[CodeView] Minimal support for S_UNAMESPACE records

[aganea]

Summary

S_UNAMESPACE records are generated by the MSVC compiler when using namespace... directives are used.

Such as:

// SimpleFunction.cpp
// > cl /c SimpleFunction.cpp /Z7
namespace test { }
using namespace test;
int main(void) {
  return 0;
}

This change inherently fixes some 'ignoring unknown symbol record with kind 0x' when passing lld-link -verbose with MSVC .objs.

[zturner]

Don't think I've ever seen S_UNAMESPACE before. Do you know what it is?

[aganea]

In D50007#1180965, @zturner wrote:

Don't think I've ever seen S_UNAMESPACE before. Do you know what it is?

It seems to be related to "using namespace" directives and precompiled headers (/Yc).

Compiling the following as precompiled headers (with cl.exe /Yc):

#include <stdio.h>
#include <Windows.h>

...generates:

C:\Users\aganea\Desktop\LLD verbose>cvdump temp\lib.obj | more
Microsoft (R) Debugging Information Dumper  Version 14.00.23611
Copyright (C) Microsoft Corporation.  All rights reserved.


***** SECTION #2

*** SYMBOLS

(00000C) S_OBJNAME: Signature: 1FD3C851, C:\Users\aganea\Desktop\LLD verbose\temp\lib.obj

(000045) S_COMPILE3:
         Language: C++
         Target processor: x64
         Compiled for edit and continue: no
         Compiled without debugging info: no
         Compiled with LTCG: no
         Compiled with /bzalign: no
         Managed code present: no
         Compiled with /GS: yes
         Compiled with /hotpatch: yes
         Converted by CVTCIL: no
         MSIL module: no
         Compiled with /sdl: no
         Compiled with pgo: no
         .EXP module: no
         Pad bits = 0x0000
         Frontend Version: Major = 19, Minor = 14, Build = 26431, QFE = 0
         Backend Version: Major = 19, Minor = 14, Build = 26431, QFE = 0
         Version string: Microsoft (R) Optimizing Compiler

(000081) S_UNAMESPACE: __vc_attributes
(000095) S_UNAMESPACE: helper_attributes
(0000AB) S_UNAMESPACE: atl
(0000B3) S_UNAMESPACE: std

[aganea]

This will be covered by D45213, but if you prefer I could add a test along with this?

[zturner]

Can you update add a proper class for this symbol? Even though it's very simple, just having that class there serves as useful documentation of its format. The steps would be, roughly:

1. Update CodeViewSymbols.def and change this from a CV_SYMBOL to a SYMBOL_RECORD.
2. In SymbolRecord.h, add a class for this (looks like it probably has only 1 field).
3. Update getSymbolNameOffset function in RecordName.cpp to return the correct offset for this symbol.
4. Fix up any compiler errors that occur as a result of needing to implement virtual functions on things that were ignoring this record before.

It should be pretty straightforward with just boilerplate changes. But after that I think this will be good to go.

[zturner]

Very nice, thanks!

llvm/trunk/include/llvm/DebugInfo/CodeView/CodeViewSymbols.def

 
 // Managed code symbols.
 CV_SYMBOL(S_MANFRAMEREL   , 0x111e)
 CV_SYMBOL(S_MANREGISTER   , 0x111f)
 CV_SYMBOL(S_MANSLOT       , 0x1120)
 CV_SYMBOL(S_MANMANYREG    , 0x1121)
 CV_SYMBOL(S_MANREGREL     , 0x1122)
 CV_SYMBOL(S_MANMANYREG2   , 0x1123)
-CV_SYMBOL(S_UNAMESPACE    , 0x1124)
 CV_SYMBOL(S_DATAREF       , 0x1126)
 CV_SYMBOL(S_ANNOTATIONREF , 0x1128)
 CV_SYMBOL(S_TOKENREF      , 0x1129)
 CV_SYMBOL(S_GMANPROC      , 0x112a)
 CV_SYMBOL(S_LMANPROC      , 0x112b)
 CV_SYMBOL(S_ATTR_FRAMEREL , 0x112e)
 CV_SYMBOL(S_ATTR_REGISTER , 0x112f)
 CV_SYMBOL(S_ATTR_REGREL   , 0x1130)

 SYMBOL_RECORD(S_LDATA32       , 0x110c, DataSym)
 SYMBOL_RECORD_ALIAS(S_GDATA32       , 0x110d, GlobalData, DataSym)
 SYMBOL_RECORD_ALIAS(S_LMANDATA      , 0x111c, ManagedLocalData, DataSym)
 SYMBOL_RECORD_ALIAS(S_GMANDATA      , 0x111d, ManagedGlobalData, DataSym)
 
 SYMBOL_RECORD(S_LTHREAD32     , 0x1112, ThreadLocalDataSym)
 SYMBOL_RECORD_ALIAS(S_GTHREAD32     , 0x1113, GlobalTLS, ThreadLocalDataSym)
 
+SYMBOL_RECORD(S_UNAMESPACE    , 0x1124, UsingNamespaceSym)
 
 #undef CV_SYMBOL
 #undef SYMBOL_RECORD
 #undef SYMBOL_RECORD_ALIAS

llvm/trunk/include/llvm/DebugInfo/CodeView/SymbolRecord.h

public:
   TypeIndex Type;
   uint32_t DataOffset;
   uint16_t Segment;
   StringRef Name;
 
   uint32_t RecordOffset;
 };
 
+// S_UNAMESPACE
+class UsingNamespaceSym : public SymbolRecord {
+public:
+  explicit UsingNamespaceSym(SymbolRecordKind Kind) : SymbolRecord(Kind) {}
+  explicit UsingNamespaceSym(uint32_t RecordOffset)
+      : SymbolRecord(SymbolRecordKind::RegRelativeSym),
+        RecordOffset(RecordOffset) {}
+
+  StringRef Name;
+
+  uint32_t RecordOffset;
+};
+
 // S_ANNOTATION
 
 using CVSymbol = CVRecord<SymbolKind>;
 using CVSymbolArray = VarStreamArray<CVSymbol>;
 
 Expected<CVSymbol> readSymbolFromStream(BinaryStreamRef Stream,
                                         uint32_t Offset);
 
 } // end namespace codeview
 } // end namespace llvm
 
 #endif // LLVM_DEBUGINFO_CODEVIEW_SYMBOLRECORD_H

llvm/trunk/lib/DebugInfo/CodeView/RecordName.cpp

static int getSymbolNameOffset(CVSymbol Sym) {
   // See ObjNameSym, ExportSym, and UDTSym
   case SymbolKind::S_OBJNAME:
   case SymbolKind::S_EXPORT:
   case SymbolKind::S_UDT:
     return 4;
   // See BPRelativeSym
   case SymbolKind::S_BPREL32:
     return 8;
+  // See UsingNamespaceSym
+  case SymbolKind::S_UNAMESPACE:
+    return 0;
   default:
     return -1;
   }
 }
 
 StringRef llvm::codeview::getSymbolName(CVSymbol Sym) {
   if (Sym.kind() == SymbolKind::S_CONSTANT) {
     // S_CONSTANT is preceded by an APSInt, which has a variable length.  So we

llvm/trunk/lib/DebugInfo/CodeView/SymbolDumper.cpp

 }
 
 Error CVSymbolDumperImpl::visitKnownRecord(CVSymbol &CVR, UDTSym &UDT) {
   printTypeIndex("Type", UDT.Type);
   W.printString("UDTName", UDT.Name);
   return Error::success();
 }
 
+Error CVSymbolDumperImpl::visitKnownRecord(CVSymbol &CVR,
+                                           UsingNamespaceSym &UN) {
+  W.printString("Namespace", UN.Name);
+  return Error::success();
+}
+
 Error CVSymbolDumperImpl::visitUnknownSymbol(CVSymbol &CVR) {
   W.printNumber("Length", CVR.length());
   return Error::success();
 }
 
 Error CVSymbolDumper::dump(CVRecord<SymbolKind> &Record) {
   SymbolVisitorCallbackPipeline Pipeline;
   SymbolDeserializer Deserializer(ObjDelegate.get(), Container);

llvm/trunk/lib/DebugInfo/CodeView/SymbolRecordMapping.cpp

 
 Error SymbolRecordMapping::visitKnownRecord(CVSymbol &CVR, UDTSym &UDT) {
 
   error(IO.mapInteger(UDT.Type));
   error(IO.mapStringZ(UDT.Name));
 
   return Error::success();
 }
+
+Error SymbolRecordMapping::visitKnownRecord(CVSymbol &CVR,
+                                            UsingNamespaceSym &UN) {
+
+  error(IO.mapStringZ(UN.Name));
+
+  return Error::success();
+}

llvm/trunk/lib/DebugInfo/CodeView/TypeIndexDiscovery.cpp

static bool discoverTypeIndices(ArrayRef<uint8_t> Content, SymbolKind Kind,
   case SymbolKind::S_DEFRANGE_REGISTER:
   case SymbolKind::S_DEFRANGE_REGISTER_REL:
   case SymbolKind::S_DEFRANGE_FRAMEPOINTER_REL:
   case SymbolKind::S_DEFRANGE_FRAMEPOINTER_REL_FULL_SCOPE:
   case SymbolKind::S_DEFRANGE_SUBFIELD_REGISTER:
   case SymbolKind::S_DEFRANGE_SUBFIELD:
     break;
 
-  // No type refernces.
+  // No type references.
   case SymbolKind::S_LABEL32:
   case SymbolKind::S_OBJNAME:
   case SymbolKind::S_COMPILE:
   case SymbolKind::S_COMPILE2:
   case SymbolKind::S_COMPILE3:
   case SymbolKind::S_ENVBLOCK:
   case SymbolKind::S_BLOCK32:
   case SymbolKind::S_FRAMEPROC:
   case SymbolKind::S_THUNK32:
   case SymbolKind::S_FRAMECOOKIE:
+  case SymbolKind::S_UNAMESPACE:
     break;
   // Scope ending symbols.
   case SymbolKind::S_END:
   case SymbolKind::S_INLINESITE_END:
   case SymbolKind::S_PROC_ID_END:
     break;
   default:
     return false; // Unknown symbol.

llvm/trunk/lib/ObjectYAML/CodeViewYAMLSymbols.cpp

 
 template <> void SymbolRecordImpl<ThreadLocalDataSym>::map(IO &IO) {
   IO.mapRequired("Type", Symbol.Type);
   IO.mapOptional("Offset", Symbol.DataOffset, 0U);
   IO.mapOptional("Segment", Symbol.Segment, uint16_t(0));
   IO.mapRequired("DisplayName", Symbol.Name);
 }
 
+template <> void SymbolRecordImpl<UsingNamespaceSym>::map(IO &IO) {
+  IO.mapRequired("Namespace", Symbol.Name);
+}
+
 } // end namespace detail
 } // end namespace CodeViewYAML
 } // end namespace llvm
 
 CVSymbol CodeViewYAML::SymbolRecord::toCodeViewSymbol(
     BumpPtrAllocator &Allocator, CodeViewContainer Container) const {
   return Symbol->toCodeViewSymbol(Allocator, Container);
 }

llvm/trunk/test/DebugInfo/PDB/using-namespace.test

+# RUN: yaml2obj < %s > %t.obj
+# RUN: llvm-readobj -codeview %t.obj | FileCheck %s
+
+# CHECK:         Kind: S_UNAMESPACE (0x1124)
+# CHECK-NEXT:    Namespace: __vc_attributes
+
+--- !COFF
+header:          
+  Machine:         IMAGE_FILE_MACHINE_AMD64
+  Characteristics: [  ]
+sections:        
+  - Name:            '.debug$S'
+    Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_DISCARDABLE, IMAGE_SCN_MEM_READ ]
+    Alignment:       1
+    Subsections:     
+      - !Symbols
+        Records:         
+          - Kind:            S_OBJNAME
+            ObjNameSym:      
+              Signature:       0
+              ObjectName:      'SimpleFunction.obj'
+          - Kind:            S_COMPILE3
+            Compile3Sym:     
+              Flags:           [ SecurityChecks, HotPatch ]
+              Machine:         X64
+              FrontendMajor:   19
+              FrontendMinor:   14
+              FrontendBuild:   26433
+              FrontendQFE:     0
+              BackendMajor:    19
+              BackendMinor:    14
+              BackendBuild:    26433
+              BackendQFE:      0
+              Version:         'Microsoft (R) Optimizing Compiler'
+          - Kind:            S_UNAMESPACE
+            UsingNamespaceSym: 
+              Namespace:       __vc_attributes
+          - Kind:            S_UNAMESPACE
+            UsingNamespaceSym: 
+              Namespace:       helper_attributes
+          - Kind:            S_UNAMESPACE
+            UsingNamespaceSym: 
+              Namespace:       atl
+          - Kind:            S_UNAMESPACE
+            UsingNamespaceSym: 
+              Namespace:       std
+      - !StringTable
+        Strings:
+          - 'SimpleFunction.c'
+symbols:         
+...

llvm/trunk/tools/llvm-pdbutil/MinimalSymbolDumper.cpp

 }
 
 Error MinimalSymbolDumper::visitKnownRecord(CVSymbol &CVR, UDTSym &UDT) {
   P.format(" `{0}`", UDT.Name);
   AutoIndent Indent(P, 7);
   P.formatLine("original type = {0}", UDT.Type);
   return Error::success();
 }
+
+Error MinimalSymbolDumper::visitKnownRecord(CVSymbol &CVR,
+                                            UsingNamespaceSym &UN) {
+  P.format(" `{0}`", UN.Name);
+  return Error::success();
+}

llvm/trunk/unittests/DebugInfo/CodeView/TypeIndexDiscoveryTest.cpp

 }
 
 // This is a test for getEncodedIntegerLength()
 TEST_F(TypeIndexIteratorTest, VariableSizeIntegers) {
   BaseClassRecord BaseClass1(MemberAccess::Public, TypeIndex(47), (uint64_t)-1);
   BaseClassRecord BaseClass2(MemberAccess::Public, TypeIndex(48), 1);
   writeFieldList(BaseClass1, BaseClass2);
   checkTypeReferences(0, TypeIndex(47), TypeIndex(48));
-}
+}
+
+TEST_F(TypeIndexIteratorTest, UsingNamespace) {
+  UsingNamespaceSym UN(SymbolRecordKind::UsingNamespaceSym);
+  UN.Name = "std";
+  writeSymbolRecords(UN);
+  checkTypeReferences(0);
+}