[llvm][DebugInfo] Add 'sourceLanguageVersion' field support to DICompileUnit (#162632)

Depends on:
* #162445

In preparation to emit DWARFv6's `DW_AT_language_version`.

Author: Michael137

llvm/lib/AsmParser/LLParser.cpp

@@ -5876,6 +5876,7 @@ bool LLParser::parseDICompileUnit(MDNode *&Result, bool IsDistinct) {
   REQUIRED(file, MDField, (/* AllowNull */ false));                            \
   OPTIONAL(language, DwarfLangField, );                                        \
   OPTIONAL(sourceLanguageName, DwarfSourceLangNameField, );                    \
+  OPTIONAL(sourceLanguageVersion, MDUnsignedField, (0, UINT32_MAX));           \
   OPTIONAL(producer, MDStringField, );                                         \
   OPTIONAL(isOptimized, MDBoolField, );                                        \
   OPTIONAL(flags, MDStringField, );                                            \
@@ -5905,10 +5906,15 @@ bool LLParser::parseDICompileUnit(MDNode *&Result, bool IsDistinct) {
     return error(Loc, "can only specify one of 'language' and "
                       "'sourceLanguageName' on !DICompileUnit");
 
+  if (sourceLanguageVersion.Seen && !sourceLanguageName.Seen)
+    return error(Loc, "'sourceLanguageVersion' requires an associated "
+                      "'sourceLanguageName' on !DICompileUnit");
+
   Result = DICompileUnit::getDistinct(
       Context,
       language.Seen ? DISourceLanguageName(language.Val)
-                    : DISourceLanguageName(sourceLanguageName.Val, 0),
+                    : DISourceLanguageName(sourceLanguageName.Val,
+                                           sourceLanguageVersion.Val),
       file.Val, producer.Val, isOptimized.Val, flags.Val, runtimeVersion.Val,
       splitDebugFilename.Val, emissionKind.Val, enums.Val, retainedTypes.Val,
       globals.Val, imports.Val, macros.Val, dwoId.Val, splitDebugInlining.Val,

llvm/lib/Bitcode/Reader/MetadataLoader.cpp

@@ -1860,7 +1860,7 @@ Error MetadataLoader::MetadataLoaderImpl::parseOneMetadata(
     break;
   }
   case bitc::METADATA_COMPILE_UNIT: {
-    if (Record.size() < 14 || Record.size() > 22)
+    if (Record.size() < 14 || Record.size() > 23)
       return error("Invalid record");
 
     // Ignore Record[0], which indicates whether this compile unit is
@@ -1869,11 +1869,13 @@ Error MetadataLoader::MetadataLoaderImpl::parseOneMetadata(
 
     const auto LangVersionMask = (uint64_t(1) << 63);
     const bool HasVersionedLanguage = Record[1] & LangVersionMask;
+    const uint32_t LanguageVersion = Record.size() > 22 ? Record[22] : 0;
 
     auto *CU = DICompileUnit::getDistinct(
         Context,
         HasVersionedLanguage
-            ? DISourceLanguageName(Record[1] & ~LangVersionMask, 0)
+            ? DISourceLanguageName(Record[1] & ~LangVersionMask,
+                                   LanguageVersion)
             : DISourceLanguageName(Record[1]),
         getMDOrNull(Record[2]), getMDString(Record[3]), Record[4],
         getMDString(Record[5]), Record[6], getMDString(Record[7]), Record[8],

llvm/lib/Bitcode/Writer/BitcodeWriter.cpp

@@ -2142,6 +2142,7 @@ void ModuleBitcodeWriter::writeDICompileUnit(const DICompileUnit *N,
   Record.push_back(N->getRangesBaseAddress());
   Record.push_back(VE.getMetadataOrNullID(N->getRawSysRoot()));
   Record.push_back(VE.getMetadataOrNullID(N->getRawSDK()));
+  Record.push_back(Lang.hasVersionedName() ? Lang.getVersion() : 0);
 
   Stream.EmitRecord(bitc::METADATA_COMPILE_UNIT, Record, Abbrev);
   Record.clear();

llvm/lib/IR/AsmWriter.cpp

@@ -2374,16 +2374,21 @@ static void writeDICompileUnit(raw_ostream &Out, const DICompileUnit *N,
   Out << "!DICompileUnit(";
   MDFieldPrinter Printer(Out, WriterCtx);
 
-  auto Lang = N->getSourceLanguage();
-  if (Lang.hasVersionedName())
+  DISourceLanguageName Lang = N->getSourceLanguage();
+
+  if (Lang.hasVersionedName()) {
     Printer.printDwarfEnum(
         "sourceLanguageName",
         static_cast<llvm::dwarf::SourceLanguageName>(Lang.getName()),
         dwarf::SourceLanguageNameString,
         /* ShouldSkipZero */ false);
-  else
+
+    Printer.printInt("sourceLanguageVersion", Lang.getVersion(),
+                     /*ShouldSkipZero=*/true);
+  } else {
     Printer.printDwarfEnum("language", Lang.getName(), dwarf::LanguageString,
                            /* ShouldSkipZero */ false);
+  }
 
   Printer.printMetadata("file", N->getRawFile(), /* ShouldSkipNull */ false);
   Printer.printString("producer", N->getProducer());

llvm/test/Assembler/dicompileunit-invalid-language-version.ll

@@ -0,0 +1,25 @@
+; RUN: split-file %s %t
+; RUN: not llvm-as < %t/dw_lang_with_version.ll -disable-output 2>&1 | FileCheck %s --check-prefix=WRONG-ATTR
+; RUN: not llvm-as < %t/overflow.ll -disable-output 2>&1 | FileCheck %s --check-prefix=OVERFLOW
+; RUN: not llvm-as < %t/version_without_name.ll -disable-output 2>&1 | FileCheck %s --check-prefix=NO-NAME
+; RUN: not llvm-as < %t/negative.ll -disable-output 2>&1 | FileCheck %s --check-prefix=NEGATIVE
+
+; WRONG-ATTR: error: 'sourceLanguageVersion' requires an associated 'sourceLanguageName' on !DICompileUnit
+; OVERFLOW: error: value for 'sourceLanguageVersion' too large, limit is 4294967295
+; NEGATIVE: error: expected unsigned integer
+; NO-NAME: error: missing one of 'language' or 'sourceLanguageName', required for !DICompileUnit
+
+;--- dw_lang_with_version.ll
+!0 = distinct !DICompileUnit(language: DW_LANG_C, sourceLanguageVersion: 1,
+                             file: !DIFile(filename: "", directory: ""))
+
+;--- overflow.ll
+!0 = distinct !DICompileUnit(sourceLanguageName: DW_LNAME_C, sourceLanguageVersion: 4294967298)
+
+;--- negative.ll
+!0 = distinct !DICompileUnit(sourceLanguageName: DW_LNAME_C, sourceLanguageVersion: -1,
+                             file: !DIFile(filename: "", directory: ""))
+
+;--- version_without_name.ll
+!0 = distinct !DICompileUnit(sourceLanguageVersion: 1,
+                             file: !DIFile(filename: "", directory: ""))

llvm/test/Bitcode/Inputs/compile-unit-no-versioned-language.bc

@@ -0,0 +1,17 @@
+; RUN: llvm-as < %s | llvm-dis | llvm-as | llvm-dis | FileCheck %s --implicit-check-not "sourceLanguageVersion: 0"
+
+; CHECK: sourceLanguageVersion: 120
+
+source_filename = "cu.cpp"
+target triple = "arm64-apple-macosx"
+
+!llvm.dbg.cu = !{!0, !5}
+!llvm.module.flags = !{!3, !4}
+
+!0 = distinct !DICompileUnit(sourceLanguageName: DW_LNAME_ObjC_plus_plus, sourceLanguageVersion: 120, file: !1, producer: "handwritten", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, globals: !2, splitDebugInlining: false, nameTableKind: Apple, sysroot: "/")
+!1 = !DIFile(filename: "cu.cpp", directory: "/tmp")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 5}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = distinct !DICompileUnit(sourceLanguageName: DW_LNAME_ObjC_plus_plus, sourceLanguageVersion: 0, file: !6, producer: "handwritten", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, globals: !2, splitDebugInlining: false, nameTableKind: Apple, sysroot: "/")
+!6 = !DIFile(filename: "cu2.cpp", directory: "/tmp")

llvm/test/Bitcode/dwarf-source-language-version.ll

@@ -0,0 +1,21 @@
+; Test loading metadata which was not aware of versioned language names.
+;
+; RUN: llvm-dis -o - %p/Inputs/compile-unit-no-versioned-language.bc \
+; RUN:     | FileCheck %s --implicit-check-not "sourceLanguageName" --implicit-check-not "sourceLanguageVersion"
+
+; Input bitcode file was compiled from following source on
+; LLVM commit `fc22b58c25963ece6b041cadbdc931c2338955e4`:
+;
+; source_filename = "cu.cpp"
+; target triple = "arm64-apple-macosx"
+; 
+; !llvm.dbg.cu = !{!0}
+; !llvm.module.flags = !{!3, !4}
+; 
+; !0 = distinct !DICompileUnit(language: DW_LANG_ObjC, file: !1, producer: "handwritten", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, globals: !2, splitDebugInlining: false, nameTableKind: Apple, sysroot: "/")
+; !1 = !DIFile(filename: "cu.cpp", directory: "/tmp")
+; !2 = !{}
+; !3 = !{i32 7, !"Dwarf Version", i32 5}
+; !4 = !{i32 2, !"Debug Info Version", i32 3}
+
+; CHECK: distinct !DICompileUnit(language: DW_LANG_ObjC,