[dsymutil] Better support for symbol aliases

[JDevlieghere]

Summary

This patch adds logic to follow a symbol's aliases when the symbol name
cannot be found in the current object file. It checks the main binary
for the symbol's address and queries the current object for its aliases
(symbols with the same address) before printing out a warning.

[aprantl]

tools/dsymutil/MachODebugMapParser.cpp

400 ↗

(On Diff #116508)

The LLVM style usually avoids bracing single statements.

450 ↗

(On Diff #116508)

range-based for?

[friss]

tools/dsymutil/MachODebugMapParser.cpp

389–395 ↗

(On Diff #116508)

I'm not sure I could construct a test case to trigger a bug here, but I think it would be safer to add as "Type != MachO::N_STSYM" condition to this if. I don't think it's possible to alias static symbols.

[JDevlieghere]

• Feedback Adrian & Fred

llvm/trunk/test/tools/dsymutil/X86/alias.test

+# RUN: llvm-dsymutil -f -oso-prepend-path=%p/../Inputs/alias \
+# RUN: %p/../Inputs/alias/foobar -o - 2>&1 | llvm-dwarfdump - | FileCheck %s
+# CHECK-NOT: could not find object file symbol for symbol
+# CHECK: DW_AT_name ("foo.c")
+# CHECK: DW_AT_name ("bar.c")
+
+# Source:
+#   $ cat foo.c
+#   int foo = 1;
+#   $ cat bar.c
+#   extern int bar;
+#   int main() {
+#     return bar;
+#   }
+
+# Compile with:
+#   $ clang -g -O0 bar.c -c -o bar.o
+#   $ clang -g -O0 foo.c -c -o foo.o
+#   $ ld -arch x86_64 -macosx_version_min 10.13.0 foo.o bar.o -lSystem -alias _foo _bar -o foobar

llvm/trunk/tools/dsymutil/MachODebugMapParser.cpp

private:
   std::unique_ptr<DebugMap> parseOneBinary(const MachOObjectFile &MainBinary,
                                            StringRef BinaryPath);
 
   void
   switchToNewDebugMapObject(StringRef Filename,
                             sys::TimePoint<std::chrono::seconds> Timestamp);
   void resetParserState();
   uint64_t getMainBinarySymbolAddress(StringRef Name);
+  std::vector<StringRef> getMainBinarySymbolNames(uint64_t Value);
   void loadMainBinarySymbols(const MachOObjectFile &MainBinary);
   void loadCurrentObjectFileSymbols(const object::MachOObjectFile &Obj);
   void handleStabSymbolTableEntry(uint32_t StringIndex, uint8_t Type,
                                   uint8_t SectionIndex, uint16_t Flags,
                                   uint64_t Value);
 
   template <typename STEType> void handleStabDebugMapEntry(const STEType &STE) {
     handleStabSymbolTableEntry(STE.n_strx, STE.n_type, STE.n_sect, STE.n_desc,
case MachO::N_FUN:
     }
   case MachO::N_STSYM:
     break;
   default:
     return;
   }
 
   auto ObjectSymIt = CurrentObjectAddresses.find(Name);
+
+  // If the name of a (non-static) symbol is not in the current object, we
+  // check all its aliases from the main binary.
+  if (ObjectSymIt == CurrentObjectAddresses.end() && Type != MachO::N_STSYM) {
+    for (const auto &Alias : getMainBinarySymbolNames(Value)) {
+      ObjectSymIt = CurrentObjectAddresses.find(Alias);
+      if (ObjectSymIt != CurrentObjectAddresses.end())
+        break;
+    }
+  }
+
   if (ObjectSymIt == CurrentObjectAddresses.end())
     return Warning("could not find object file symbol for symbol " +
                    Twine(Name));
+
   if (!CurrentDebugMapObject->addSymbol(Name, ObjectSymIt->getValue(), Value,
                                         Size))
     return Warning(Twine("failed to insert symbol '") + Name +
                    "' in the debug map.");
 }
 
 /// Load the current object file symbols into CurrentObjectAddresses.
 void MachODebugMapParser::loadCurrentObjectFileSymbols(

 /// address is available through this function.
 uint64_t MachODebugMapParser::getMainBinarySymbolAddress(StringRef Name) {
   auto Sym = MainBinarySymbolAddresses.find(Name);
   if (Sym == MainBinarySymbolAddresses.end())
     return 0;
   return Sym->second;
 }
 
+/// Get all symbol names in the main binary for the given value.
+std::vector<StringRef>
+MachODebugMapParser::getMainBinarySymbolNames(uint64_t Value) {
+  std::vector<StringRef> Names;
+  for (const auto &Entry : MainBinarySymbolAddresses) {
+    if (Entry.second == Value)
+      Names.push_back(Entry.first());
+  }
+  return Names;
+}
+
 /// Load the interesting main binary symbols' addresses into
 /// MainBinarySymbolAddresses.
 void MachODebugMapParser::loadMainBinarySymbols(
     const MachOObjectFile &MainBinary) {
   section_iterator Section = MainBinary.section_end();
   MainBinarySymbolAddresses.clear();
   for (const auto &Sym : MainBinary.symbols()) {
     Expected<SymbolRef::Type> TypeOrErr = Sym.getType();