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

Michael137 · 2025-10-09T10:37:44Z

Depends on:

[llvm][DebugInfo] Add 'sourceLanguageName' field support to DICompileUnit #162445

In preparation to emit DWARFv6's DW_AT_language_version.

… support to DICompileUnit

llvmbot · 2025-10-13T08:18:19Z

@llvm/pr-subscribers-llvm-ir

Author: Michael Buch (Michael137)

Changes

Depends on:

[llvm][DebugInfo] Add 'sourceLanguageName' field support to DICompileUnit #162445

In preparation to emit DWARFv6's DW_AT_language_version.

Full diff: https://github.com/llvm/llvm-project/pull/162632.diff

6 Files Affected:

(modified) llvm/lib/AsmParser/LLParser.cpp (+7-1)
(modified) llvm/lib/Bitcode/Reader/MetadataLoader.cpp (+4-2)
(modified) llvm/lib/Bitcode/Writer/BitcodeWriter.cpp (+1)
(modified) llvm/lib/IR/AsmWriter.cpp (+8-3)
(added) llvm/test/Assembler/dicompileunit-invalid-language-version.ll (+25)
(added) llvm/test/Bitcode/dwarf-source-language-version.ll (+17)

diff --git a/llvm/lib/AsmParser/LLParser.cpp b/llvm/lib/AsmParser/LLParser.cpp
index 380b19296a3c4..bb9f8442610d8 100644
--- a/llvm/lib/AsmParser/LLParser.cpp
+++ b/llvm/lib/AsmParser/LLParser.cpp
@@ -5865,6 +5865,7 @@ 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, );                                            \
@@ -5894,10 +5895,15 @@ 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,
diff --git a/llvm/lib/Bitcode/Reader/MetadataLoader.cpp b/llvm/lib/Bitcode/Reader/MetadataLoader.cpp
index cdcf7a80ffac7..ed0443f599a44 100644
--- a/llvm/lib/Bitcode/Reader/MetadataLoader.cpp
+++ b/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],
diff --git a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp b/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
index 54e916e2dcfe1..8ff3aa9817571 100644
--- a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
+++ b/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();
diff --git a/llvm/lib/IR/AsmWriter.cpp b/llvm/lib/IR/AsmWriter.cpp
index 2430d988ddb04..3908a78f48412 100644
--- a/llvm/lib/IR/AsmWriter.cpp
+++ b/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());
diff --git a/llvm/test/Assembler/dicompileunit-invalid-language-version.ll b/llvm/test/Assembler/dicompileunit-invalid-language-version.ll
new file mode 100644
index 0000000000000..b3794ac749f60
--- /dev/null
+++ b/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: ""))
diff --git a/llvm/test/Bitcode/dwarf-source-language-version.ll b/llvm/test/Bitcode/dwarf-source-language-version.ll
new file mode 100644
index 0000000000000..311afd5a99afe
--- /dev/null
+++ b/llvm/test/Bitcode/dwarf-source-language-version.ll
@@ -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")

adrian-prantl

Can you confirm that there is a .bc file checked in with RecordSize == 22 that tests the upgrade code?

Michael137 · 2025-10-14T09:29:50Z

Can you confirm that there is a .bc file checked in with RecordSize == 22 that tests the upgrade code?

There probably isn't. Do we have raw bitcode files that test the upgrade code for other metadata? Happy to add one if checking in bitcode is fine

adrian-prantl · 2025-10-14T15:35:23Z

ls llvm/test/Bitcode/*upgrade*

Michael137 · 2025-10-14T23:33:18Z

ls llvm/test/Bitcode/*upgrade*

Added. Turns out there were a couple already there that exercise this upgrade codepath, but I added a dedicated test just in case.

Michael137 · 2025-10-15T09:17:25Z

Failing tests in CI is a LLDB-DAP test. Unrelated

…163208) Depends on: * #163348 * #162632 With this patch Clang will start emitting `DW_AT_language_{name, version}` for C++/C/Objective-C/Objective-C++ when using `-gdwarf-6`. We adjust the `DISourceLanguageName` (which we pass to `DICompileUnit`) to hold a `DW_AT_language_name_` and version code when in DWARFv6. Otherwise we continue using the `DW_LANG_` version of `DISourceLanguageName`. We didn't back-port emitting `DW_AT_language_name`/`DW_AT_language_version` to DWARFv5 (unlike GCC, which emits both the new and old language attributes in DWARFv5) because there wasn't a compelling reason to do so (yet).

… DWARFv6 (#163208) Depends on: * llvm/llvm-project#163348 * llvm/llvm-project#162632 With this patch Clang will start emitting `DW_AT_language_{name, version}` for C++/C/Objective-C/Objective-C++ when using `-gdwarf-6`. We adjust the `DISourceLanguageName` (which we pass to `DICompileUnit`) to hold a `DW_AT_language_name_` and version code when in DWARFv6. Otherwise we continue using the `DW_LANG_` version of `DISourceLanguageName`. We didn't back-port emitting `DW_AT_language_name`/`DW_AT_language_version` to DWARFv5 (unlike GCC, which emits both the new and old language attributes in DWARFv5) because there wasn't a compelling reason to do so (yet).

…63147) Depends on: * #162632 Emit `DW_AT_language_version` (new in DWARFv6) if DICompileUnit has a `sourceLanguageVersion` field. Omit it if it has the default value of `0` (since it's the default lower bound of any language without a version scheme).

…version (#163147) Depends on: * llvm/llvm-project#162632 Emit `DW_AT_language_version` (new in DWARFv6) if DICompileUnit has a `sourceLanguageVersion` field. Omit it if it has the default value of `0` (since it's the default lower bound of any language without a version scheme).

Michael137 force-pushed the llvm/dwarf-language-attrs-2-part-9 branch 2 times, most recently from 2cf20c8 to dc6391f Compare October 11, 2025 12:39

[llvm][DebugInfo] [llvm][DebugInfo] Add 'sourceLanguageVersion' field…

2a2bd98

… support to DICompileUnit

Michael137 force-pushed the llvm/dwarf-language-attrs-2-part-9 branch from dc6391f to 2a2bd98 Compare October 11, 2025 12:44

Michael137 marked this pull request as ready for review October 13, 2025 08:17

llvmbot added the llvm:ir label Oct 13, 2025

This was referenced Oct 13, 2025

[llvm][DebugInfo] Add support for emitting DW_AT_language_version #163147

Merged

[clang][DebugInfo] Emit DW_AT_language_{name, version} for DWARFv6 #163208

Merged

adrian-prantl reviewed Oct 13, 2025

View reviewed changes

fixup! add dedicated bitcode upgrade test

831f6b9

Michael137 added 2 commits October 15, 2025 10:20

fixup! relax test assertion

8ecd4bc

fixup! fix test assertion

f55cca0

adrian-prantl approved these changes Oct 15, 2025

View reviewed changes

Michael137 merged commit cf1cdde into llvm:main Oct 15, 2025
10 checks passed

Michael137 deleted the llvm/dwarf-language-attrs-2-part-9 branch October 15, 2025 15:52

Michael137 changed the title ~~[llvm][DebugInfo] [llvm][DebugInfo] Add 'sourceLanguageVersion' field support to DICompileUnit~~ [llvm][DebugInfo] Add 'sourceLanguageVersion' field support to DICompileUnit Oct 15, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

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

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

Uh oh!

Michael137 commented Oct 9, 2025

Uh oh!

llvmbot commented Oct 13, 2025

Uh oh!

adrian-prantl left a comment

Uh oh!

Michael137 commented Oct 14, 2025

Uh oh!

adrian-prantl commented Oct 14, 2025

Uh oh!

Michael137 commented Oct 14, 2025

Uh oh!

Michael137 commented Oct 15, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

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

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

Uh oh!

Conversation

Michael137 commented Oct 9, 2025

Uh oh!

llvmbot commented Oct 13, 2025

Uh oh!

adrian-prantl left a comment

Choose a reason for hiding this comment

Uh oh!

Michael137 commented Oct 14, 2025

Uh oh!

adrian-prantl commented Oct 14, 2025

Uh oh!

Michael137 commented Oct 14, 2025

Uh oh!

Michael137 commented Oct 15, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants