[llvm] Emit DW_AT_identifier_case for compile units (PR #216197)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Aug 14 09:14:18 PDT 2026
https://github.com/timsmith78 updated https://github.com/llvm/llvm-project/pull/216197
>From 4ce144e92cac2b5daab8ae7a7f0e24f6c9d0bb4e Mon Sep 17 00:00:00 2001
From: Timothy Smith <timothy.smith at hpe.com>
Date: Thu, 13 Aug 2026 11:19:57 -0500
Subject: [PATCH 1/3] Emit DW_AT_identifier_case for compile units
DWARF compile unit tags can contain an attribute that indicates the case
sensitivity of symbols in the source language. This attribute will be
useful in supporting Fortran in lldb (issue# 205872)
Co-developed with Claude Opus 4.6
---
llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp | 32 +++++++++++++++
.../Generic/compileunit-identifier-case.ll | 41 +++++++++++++++++++
2 files changed, 73 insertions(+)
create mode 100644 llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
index 88e75e3c96106..566cc34476e26 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
@@ -1094,6 +1094,35 @@ void DwarfDebug::addGnuPubAttributes(DwarfCompileUnit &U, DIE &D) const {
U.addFlag(D, dwarf::DW_AT_GNU_pubnames);
}
+static bool isLangCaseSensitive(const DISourceLanguageName &Lang) {
+ if (!Lang.hasVersionedName()) {
+ switch (Lang.getName()) {
+ case dwarf::DW_LANG_Cobol74:
+ case dwarf::DW_LANG_Cobol85:
+ case dwarf::DW_LANG_Fortran77:
+ case dwarf::DW_LANG_Fortran90:
+ case dwarf::DW_LANG_Fortran95:
+ case dwarf::DW_LANG_Fortran03:
+ case dwarf::DW_LANG_Fortran08:
+ case dwarf::DW_LANG_Fortran18:
+ case dwarf::DW_LANG_Fortran23:
+ case dwarf::DW_LANG_Pascal83:
+ return false;
+ default:
+ return true;
+ }
+ } else {
+ switch (Lang.getName()) {
+ case dwarf::DW_LNAME_Fortran:
+ case dwarf::DW_LNAME_Cobol:
+ case dwarf::DW_LNAME_Pascal:
+ return false;
+ default:
+ return true;
+ }
+ }
+}
+
void DwarfDebug::finishUnitAttributes(const DICompileUnit *DIUnit,
DwarfCompileUnit &NewCU) {
DIE &Die = NewCU.getUnitDie();
@@ -1119,6 +1148,9 @@ void DwarfDebug::finishUnitAttributes(const DICompileUnit *DIUnit,
Lang.getName());
}
+ if (!isLangCaseSensitive(DIUnit->getSourceLanguage()))
+ NewCU.addUInt(Die, dwarf::DW_AT_identifier_case, dwarf::DW_FORM_data1,
+ dwarf::DW_ID_case_insensitive);
NewCU.addString(Die, dwarf::DW_AT_name, FN);
finishTargetUnitAttributes(*DIUnit, NewCU);
diff --git a/llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll b/llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll
new file mode 100644
index 0000000000000..754e1d759836f
--- /dev/null
+++ b/llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll
@@ -0,0 +1,41 @@
+; AIX and zOS don't support DWARF 6 DW_AT_language_name
+; XFAIL: target={{.*}}-zos{{.*}}, target={{.*}}-aix{{.*}}
+; RUN: %llc_dwarf -filetype=obj -O0 < %s | llvm-dwarfdump -debug-info - | FileCheck %s
+
+; Test that DW_AT_identifier_case is emitted for case-insensitive languages
+; (Fortran) and omitted for case-sensitive languages (C++).
+
+; Fortran via old-style language field
+; CHECK: DW_AT_language (DW_LANG_Fortran90)
+; CHECK-NEXT: DW_AT_identifier_case (DW_ID_case_insensitive)
+
+; C++ via old-style language field — no DW_AT_identifier_case emitted
+; CHECK: DW_AT_language (DW_LANG_C_plus_plus)
+; CHECK-NOT: DW_AT_identifier_case
+; CHECK: DW_AT_name
+
+; Fortran via sourceLanguageName
+; CHECK: DW_AT_language_name (DW_LNAME_Fortran)
+; CHECK: DW_AT_identifier_case (DW_ID_case_insensitive)
+
+; C++ via sourceLanguageName — no DW_AT_identifier_case emitted
+; CHECK: DW_AT_language_name (DW_LNAME_C_plus_plus)
+; CHECK-NOT: DW_AT_identifier_case
+; CHECK: DW_AT_name
+
+ at x = global i32 0, align 4, !dbg !0
+
+!llvm.dbg.cu = !{!2, !8, !9, !10}
+!llvm.module.flags = !{!6, !7}
+
+!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
+!1 = !DIGlobalVariable(name: "x", scope: !2, file: !3, line: 1, type: !5, isLocal: false, isDefinition: true)
+!2 = distinct !DICompileUnit(language: DW_LANG_Fortran90, file: !3, producer: "handwritten", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, globals: !4, splitDebugInlining: false, nameTableKind: Apple, sysroot: "/")
+!3 = !DIFile(filename: "test.f90", directory: "/tmp")
+!4 = !{!0}
+!5 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!6 = !{i32 7, !"Dwarf Version", i32 5}
+!7 = !{i32 2, !"Debug Info Version", i32 3}
+!8 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !3, producer: "handwritten", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, globals: !4, splitDebugInlining: false, nameTableKind: Apple, sysroot: "/")
+!9 = distinct !DICompileUnit(sourceLanguageName: DW_LNAME_Fortran, sourceLanguageVersion: 1990, file: !3, producer: "handwritten", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, globals: !4, splitDebugInlining: false, nameTableKind: Apple, sysroot: "/")
+!10 = distinct !DICompileUnit(sourceLanguageName: DW_LNAME_C_plus_plus, sourceLanguageVersion: 201100, file: !3, producer: "handwritten", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, globals: !4, splitDebugInlining: false, nameTableKind: Apple, sysroot: "/")
>From f7567f090cfb820b43c06576870e5e261bb61798 Mon Sep 17 00:00:00 2001
From: Timothy Smith <timothy.smith at hpe.com>
Date: Fri, 14 Aug 2026 11:13:15 -0500
Subject: [PATCH 2/3] Optimize language case-sensitivity check
---
llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp | 33 +++++++++++-----------
1 file changed, 16 insertions(+), 17 deletions(-)
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
index 566cc34476e26..502e510db18aa 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
@@ -1095,23 +1095,7 @@ void DwarfDebug::addGnuPubAttributes(DwarfCompileUnit &U, DIE &D) const {
}
static bool isLangCaseSensitive(const DISourceLanguageName &Lang) {
- if (!Lang.hasVersionedName()) {
- switch (Lang.getName()) {
- case dwarf::DW_LANG_Cobol74:
- case dwarf::DW_LANG_Cobol85:
- case dwarf::DW_LANG_Fortran77:
- case dwarf::DW_LANG_Fortran90:
- case dwarf::DW_LANG_Fortran95:
- case dwarf::DW_LANG_Fortran03:
- case dwarf::DW_LANG_Fortran08:
- case dwarf::DW_LANG_Fortran18:
- case dwarf::DW_LANG_Fortran23:
- case dwarf::DW_LANG_Pascal83:
- return false;
- default:
- return true;
- }
- } else {
+ if (Lang.hasVersionedName()) {
switch (Lang.getName()) {
case dwarf::DW_LNAME_Fortran:
case dwarf::DW_LNAME_Cobol:
@@ -1121,6 +1105,21 @@ static bool isLangCaseSensitive(const DISourceLanguageName &Lang) {
return true;
}
}
+ switch (Lang.getName()) {
+ case dwarf::DW_LANG_Cobol74:
+ case dwarf::DW_LANG_Cobol85:
+ case dwarf::DW_LANG_Fortran77:
+ case dwarf::DW_LANG_Fortran90:
+ case dwarf::DW_LANG_Fortran95:
+ case dwarf::DW_LANG_Fortran03:
+ case dwarf::DW_LANG_Fortran08:
+ case dwarf::DW_LANG_Fortran18:
+ case dwarf::DW_LANG_Fortran23:
+ case dwarf::DW_LANG_Pascal83:
+ return false;
+ default:
+ return true;
+ }
}
void DwarfDebug::finishUnitAttributes(const DICompileUnit *DIUnit,
>From 85a9c6b786d7fe931a84ffdd433f7212b71fa09e Mon Sep 17 00:00:00 2001
From: Timothy Smith <timothy.smith at hpe.com>
Date: Fri, 14 Aug 2026 11:13:49 -0500
Subject: [PATCH 3/3] Strengthen case-sensitivity test
---
llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll b/llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll
index 754e1d759836f..965832bea075a 100644
--- a/llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll
+++ b/llvm/test/DebugInfo/Generic/compileunit-identifier-case.ll
@@ -1,6 +1,7 @@
; AIX and zOS don't support DWARF 6 DW_AT_language_name
; XFAIL: target={{.*}}-zos{{.*}}, target={{.*}}-aix{{.*}}
-; RUN: %llc_dwarf -filetype=obj -O0 < %s | llvm-dwarfdump -debug-info - | FileCheck %s
+; RUN: %llc_dwarf -filetype=obj -O0 < %s | llvm-dwarfdump -debug-info - \
+; RUN: | FileCheck %s --implicit-check-not=DW_AT_identifier_case
; Test that DW_AT_identifier_case is emitted for case-insensitive languages
; (Fortran) and omitted for case-sensitive languages (C++).
@@ -11,8 +12,6 @@
; C++ via old-style language field — no DW_AT_identifier_case emitted
; CHECK: DW_AT_language (DW_LANG_C_plus_plus)
-; CHECK-NOT: DW_AT_identifier_case
-; CHECK: DW_AT_name
; Fortran via sourceLanguageName
; CHECK: DW_AT_language_name (DW_LNAME_Fortran)
@@ -20,8 +19,6 @@
; C++ via sourceLanguageName — no DW_AT_identifier_case emitted
; CHECK: DW_AT_language_name (DW_LNAME_C_plus_plus)
-; CHECK-NOT: DW_AT_identifier_case
-; CHECK: DW_AT_name
@x = global i32 0, align 4, !dbg !0
More information about the llvm-commits
mailing list