Rename embedded bitcode section in MachO

[steven_wu]

Summary

Rename the section embeds bitcode from ".llvmbc,.llvmbc" to "LLVM,bitcode".
The new name matches MachO section naming convention.

[• rafael]

Don't you also need to check the segment name on macho? Or it is just invalid to create a FOO, bitcode?

[davide]

lib/Object/FunctionIndexObjectFile.cpp

42

You can just inline the call here (also true in the other case), as there's only one use of this variable.

lib/Object/ObjectFile.cpp

119

this can be spelled in a more compact form.
return (Obj.isMachO()) ? "__bitcode" : ".llvmbc";

[steven_wu]

It will probably be better to check the segment name. I was hesitate to do
that because I couldn't figure out a way without adding more virtual method.
Here is a try to make it more robust. Note it does slightly change the
behavior because std::error not longer propagates through findBitcodeInObject.

[• rafael]

I think this is fine given that it just mean that a section with a corrupted name is not bitcode. That is the same answer you would get if the corruption had changed the name to __bitcodx.

pcc, what do you think?

[steven_wu]

Ping.

[pcc]

LGTM

docs/BitCodeFormat.rst

 encode the CPU of the target.
 
 .. _native object file:
 
 Native Object File Wrapper Format
 =================================
 
 Bitcode files for LLVM IR may also be wrapped in a native object file
-(i.e. ELF, COFF, Mach-O).  The bitcode must be stored in a section of the
-object file named ``.llvmbc``.  This wrapper format is useful for accommodating
-LTO in compilation pipelines where intermediate objects must be native object
-files which contain metadata in other sections.
+(i.e. ELF, COFF, Mach-O).  The bitcode must be stored in a section of the object
+file named ``__LLVM,__bitcode`` for MachO and ``.llvmbc`` for the other object
+formats.  This wrapper format is useful for accommodating LTO in compilation
+pipelines where intermediate objects must be native object files which contain
+metadata in other sections.
 
 Not all tools support this format.
 
 .. _encoding of LLVM IR:
 
 LLVM IR Encoding
 ================
 

include/llvm/Object/ObjectFile.h

public:
   static ErrorOr<std::unique_ptr<COFFObjectFile>>
   createCOFFObjectFile(MemoryBufferRef Object);
 
   static ErrorOr<std::unique_ptr<ObjectFile>>
   createELFObjectFile(MemoryBufferRef Object);
 
   static ErrorOr<std::unique_ptr<MachOObjectFile>>
   createMachOObjectFile(MemoryBufferRef Object);
+
+  static const char* getSectionNameForBitcode(const ObjectFile &Obj);
+
 };
 
 // Inline function definitions.
 inline SymbolRef::SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner)
     : BasicSymbolRef(SymbolP, Owner) {}
 
 inline ErrorOr<StringRef> SymbolRef::getName() const {
   return getObject()->getSymbolName(getRawDataRefImpl());

lib/Object/FunctionIndexObjectFile.cpp

 FunctionIndexObjectFile::~FunctionIndexObjectFile() {}
 
 std::unique_ptr<FunctionInfoIndex> FunctionIndexObjectFile::takeIndex() {
   return std::move(Index);
 }
 
 ErrorOr<MemoryBufferRef>
 FunctionIndexObjectFile::findBitcodeInObject(const ObjectFile &Obj) {
+  const char* BitcodeSectName = ObjectFile::getSectionNameForBitcode(Obj);
   for (const SectionRef &Sec : Obj.sections()) {
     StringRef SecName;
     if (std::error_code EC = Sec.getName(SecName))
       return EC;
-    if (SecName == ".llvmbc") {
+    if (SecName == BitcodeSectName) {
       StringRef SecContents;
       if (std::error_code EC = Sec.getContents(SecContents))
         return EC;
       return MemoryBufferRef(SecContents, Obj.getFileName());
     }
   }
 
   return object_error::bitcode_section_not_found;

lib/Object/IRObjectFile.cpp

basic_symbol_iterator IRObjectFile::symbol_end_impl() const {
   DataRefImpl Ret;
   uint64_t NumAsm = AsmSymbols.size();
   NumAsm <<= 2;
   Ret.p = 3 | NumAsm;
   return basic_symbol_iterator(BasicSymbolRef(Ret, this));
 }
 
 ErrorOr<MemoryBufferRef> IRObjectFile::findBitcodeInObject(const ObjectFile &Obj) {
+  const char* BitcodeSectName = ObjectFile::getSectionNameForBitcode(Obj);
   for (const SectionRef &Sec : Obj.sections()) {
     StringRef SecName;
     if (std::error_code EC = Sec.getName(SecName))
       return EC;
-    if (SecName == ".llvmbc") {
+    if (SecName == BitcodeSectName) {
       StringRef SecContents;
       if (std::error_code EC = Sec.getContents(SecContents))
         return EC;
       return MemoryBufferRef(SecContents, Obj.getFileName());
     }
   }
 
   return object_error::bitcode_section_not_found;

lib/Object/ObjectFile.cpp

ObjectFile::createObjectFile(StringRef ObjectPath) {
   ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
       createObjectFile(Buffer->getMemBufferRef());
   if (std::error_code EC = ObjOrErr.getError())
     return EC;
   std::unique_ptr<ObjectFile> Obj = std::move(ObjOrErr.get());
 
   return OwningBinary<ObjectFile>(std::move(Obj), std::move(Buffer));
 }
+
+const char* ObjectFile::getSectionNameForBitcode(const ObjectFile &Obj) {
+  if (Obj.isMachO())
+    return "__bitcode";
+  else
+    return ".llvmbc";
+}

test/LTO/X86/Inputs/bcsection.macho.s

-.section .llvmbc,.llvmbc
+.section __LLVM,__bitcode
 .incbin "bcsection.bc"