[DWARF] Do not emit DWARF5 symbols in the case of DWARF2/3 + non-lldb

[linsinan1995]

Modify other legacy dwarf versions to align with the dwarf4 handling approach when determining whether to generate DWARF5 or GNU extensions.

[llvmbot]

@llvm/pr-subscribers-debuginfo

Author: sinan (linsinan1995)

Changes

Modify other legacy dwarf versions to align with the dwarf4 handling approach when determining whether to generate DWARF5 or GNU extensions.

---

Full diff: https://github.com/llvm/llvm-project/pull/110120.diff

2 Files Affected:

• (modified) llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp (+1-1)
• (modified) llvm/test/DebugInfo/MIR/X86/call-site-gnu-vs-dwarf5-attrs.mir (+15)

diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
index 0a1ff189bedbc4..ad7fdb12e9ccb6 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
@@ -1209,7 +1209,7 @@ void DwarfCompileUnit::constructAbstractSubprogramScopeDIE(
 }
 
 bool DwarfCompileUnit::useGNUAnalogForDwarf5Feature() const {
-  return DD->getDwarfVersion() == 4 && !DD->tuneForLLDB();
+  return DD->getDwarfVersion() <= 4 && !DD->tuneForLLDB();
 }
 
 dwarf::Tag DwarfCompileUnit::getDwarf5OrGNUTag(dwarf::Tag Tag) const {
diff --git a/llvm/test/DebugInfo/MIR/X86/call-site-gnu-vs-dwarf5-attrs.mir b/llvm/test/DebugInfo/MIR/X86/call-site-gnu-vs-dwarf5-attrs.mir
index 1790f761585c35..a9c20d774822ec 100644
--- a/llvm/test/DebugInfo/MIR/X86/call-site-gnu-vs-dwarf5-attrs.mir
+++ b/llvm/test/DebugInfo/MIR/X86/call-site-gnu-vs-dwarf5-attrs.mir
@@ -30,6 +30,21 @@
 # RUN:     -debug-entry-values -mtriple=x86_64-unknown-unknown \
 # RUN:     -start-after=machineverifier -o - %s | llvm-dwarfdump - | FileCheck %s -check-prefixes=CHECK-DWARF5
 
+## === DWARF3, tune for gdb ===
+# RUN: llc -emit-call-site-info -dwarf-version 3 -debugger-tune=gdb -filetype=obj \
+# RUN:     -mtriple=x86_64-unknown-unknown -start-after=machineverifier -o - %s  \
+# RUN:     | llvm-dwarfdump - | FileCheck %s -implicit-check-not=DW_AT_call
+
+## === DWARF3, tune for lldb ===
+# RUN: llc -dwarf-version 3 -debugger-tune=lldb -emit-call-site-info -filetype=obj \
+# RUN:     -mtriple=x86_64-unknown-unknown -start-after=machineverifier -o - %s   \
+# RUN:     | llvm-dwarfdump - | FileCheck %s -implicit-check-not=DW_AT_GNU_call
+
+## === DWARF3, tune for sce ===
+# RUN: llc -emit-call-site-info -dwarf-version 3 -filetype=obj -debugger-tune=sce \
+# RUN:     -debug-entry-values -mtriple=x86_64-unknown-unknown \
+# RUN:     -start-after=machineverifier -o - %s | llvm-dwarfdump - | FileCheck %s -implicit-check-not=DW_AT_call
+
 ## This is based on the following reproducer:
 ##
 ## extern void fn();

[pogo59]

LGTM. Probably should have been this way from the start.

[linsinan1995]

I encountered an issue where clang may emit DWARF version 5 symbols when a dwarf version less than 4 is specified. This can lead to errors during software packaging on some systems due to dwz (dwz: Unknown DWARF DW_OP_163, which is DW_OP_entry_value), generating problematic debug information.

A simple reproduce case in X86:

void use(int * x, int * y) {
  *x += 4;
  *y -= 2;
}

int helper(int z_, int d_);
int x = 0;
int y = 1;

int  main(int argc, char *argv[]) {
   x = argc;
   y = argc + 3;
   use(&x, &y);
   return helper(x, y);
}

$ clang++ -O1 -gdwarf-3 test.cc  -c
$ llvm-dwarfdump --show-form --verbose --debug-info test.o | grep DW_OP_entry_value
                     [0x0000000000000024, 0x0000000000000029) ".text": DW_OP_entry_value(DW_OP_reg5 RDI))
                     [0x000000000000001c, 0x0000000000000029) ".text": DW_OP_entry_value(DW_OP_reg4 RSI))

[djtodoro]

LGTM, thanks. Please make sure that the commit title is shorter (and still concise).