733               if (sc.function == nullptr) {
734                 auto pdb_func = llvm::dyn_cast<PDBSymbolFunc>(symbol_up.get());
735                 assert(pdb_func);
736                 sc.function = ParseCompileUnitFunctionForPDBFunc(*pdb_func, sc);
737               }
738               if (sc.function && (resolve_scope & eSymbolContextBlock)) {
739                 Block &block = sc.function->GetBlock(true);
740                 sc.block = block.FindBlockByID(sc.function->GetID());
741               }
742             }
743             sc_list.Append(sc);
744           }
745         } else if (has_line_table) {
746           // We can parse line table for the compile unit. But no query to
747           // resolve function or block. We append `sc` to the list anyway.
748           sc_list.Append(sc);
749         }
750       } else {
751         // No query for line entry, function or block. But we have a valid
752         // compile unit, append `sc` to the list.
753         sc_list.Append(sc);
754       }
755     }
756   }
757   return sc_list.GetSize() - old_size;
758 }
759 
760 std::string SymbolFilePDB::GetMangledForPDBData(const PDBSymbolData &pdb_data) {
761   std::string decorated_name;
762   auto vm_addr = pdb_data.getVirtualAddress();
763   if (vm_addr != LLDB_INVALID_ADDRESS && vm_addr) {
764     auto result_up =
765         m_global_scope_up->findAllChildren(PDB_SymType::PublicSymbol);
766     if (result_up) {
767       while (auto symbol_up = result_up->getNext()) {
768         if (symbol_up->getRawSymbol().getVirtualAddress() == vm_addr) {
769           decorated_name = symbol_up->getRawSymbol().getName();
770           break;
771         }
772       }
773     }
774   }
775   if (!decorated_name.empty())
776     return decorated_name;
777 
778   return std::string();
779 }
780 
781 VariableSP SymbolFilePDB::ParseVariableForPDBData(
782     const lldb_private::SymbolContext &sc,
783     const llvm::pdb::PDBSymbolData &pdb_data) {
784   VariableSP var_sp;
785   uint32_t var_uid = pdb_data.getSymIndexId();
786   auto result = m_variables.find(var_uid);
787   if (result != m_variables.end())
788     return result->second;
789 
790   ValueType scope = eValueTypeInvalid;
791   bool is_static_member = false;
792   bool is_external = false;
793   bool is_artificial = false;
794 
795   switch (pdb_data.getDataKind()) {
796   case PDB_DataKind::Global:
797     scope = eValueTypeVariableGlobal;
798     is_external = true;
799     break;
800   case PDB_DataKind::Local:
801     scope = eValueTypeVariableLocal;
802     break;
803   case PDB_DataKind::FileStatic:
804     scope = eValueTypeVariableStatic;
805     break;
806   case PDB_DataKind::StaticMember:
807     is_static_member = true;
808     scope = eValueTypeVariableStatic;
809     break;
810   case PDB_DataKind::Member:
811     scope = eValueTypeVariableStatic;
812     break;
813   case PDB_DataKind::Param:
814     scope = eValueTypeVariableArgument;
815     break;
816   case PDB_DataKind::Constant:
817     scope = eValueTypeConstResult;
818     break;
819   default:
820     break;
821   }
822 
823   switch (pdb_data.getLocationType()) {
824   case PDB_LocType::TLS:
825     scope = eValueTypeVariableThreadLocal;
826     break;
827   case PDB_LocType::RegRel: {
828     // It is a `this` pointer.
829     if (pdb_data.getDataKind() == PDB_DataKind::ObjectPtr) {
830       scope = eValueTypeVariableArgument;
831       is_artificial = true;
832     }
833   } break;
834   default:
835     break;
836   }
837 
838   Declaration decl;
839   if (!is_artificial && !pdb_data.isCompilerGenerated()) {
840     if (auto lines = pdb_data.getLineNumbers()) {
841       if (auto first_line = lines->getNext()) {
842         uint32_t src_file_id = first_line->getSourceFileId();
843         auto src_file = m_session_up->getSourceFileById(src_file_id);
844         if (src_file) {
845           FileSpec spec(src_file->getFileName(), /*resolve_path*/ false);
846           decl.SetFile(spec);
847           decl.SetColumn(first_line->getColumnNumber());
848           decl.SetLine(first_line->getLineNumber());
849         }
850       }
851     }
852   }
853 
854   Variable::RangeList ranges;
855   SymbolContextScope *context_scope = sc.comp_unit;
856   if (scope == eValueTypeVariableLocal) {
857     if (sc.function) {
858       context_scope = sc.function->GetBlock(true).FindBlockByID(
859           pdb_data.getClassParentId());
860       if (context_scope == nullptr)
861         context_scope = sc.function;
862     }
863   }
864 
865   SymbolFileTypeSP type_sp =
866       std::make_shared<SymbolFileType>(*this, pdb_data.getTypeId());
867 
868   auto var_name = pdb_data.getName();
869   auto mangled = GetMangledForPDBData(pdb_data);
870   auto mangled_cstr = mangled.empty() ? nullptr : mangled.c_str();
871 
872   DWARFExpression location(nullptr);
873 
874   var_sp = std::make_shared<Variable>(
875       var_uid, var_name.c_str(), mangled_cstr, type_sp, scope, context_scope,
876       ranges, &decl, location, is_external, is_artificial, is_static_member);
877 
878   m_variables.insert(std::make_pair(var_uid, var_sp));
879   return var_sp;
880 }
881 
882 size_t
883 SymbolFilePDB::ParseVariables(const lldb_private::SymbolContext &sc,
884                               const llvm::pdb::PDBSymbol &pdb_symbol,
885                               lldb_private::VariableList *variable_list) {
886   size_t num_added = 0;
887 
888   if (auto pdb_data = llvm::dyn_cast<PDBSymbolData>(&pdb_symbol)) {
889     VariableListSP local_variable_list_sp;
890 
891     auto result = m_variables.find(pdb_data->getSymIndexId());
892     if (result != m_variables.end()) {
893       if (variable_list)
894         variable_list->AddVariableIfUnique(result->second);
895     } else {
896       // Prepare right VariableList for this variable.
897       if (auto lexical_parent = pdb_data->getLexicalParent()) {
898         switch (lexical_parent->getSymTag()) {
899         case PDB_SymType::Exe:
900           assert(sc.comp_unit);
901           LLVM_FALLTHROUGH;
902         case PDB_SymType::Compiland: {
903           if (sc.comp_unit) {
904             local_variable_list_sp = sc.comp_unit->GetVariableList(false);
905             if (!local_variable_list_sp) {
906               local_variable_list_sp = std::make_shared<VariableList>();
907               sc.comp_unit->SetVariableList(local_variable_list_sp);
908             }
909           }
910         } break;
911         case PDB_SymType::Block:
912         case PDB_SymType::Function: {
913           if (sc.function) {
914             Block *block = sc.function->GetBlock(true).FindBlockByID(
915                 lexical_parent->getSymIndexId());
916             if (block) {
917               local_variable_list_sp = block->GetBlockVariableList(false);
918               if (!local_variable_list_sp) {
919                 local_variable_list_sp = std::make_shared<VariableList>();
920                 block->SetVariableList(local_variable_list_sp);
921               }
922             }
923           }
924         } break;
925         default:
926           break;
927         }
928       }
929 
930       if (local_variable_list_sp) {
931         if (auto var_sp = ParseVariableForPDBData(sc, *pdb_data)) {
932           local_variable_list_sp->AddVariableIfUnique(var_sp);
933           if (variable_list)
934             variable_list->AddVariableIfUnique(var_sp);
935           ++num_added;
936         }
937       }
938     }
939   }
940 
941   if (auto results = pdb_symbol.findAllChildren()) {
942     while (auto result = results->getNext())
943       num_added += ParseVariables(sc, *result, variable_list);
944   }
945 
946   return num_added;
947 }
948 
949 uint32_t SymbolFilePDB::FindGlobalVariables(
950     const lldb_private::ConstString &name,
951     const lldb_private::CompilerDeclContext *parent_decl_ctx,
952     uint32_t max_matches, lldb_private::VariableList &variables) {
953   if (!DeclContextMatchesThisSymbolFile(parent_decl_ctx))
954     return 0;
955   if (name.IsEmpty())
956     return 0;
957 
958   auto results =
959       m_global_scope_up->findChildren(PDB_SymType::Data, name.GetStringRef(),
960                                       PDB_NameSearchFlags::NS_CaseSensitive);
961   if (!results)
962     return 0;
963 
964   uint32_t matches = 0;
965   size_t old_size = variables.GetSize();
966   while (auto result = results->getNext()) {
967     auto pdb_data = llvm::dyn_cast<PDBSymbolData>(result.get());
968     if (max_matches > 0 && matches >= max_matches)
969       break;
970 
971     SymbolContext sc;
972     sc.module_sp = m_obj_file->GetModule();
973     lldbassert(sc.module_sp.get());
974 
975     sc.comp_unit = ParseCompileUnitForUID(pdb_data->getCompilandId()).get();
976     // FIXME: We are not able to determine the compile unit.
977     if (sc.comp_unit == nullptr)
978       continue;
979 
980     ParseVariables(sc, *pdb_data, &variables);
981     matches = variables.GetSize() - old_size;
982   }
983 
984   return matches;
985 }
986 
987 uint32_t
988 SymbolFilePDB::FindGlobalVariables(const lldb_private::RegularExpression &regex,
989                                    uint32_t max_matches,
990                                    lldb_private::VariableList &variables) {
991   if (!regex.IsValid())
992     return 0;
993   auto results = m_global_scope_up->findAllChildren<PDBSymbolData>();
994   if (!results)
995     return 0;
996 
997   uint32_t matches = 0;
998   size_t old_size = variables.GetSize();
999   while (auto pdb_data = results->getNext()) {
1000     if (max_matches > 0 && matches >= max_matches)
1001       break;
1002 
1003     auto var_name = pdb_data->getName();
1004     if (var_name.empty())
1005       continue;
1006     if (!regex.Execute(var_name))
1007       continue;
1008     SymbolContext sc;
1009     sc.module_sp = m_obj_file->GetModule();
1010     lldbassert(sc.module_sp.get());
1011 
1012     sc.comp_unit = ParseCompileUnitForUID(pdb_data->getCompilandId()).get();
1013     // FIXME: We are not able to determine the compile unit.
1014     if (sc.comp_unit == nullptr)
1015       continue;
1016 
1017     ParseVariables(sc, *pdb_data, &variables);
1018     matches = variables.GetSize() - old_size;
1019   }
1020 
1021   return matches;
1022 }
1023 
1024 bool SymbolFilePDB::ResolveFunction(const llvm::pdb::PDBSymbolFunc &pdb_func,
1025                                     bool include_inlines,
1026                                     lldb_private::SymbolContextList &sc_list) {
1027   lldb_private::SymbolContext sc;
1028   sc.comp_unit = ParseCompileUnitForUID(pdb_func.getCompilandId()).get();
1029   if (!sc.comp_unit)
1030     return false;
1031   sc.module_sp = sc.comp_unit->GetModule();
1032   sc.function = ParseCompileUnitFunctionForPDBFunc(pdb_func, sc);
1033   if (!sc.function)
1034     return false;
1035 
1036   sc_list.Append(sc);
1037   return true;
1038 }
1039 
1040 bool SymbolFilePDB::ResolveFunction(uint32_t uid, bool include_inlines,
1041                                     lldb_private::SymbolContextList &sc_list) {
1042   auto pdb_func_up = m_session_up->getConcreteSymbolById<PDBSymbolFunc>(uid);
1043   if (!pdb_func_up && !(include_inlines && pdb_func_up->hasInlineAttribute()))
1044     return false;
1045   return ResolveFunction(*pdb_func_up, include_inlines, sc_list);
1046 }
1047 
1048 void SymbolFilePDB::CacheFunctionNames() {
1049   if (!m_func_full_names.IsEmpty())
1050     return;
1051 
1052   std::map<uint64_t, uint32_t> addr_ids;
1053 
1054   if (auto results_up = m_global_scope_up->findAllChildren<PDBSymbolFunc>()) {
1055     while (auto pdb_func_up = results_up->getNext()) {
1056       if (pdb_func_up->isCompilerGenerated())
1057         continue;
1058 
1059       auto name = pdb_func_up->getName();
1060       auto demangled_name = pdb_func_up->getUndecoratedName();
1061       if (name.empty() && demangled_name.empty())
1062         continue;
1063 
1064       auto uid = pdb_func_up->getSymIndexId();
1065       if (!demangled_name.empty() && pdb_func_up->getVirtualAddress())
1066         addr_ids.insert(std::make_pair(pdb_func_up->getVirtualAddress(), uid));
1067 
1068       if (auto parent = pdb_func_up->getClassParent()) {
1069 
1070         // PDB have symbols for class/struct methods or static methods in Enum
1071         // Class. We won't bother to check if the parent is UDT or Enum here.
1072         m_func_method_names.Append(ConstString(name), uid);
1073 
1074         ConstString cstr_name(name);
1075 
1076         // To search a method name, like NS::Class:MemberFunc, LLDB searches
1077         // its base name, i.e. MemberFunc by default. Since PDBSymbolFunc does
1078         // not have inforamtion of this, we extract base names and cache them
1079         // by our own effort.
1080         llvm::StringRef basename;
1081         CPlusPlusLanguage::MethodName cpp_method(cstr_name);
1082         if (cpp_method.IsValid()) {
1083           llvm::StringRef context;
1084           basename = cpp_method.GetBasename();
1085           if (basename.empty())
1086             CPlusPlusLanguage::ExtractContextAndIdentifier(name.c_str(),
1087                                                            context, basename);
1088         }
1089 
1090         if (!basename.empty())
1091           m_func_base_names.Append(ConstString(basename), uid);
1092         else {
1093           m_func_base_names.Append(ConstString(name), uid);
1094         }
1095 
1096         if (!demangled_name.empty())
1097           m_func_full_names.Append(ConstString(demangled_name), uid);
1098 
1099       } else {
1100         // Handle not-method symbols.
1101 
1102         // The function name might contain namespace, or its lexical scope. It
1103         // is not safe to get its base name by applying same scheme as we deal
1104         // with the method names.
1105         // FIXME: Remove namespace if function is static in a scope.
1106         m_func_base_names.Append(ConstString(name), uid);
1107 
1108         if (name == "main") {
1109           m_func_full_names.Append(ConstString(name), uid);
1110 
1111           if (!demangled_name.empty() && name != demangled_name) {
1112             m_func_full_names.Append(ConstString(demangled_name), uid);
1113             m_func_base_names.Append(ConstString(demangled_name), uid);
1114           }
1115         } else if (!demangled_name.empty()) {
1116           m_func_full_names.Append(ConstString(demangled_name), uid);
1117         } else {
1118           m_func_full_names.Append(ConstString(name), uid);
1119         }
1120       }
1121     }
1122   }
1123 
1124   if (auto results_up =
1125           m_global_scope_up->findAllChildren<PDBSymbolPublicSymbol>()) {
1126     while (auto pub_sym_up = results_up->getNext()) {
1127       if (!pub_sym_up->isFunction())
1128         continue;
1129       auto name = pub_sym_up->getName();
1130       if (name.empty())
1131         continue;
1132 
1133       if (CPlusPlusLanguage::IsCPPMangledName(name.c_str())) {
1134         auto vm_addr = pub_sym_up->getVirtualAddress();
1135 
1136         // PDB public symbol has mangled name for its associated function.
1137         if (vm_addr && addr_ids.find(vm_addr) != addr_ids.end()) {
1138           // Cache mangled name.
1139           m_func_full_names.Append(ConstString(name), addr_ids[vm_addr]);
1140         }
1141       }
1142     }
1143   }
1144   // Sort them before value searching is working properly
1145   m_func_full_names.Sort();
1146   m_func_full_names.SizeToFit();
1147   m_func_method_names.Sort();
1148   m_func_method_names.SizeToFit();
1149   m_func_base_names.Sort();
1150   m_func_base_names.SizeToFit();
1151 }
1152 
1153 uint32_t SymbolFilePDB::FindFunctions(
1154     const lldb_private::ConstString &name,
1155     const lldb_private::CompilerDeclContext *parent_decl_ctx,
1156     uint32_t name_type_mask, bool include_inlines, bool append,
1157     lldb_private::SymbolContextList &sc_list) {
1158   if (!append)
1159     sc_list.Clear();
1160   lldbassert((name_type_mask & eFunctionNameTypeAuto) == 0);
1161 
1162   if (name_type_mask == eFunctionNameTypeNone)
1163     return 0;
1164   if (!DeclContextMatchesThisSymbolFile(parent_decl_ctx))
1165     return 0;
1166   if (name.IsEmpty())
1167     return 0;
1168 
1169   auto old_size = sc_list.GetSize();
1170   if (name_type_mask & eFunctionNameTypeFull ||
1171       name_type_mask & eFunctionNameTypeBase ||
1172       name_type_mask & eFunctionNameTypeMethod) {
1173     CacheFunctionNames();
1174 
1175     std::set<uint32_t> resolved_ids;
1176     auto ResolveFn = [include_inlines, &name, &sc_list, &resolved_ids,
1177                       this](UniqueCStringMap<uint32_t> &Names) {
1178       std::vector<uint32_t> ids;
1179       if (Names.GetValues(name, ids)) {
1180         for (auto id : ids) {
1181           if (resolved_ids.find(id) == resolved_ids.end()) {
1182             if (ResolveFunction(id, include_inlines, sc_list))
1183               resolved_ids.insert(id);
1184           }
1185         }
1186       }
1187     };
1188     if (name_type_mask & eFunctionNameTypeFull) {
1189       ResolveFn(m_func_full_names);
1190     }
1191     if (name_type_mask & eFunctionNameTypeBase) {
1192       ResolveFn(m_func_base_names);
1193     }
1194     if (name_type_mask & eFunctionNameTypeMethod) {
1195       ResolveFn(m_func_method_names);
1196     }
1197   }
1198   return sc_list.GetSize() - old_size;
1199 }
1200 
1201 uint32_t
1202 SymbolFilePDB::FindFunctions(const lldb_private::RegularExpression &regex,
1203                              bool include_inlines, bool append,
1204                              lldb_private::SymbolContextList &sc_list) {
1205   if (!append)
1206     sc_list.Clear();
1207   if (!regex.IsValid())
1208     return 0;
1209 
1210   auto old_size = sc_list.GetSize();
1211   CacheFunctionNames();
1212 
1213   std::set<uint32_t> resolved_ids;
1214   auto ResolveFn = [&regex, include_inlines, &sc_list, &resolved_ids,
1215                     this](UniqueCStringMap<uint32_t> &Names) {
1216     std::vector<uint32_t> ids;
1217     if (Names.GetValues(regex, ids)) {
1218       for (auto id : ids) {
1219         if (resolved_ids.find(id) == resolved_ids.end())
1220           if (ResolveFunction(id, include_inlines, sc_list))
1221             resolved_ids.insert(id);
1222       }
1223     }
1224   };
1225   ResolveFn(m_func_full_names);
1226   ResolveFn(m_func_base_names);
1227 
1228   return sc_list.GetSize() - old_size;
1229 }
1230 
1231 void SymbolFilePDB::GetMangledNamesForFunction(
1232     const std::string &scope_qualified_name,
1233     std::vector<lldb_private::ConstString> &mangled_names) {}
1234 
1235 uint32_t SymbolFilePDB::FindTypes(
1236     const lldb_private::SymbolContext &sc,
1237     const lldb_private::ConstString &name,
1238     const lldb_private::CompilerDeclContext *parent_decl_ctx, bool append,
1239     uint32_t max_matches,
1240     llvm::DenseSet<lldb_private::SymbolFile *> &searched_symbol_files,
1241     lldb_private::TypeMap &types) {
1242   if (!append)
1243     types.Clear();
1244   if (!name)
1245     return 0;
1246   if (!DeclContextMatchesThisSymbolFile(parent_decl_ctx))
1247     return 0;
1248 
1249   searched_symbol_files.clear();
1250   searched_symbol_files.insert(this);
1251 
1252   std::string name_str = name.AsCString();
1253 
1254   // There is an assumption 'name' is not a regex
1255   FindTypesByName(name_str, max_matches, types);
1256 
1257   return types.GetSize();
1258 }
1259 
1260 void SymbolFilePDB::FindTypesByRegex(
1261     const lldb_private::RegularExpression &regex, uint32_t max_matches,
1262     lldb_private::TypeMap &types) {
1263   // When searching by regex, we need to go out of our way to limit the search
1264   // space as much as possible since this searches EVERYTHING in the PDB,
1265   // manually doing regex comparisons.  PDB library isn't optimized for regex
1266   // searches or searches across multiple symbol types at the same time, so the
1267   // best we can do is to search enums, then typedefs, then classes one by one,
1268   // and do a regex comparison against each of them.
1269   PDB_SymType tags_to_search[] = {PDB_SymType::Enum, PDB_SymType::Typedef,
1270                                   PDB_SymType::UDT};
1271   std::unique_ptr<IPDBEnumSymbols> results;
1272 
1273   uint32_t matches = 0;
1274 
1275   for (auto tag : tags_to_search) {
1276     results = m_global_scope_up->findAllChildren(tag);
1277     if (!results)
1278       continue;
1279 
1280     while (auto result = results->getNext()) {
1281       if (max_matches > 0 && matches >= max_matches)
1282         break;
1283 
1284       std::string type_name;
1285       if (auto enum_type = llvm::dyn_cast<PDBSymbolTypeEnum>(result.get()))
1286         type_name = enum_type->getName();
1287       else if (auto typedef_type =
1288                    llvm::dyn_cast<PDBSymbolTypeTypedef>(result.get()))
1289         type_name = typedef_type->getName();
1290       else if (auto class_type = llvm::dyn_cast<PDBSymbolTypeUDT>(result.get()))
1291         type_name = class_type->getName();
1292       else {
1293         // We're looking only for types that have names.  Skip symbols, as well
1294         // as unnamed types such as arrays, pointers, etc.
1295         continue;
1296       }
1297 
1298       if (!regex.Execute(type_name))
1299         continue;
1300 
1301       // This should cause the type to get cached and stored in the `m_types`
1302       // lookup.
1303       if (!ResolveTypeUID(result->getSymIndexId()))
1304         continue;
1305 
1306       auto iter = m_types.find(result->getSymIndexId());
1307       if (iter == m_types.end())
1308         continue;
1309       types.Insert(iter->second);
1310       ++matches;
1311     }
1312   }
1313 }
1314 
1315 void SymbolFilePDB::FindTypesByName(const std::string &name,
1316                                     uint32_t max_matches,
1317                                     lldb_private::TypeMap &types) {
1318   std::unique_ptr<IPDBEnumSymbols> results;
1319   if (name.empty())
1320     return;
1321   results = m_global_scope_up->findChildren(PDB_SymType::None, name,
1322                                             PDB_NameSearchFlags::NS_Default);
1323   if (!results)
1324     return;
1325 
1326   uint32_t matches = 0;
1327 
1328   while (auto result = results->getNext()) {
1329     if (max_matches > 0 && matches >= max_matches)
1330       break;
1331     switch (result->getSymTag()) {
1332     case PDB_SymType::Enum:
1333     case PDB_SymType::UDT:
1334     case PDB_SymType::Typedef:
1335       break;
1336     default:
1337       // We're looking only for types that have names.  Skip symbols, as well
1338       // as unnamed types such as arrays, pointers, etc.
1339       continue;
1340     }
1341 
1342     // This should cause the type to get cached and stored in the `m_types`
1343     // lookup.
1344     if (!ResolveTypeUID(result->getSymIndexId()))
1345       continue;
1346 
1347     auto iter = m_types.find(result->getSymIndexId());
1348     if (iter == m_types.end())
1349       continue;
1350     types.Insert(iter->second);
1351     ++matches;
1352   }
1353 }
1354 
1355 size_t SymbolFilePDB::FindTypes(
1356     const std::vector<lldb_private::CompilerContext> &contexts, bool append,
1357     lldb_private::TypeMap &types) {
1358   return 0;
1359 }
1360 
1361 lldb_private::TypeList *SymbolFilePDB::GetTypeList() {
1362   return m_obj_file->GetModule()->GetTypeList();
1363 }
1364 
1365 void SymbolFilePDB::GetTypesForPDBSymbol(const llvm::pdb::PDBSymbol &pdb_symbol,
1366                                          uint32_t type_mask,
1367                                          TypeCollection &type_collection) {
1368   bool can_parse = false;
1369   switch (pdb_symbol.getSymTag()) {
1370   case PDB_SymType::ArrayType:
1371     can_parse = ((type_mask & eTypeClassArray) != 0);
1372     break;
1373   case PDB_SymType::BuiltinType:
1374     can_parse = ((type_mask & eTypeClassBuiltin) != 0);
1375     break;
1376   case PDB_SymType::Enum:
1377     can_parse = ((type_mask & eTypeClassEnumeration) != 0);
1378     break;
1379   case PDB_SymType::Function:
1380   case PDB_SymType::FunctionSig:
1381     can_parse = ((type_mask & eTypeClassFunction) != 0);
1382     break;
1383   case PDB_SymType::PointerType:
1384     can_parse = ((type_mask & (eTypeClassPointer | eTypeClassBlockPointer |
1385                                eTypeClassMemberPointer)) != 0);
1386     break;
1387   case PDB_SymType::Typedef:
1388     can_parse = ((type_mask & eTypeClassTypedef) != 0);
1389     break;
1390   case PDB_SymType::UDT: {
1391     auto *udt = llvm::dyn_cast<PDBSymbolTypeUDT>(&pdb_symbol);
1392     assert(udt);
1393     can_parse = (udt->getUdtKind() != PDB_UdtType::Interface &&
1394                  ((type_mask & (eTypeClassClass | eTypeClassStruct |
1395                                 eTypeClassUnion)) != 0));
1396   } break;
1397   default:
1398     break;
1399   }
1400 
1401   if (can_parse) {
1402     if (auto *type = ResolveTypeUID(pdb_symbol.getSymIndexId())) {
1403       auto result =
1404           std::find(type_collection.begin(), type_collection.end(), type);
1405       if (result == type_collection.end())
1406         type_collection.push_back(type);
1407     }
1408   }
1409 
1410   auto results_up = pdb_symbol.findAllChildren();
1411   while (auto symbol_up = results_up->getNext())
1412     GetTypesForPDBSymbol(*symbol_up, type_mask, type_collection);
1413 }
1414 
1415 size_t SymbolFilePDB::GetTypes(lldb_private::SymbolContextScope *sc_scope,
1416                                uint32_t type_mask,
1417                                lldb_private::TypeList &type_list) {
1418   TypeCollection type_collection;
1419   uint32_t old_size = type_list.GetSize();
1420   CompileUnit *cu =
1421       sc_scope ? sc_scope->CalculateSymbolContextCompileUnit() : nullptr;
1422   if (cu) {
1423     auto compiland_up = GetPDBCompilandByUID(cu->GetID());
1424     if (!compiland_up)
1425       return 0;
1426     GetTypesForPDBSymbol(*compiland_up, type_mask, type_collection);
1427   } else {
1428     for (uint32_t cu_idx = 0; cu_idx < GetNumCompileUnits(); ++cu_idx) {
1429       auto cu_sp = ParseCompileUnitAtIndex(cu_idx);
1430       if (cu_sp) {
1431         if (auto compiland_up = GetPDBCompilandByUID(cu_sp->GetID()))
1432           GetTypesForPDBSymbol(*compiland_up, type_mask, type_collection);
1433       }
1434     }
1435   }
1436 
1437   for (auto type : type_collection) {
1438     type->GetForwardCompilerType();
1439     type_list.Insert(type->shared_from_this());
1440   }
1441   return type_list.GetSize() - old_size;
1442 }
1443 
1444 lldb_private::TypeSystem *
1445 SymbolFilePDB::GetTypeSystemForLanguage(lldb::LanguageType language) {
1446   auto type_system =
1447       m_obj_file->GetModule()->GetTypeSystemForLanguage(language);
1448   if (type_system)
1449     type_system->SetSymbolFile(this);
1450   return type_system;
1451 }
1452 
1453 lldb_private::CompilerDeclContext SymbolFilePDB::FindNamespace(
1454     const lldb_private::SymbolContext &sc,
1455     const lldb_private::ConstString &name,
1456     const lldb_private::CompilerDeclContext *parent_decl_ctx) {
1457   return lldb_private::CompilerDeclContext();
1458 }
1459 
1460 lldb_private::ConstString SymbolFilePDB::GetPluginName() {
1461   static ConstString g_name("pdb");
1462   return g_name;
1463 }
1464 
1465 uint32_t SymbolFilePDB::GetPluginVersion() { return 1; }
1466 
1467 IPDBSession &SymbolFilePDB::GetPDBSession() { return *m_session_up; }
1468 
1469 const IPDBSession &SymbolFilePDB::GetPDBSession() const {
1470   return *m_session_up;
1471 }
1472 
1473 lldb::CompUnitSP SymbolFilePDB::ParseCompileUnitForUID(uint32_t id,
1474                                                        uint32_t index) {
1475   auto found_cu = m_comp_units.find(id);
1476   if (found_cu != m_comp_units.end())
1477     return found_cu->second;
1478 
1479   auto compiland_up = GetPDBCompilandByUID(id);
1480   if (!compiland_up)
1481     return CompUnitSP();
1482 
1483   lldb::LanguageType lang;
1484   auto details = compiland_up->findOneChild<PDBSymbolCompilandDetails>();
1485   if (!details)
1486     lang = lldb::eLanguageTypeC_plus_plus;
1487   else
1488     lang = TranslateLanguage(details->getLanguage());
1489 
1490   if (lang == lldb::LanguageType::eLanguageTypeUnknown)
1491     return CompUnitSP();
1492 
1493   std::string path = compiland_up->getSourceFileFullPath();
1494   if (path.empty())
1495     return CompUnitSP();
1496 
1497   // Don't support optimized code for now, DebugInfoPDB does not return this
1498   // information.
1499   LazyBool optimized = eLazyBoolNo;
1500   auto cu_sp = std::make_shared<CompileUnit>(m_obj_file->GetModule(), nullptr,
1501                                              path.c_str(), id, lang, optimized);
1502 
1503   if (!cu_sp)
1504     return CompUnitSP();
1505 
1506   m_comp_units.insert(std::make_pair(id, cu_sp));
1507   if (index == UINT32_MAX)
1508     GetCompileUnitIndex(*compiland_up, index);
1509   lldbassert(index != UINT32_MAX);
1510   m_obj_file->GetModule()->GetSymbolVendor()->SetCompileUnitAtIndex(index,
1511                                                                     cu_sp);
1512   return cu_sp;
1513 }
1514 
1515 bool SymbolFilePDB::ParseCompileUnitLineTable(
1516     const lldb_private::SymbolContext &sc, uint32_t match_line) {
1517   lldbassert(sc.comp_unit);
1518 
1519   auto compiland_up = GetPDBCompilandByUID(sc.comp_unit->GetID());
1520   if (!compiland_up)
1521     return false;
1522 
1523   // LineEntry needs the *index* of the file into the list of support files
1524   // returned by ParseCompileUnitSupportFiles.  But the underlying SDK gives us
1525   // a globally unique idenfitifier in the namespace of the PDB.  So, we have
1526   // to do a mapping so that we can hand out indices.
1527   llvm::DenseMap<uint32_t, uint32_t> index_map;
1528   BuildSupportFileIdToSupportFileIndexMap(*compiland_up, index_map);
1529   auto line_table = llvm::make_unique<LineTable>(sc.comp_unit);
1530 
1531   // Find contributions to `compiland` from all source and header files.
1532   std::string path = sc.comp_unit->GetPath();
1533   auto files = m_session_up->getSourceFilesForCompiland(*compiland_up);
1534   if (!files)
1535     return false;
1536 
1537   // For each source and header file, create a LineSequence for contributions
1538   // to the compiland from that file, and add the sequence.
1539   while (auto file = files->getNext()) {
1540     std::unique_ptr<LineSequence> sequence(
1541         line_table->CreateLineSequenceContainer());
1542     auto lines = m_session_up->findLineNumbers(*compiland_up, *file);
1543     if (!lines)
1544       continue;
1545     int entry_count = lines->getChildCount();
1546 
1547     uint64_t prev_addr;
1548     uint32_t prev_length;
1549     uint32_t prev_line;
1550     uint32_t prev_source_idx;
1551 
1552     for (int i = 0; i < entry_count; ++i) {
1553       auto line = lines->getChildAtIndex(i);
1554 
1555       uint64_t lno = line->getLineNumber();
1556       uint64_t addr = line->getVirtualAddress();
1557       uint32_t length = line->getLength();
1558       uint32_t source_id = line->getSourceFileId();
1559       uint32_t col = line->getColumnNumber();
1560       uint32_t source_idx = index_map[source_id];
1561 
1562       // There was a gap between the current entry and the previous entry if
1563       // the addresses don't perfectly line up.
1564       bool is_gap = (i > 0) && (prev_addr + prev_length < addr);
1565 
1566       // Before inserting the current entry, insert a terminal entry at the end
1567       // of the previous entry's address range if the current entry resulted in
1568       // a gap from the previous entry.
1569       if (is_gap && ShouldAddLine(match_line, prev_line, prev_length)) {
1570         line_table->AppendLineEntryToSequence(
1571             sequence.get(), prev_addr + prev_length, prev_line, 0,
1572             prev_source_idx, false, false, false, false, true);
1573 
1574         line_table->InsertSequence(sequence.release());
1575         sequence.reset(line_table->CreateLineSequenceContainer());
1576       }
1577 
1578       if (ShouldAddLine(match_line, lno, length)) {
1579         bool is_statement = line->isStatement();
1580         bool is_prologue = false;
1581         bool is_epilogue = false;
1582         auto func =
1583             m_session_up->findSymbolByAddress(addr, PDB_SymType::Function);
1584         if (func) {
1585           auto prologue = func->findOneChild<PDBSymbolFuncDebugStart>();
1586           if (prologue)
1587             is_prologue = (addr == prologue->getVirtualAddress());
1588 
1589           auto epilogue = func->findOneChild<PDBSymbolFuncDebugEnd>();
1590           if (epilogue)
1591             is_epilogue = (addr == epilogue->getVirtualAddress());
1592         }
1593 
1594         line_table->AppendLineEntryToSequence(sequence.get(), addr, lno, col,
1595                                               source_idx, is_statement, false,
1596                                               is_prologue, is_epilogue, false);
1597       }
1598 
1599       prev_addr = addr;
1600       prev_length = length;
1601       prev_line = lno;
1602       prev_source_idx = source_idx;
1603     }
1604 
1605     if (entry_count > 0 && ShouldAddLine(match_line, prev_line, prev_length)) {
1606       // The end is always a terminal entry, so insert it regardless.
1607       line_table->AppendLineEntryToSequence(
1608           sequence.get(), prev_addr + prev_length, prev_line, 0,
1609           prev_source_idx, false, false, false, false, true);
1610     }
1611 
1612     line_table->InsertSequence(sequence.release());
1613   }
1614 
1615   if (line_table->GetSize()) {
1616     sc.comp_unit->SetLineTable(line_table.release());
1617     return true;
1618   }
1619   return false;
1620 }
1621 
1622 void SymbolFilePDB::BuildSupportFileIdToSupportFileIndexMap(
1623     const PDBSymbolCompiland &compiland,
1624     llvm::DenseMap<uint32_t, uint32_t> &index_map) const {
1625   // This is a hack, but we need to convert the source id into an index into
1626   // the support files array.  We don't want to do path comparisons to avoid
1627   // basename / full path issues that may or may not even be a problem, so we
1628   // use the globally unique source file identifiers.  Ideally we could use the
1629   // global identifiers everywhere, but LineEntry currently assumes indices.
1630   auto source_files = m_session_up->getSourceFilesForCompiland(compiland);
1631   if (!source_files)
1632     return;
1633 
1634   // LLDB uses the DWARF-like file numeration (one based)
1635   int index = 1;
1636 
1637   while (auto file = source_files->getNext()) {
1638     uint32_t source_id = file->getUniqueId();
1639     index_map[source_id] = index++;
1640   }
1641 }
1642 
1643 lldb::CompUnitSP SymbolFilePDB::GetCompileUnitContainsAddress(
1644     const lldb_private::Address &so_addr) {
1645   lldb::addr_t file_vm_addr = so_addr.GetFileAddress();
1646   if (file_vm_addr == LLDB_INVALID_ADDRESS || file_vm_addr == 0)
1647     return nullptr;
1648 
1649   // If it is a PDB function's vm addr, this is the first sure bet.
1650   if (auto lines =
1651           m_session_up->findLineNumbersByAddress(file_vm_addr, /*Length=*/1)) {
1652     if (auto first_line = lines->getNext())
1653       return ParseCompileUnitForUID(first_line->getCompilandId());
1654   }
1655 
1656   // Otherwise we resort to section contributions.
1657   if (auto sec_contribs = m_session_up->getSectionContribs()) {
1658     while (auto section = sec_contribs->getNext()) {
1659       auto va = section->getVirtualAddress();
1660       if (file_vm_addr >= va && file_vm_addr < va + section->getLength())
1661         return ParseCompileUnitForUID(section->getCompilandId());
1662     }
1663   }
1664   return nullptr;
1665 }
1666 
1667 Mangled
1668 SymbolFilePDB::GetMangledForPDBFunc(const llvm::pdb::PDBSymbolFunc &pdb_func) {
1669   Mangled mangled;
1670   auto func_name = pdb_func.getName();
1671   auto func_undecorated_name = pdb_func.getUndecoratedName();
1672   std::string func_decorated_name;
1673 
1674   // Seek from public symbols for non-static function's decorated name if any.
1675   // For static functions, they don't have undecorated names and aren't exposed
1676   // in Public Symbols either.
1677   if (!func_undecorated_name.empty()) {
1678     auto result_up = m_global_scope_up->findChildren(
1679         PDB_SymType::PublicSymbol, func_undecorated_name,
1680         PDB_NameSearchFlags::NS_UndecoratedName);
1681     if (result_up) {
1682       while (auto symbol_up = result_up->getNext()) {
1683         // For a public symbol, it is unique.
1684         lldbassert(result_up->getChildCount() == 1);
1685         if (auto *pdb_public_sym =
1686                 llvm::dyn_cast_or_null<PDBSymbolPublicSymbol>(
1687                     symbol_up.get())) {
1688           if (pdb_public_sym->isFunction()) {
1689             func_decorated_name = pdb_public_sym->getName();
1690             break;
1691           }
1692         }
1693       }
1694     }
1695   }
1696   if (!func_decorated_name.empty()) {
1697     mangled.SetMangledName(ConstString(func_decorated_name));
1698 
1699     // For MSVC, format of C funciton's decorated name depends on calling
1700     // conventon. Unfortunately none of the format is recognized by current
1701     // LLDB. For example, `_purecall` is a __cdecl C function. From PDB,
1702     // `__purecall` is retrieved as both its decorated and undecorated name
1703     // (using PDBSymbolFunc::getUndecoratedName method). However `__purecall`
1704     // string is not treated as mangled in LLDB (neither `?` nor `_Z` prefix).
1705     // Mangled::GetDemangledName method will fail internally and caches an
1706     // empty string as its undecorated name. So we will face a contradition
1707     // here for the same symbol:
1708     //   non-empty undecorated name from PDB
1709     //   empty undecorated name from LLDB
1710     if (!func_undecorated_name.empty() &&
1711         mangled.GetDemangledName(mangled.GuessLanguage()).IsEmpty())
1712       mangled.SetDemangledName(ConstString(func_undecorated_name));
1713 
1714     // LLDB uses several flags to control how a C++ decorated name is
1715     // undecorated for MSVC. See `safeUndecorateName` in Class Mangled. So the
1716     // yielded name could be different from what we retrieve from
1717     // PDB source unless we also apply same flags in getting undecorated
1718     // name through PDBSymbolFunc::getUndecoratedNameEx method.
1719     if (!func_undecorated_name.empty() &&
1720         mangled.GetDemangledName(mangled.GuessLanguage()) !=
1721             ConstString(func_undecorated_name))
1722       mangled.SetDemangledName(ConstString(func_undecorated_name));
1723   } else if (!func_undecorated_name.empty()) {
1724     mangled.SetDemangledName(ConstString(func_undecorated_name));
1725   } else if (!func_name.empty())
1726     mangled.SetValue(ConstString(func_name), false);
1727 
1728   return mangled;
1729 }
1730 
1731 bool SymbolFilePDB::DeclContextMatchesThisSymbolFile(
1732     const lldb_private::CompilerDeclContext *decl_ctx) {
1733   if (decl_ctx == nullptr || !decl_ctx->IsValid())
1734     return true;
1735 
1736   TypeSystem *decl_ctx_type_system = decl_ctx->GetTypeSystem();
1737   if (!decl_ctx_type_system)
1738     return false;
1739   TypeSystem *type_system = GetTypeSystemForLanguage(
1740       decl_ctx_type_system->GetMinimumLanguage(nullptr));
1741   if (decl_ctx_type_system == type_system)
1742     return true; // The type systems match, return true
1743 
1744   return false;
1745 }
1746