[not for review] preserve debug info in thinlto

[inglorion]

Summary

This is a possible way to get LTO to not drop debug information with
rL358069. It marks native object files generated by LTO as live if
they contain global symbols (which means these symbols are used by
other TUs or exported) so that even if MarkLive cannot see a reference
to the object, its debug information is preserved.

[inglorion]

I put this up for preservation purposes, but I'm not convinced that we should actually go this route, so abandoning.

lld/ELF/SymbolTable.cpp

template <class ELFT> void SymbolTable::addCombinedLTOObject() {
   LTO.reset(new BitcodeCompiler);
   for (BitcodeFile *F : BitcodeFiles)
     LTO->add(*F);
 
   for (InputFile *File : LTO->compile()) {
     DenseSet<CachedHashStringRef> DummyGroups;
     auto *Obj = cast<ObjFile<ELFT>>(File);
     Obj->parse(DummyGroups);
-    for (Symbol *Sym : Obj->getGlobalSymbols())
+    ArrayRef<Symbol *> Syms = Obj->getGlobalSymbols();
+    if (Syms.size() > 0) {
+      // Global symbols mean the object is live, but MarkLive may not
+      // see that because all uses may be inlined. Therefore, mark the
+      // object live here.
+      Obj->HasLiveCodeOrData = true;
+      for (Symbol *Sym : Syms)
         Sym->parseSymbolVersion();
+    }
     ObjectFiles.push_back(File);
   }
 }
 
 // Set a flag for --trace-symbol so that we can print out a log message
 // if a new symbol with the same name is inserted into the symbol table.
 void SymbolTable::trace(StringRef Name) {
   SymMap.insert({CachedHashStringRef(Name), -1});

lld/test/ELF/lto/Inputs/structfoo.ll

+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define dso_local i32 @_Z1fv() local_unnamed_addr #0 !dbg !8 {
+  %var.sroa.0 = alloca i32, align 4
+  %var.sroa.4 = alloca i32, align 4
+  call void @llvm.dbg.declare(metadata i32* %var.sroa.0, metadata !13, metadata !DIExpression(DW_OP_LLVM_fragment, 0, 32)), !dbg !19
+  call void @llvm.dbg.declare(metadata i32* %var.sroa.4, metadata !13, metadata !DIExpression(DW_OP_LLVM_fragment, 32, 32)), !dbg !19
+  store volatile i32 13, i32* %var.sroa.0, align 4, !dbg !21, !tbaa !22
+  store volatile i32 14, i32* %var.sroa.4, align 4, !dbg !27, !tbaa !28
+  %var.sroa.0.0.var.sroa.0.0.var.sroa.0.0. = load volatile i32, i32* %var.sroa.0, align 4, !dbg !29, !tbaa !22
+  %var.sroa.4.0.var.sroa.4.0.var.sroa.4.4. = load volatile i32, i32* %var.sroa.4, align 4, !dbg !30, !tbaa !28
+  %add = add nsw i32 %var.sroa.4.0.var.sroa.4.0.var.sroa.4.4., %var.sroa.0.0.var.sroa.0.0.var.sroa.0.0., !dbg !31
+  ret i32 %add, !dbg !33
+}
+
+declare void @llvm.dbg.declare(metadata, metadata, metadata) #0
+
+attributes #0 = { nounwind  }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5, !6}
+!llvm.ident = !{!7}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1, producer: "clang", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, nameTableKind: None)
+!1 = !DIFile(filename: "structfoo.cpp", directory: "/home/tests")
+!2 = !{}
+!3 = !{i32 2, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{i32 1, !"EnableSplitLTOUnit", i32 0}
+!7 = !{!"clang"}
+!8 = distinct !DISubprogram(name: "f", linkageName: "_Z1fv", scope: !1, file: !1, line: 5, type: !9, scopeLine: 5, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!9 = !DISubroutineType(types: !10)
+!10 = !{!11}
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13}
+!13 = !DILocalVariable(name: "var", scope: !8, file: !1, line: 6, type: !14)
+!14 = !DIDerivedType(tag: DW_TAG_volatile_type, baseType: !15)
+!15 = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "Foo", file: !1, line: 1, size: 64, flags: DIFlagTypePassByValue, elements: !16, identifier: "_ZTS3Foo")
+!16 = !{!17, !18}
+!17 = !DIDerivedType(tag: DW_TAG_member, name: "x", scope: !15, file: !1, line: 2, baseType: !11, size: 32)
+!18 = !DIDerivedType(tag: DW_TAG_member, name: "y", scope: !15, file: !1, line: 2, baseType: !11, size: 32, offset: 32)
+!19 = !DILocation(line: 6, column: 16, scope: !8)
+!20 = !DILocation(line: 6, column: 3, scope: !8)
+!21 = !DILocation(line: 7, column: 9, scope: !8)
+!22 = !{!23, !24, i64 0}
+!23 = !{!"_ZTS3Foo", !24, i64 0, !24, i64 4}
+!24 = !{!"int", !25, i64 0}
+!25 = !{!"omnipotent char", !26, i64 0}
+!26 = !{!"Simple C++ TBAA"}
+!27 = !DILocation(line: 8, column: 9, scope: !8)
+!28 = !{!23, !24, i64 4}
+!29 = !DILocation(line: 9, column: 14, scope: !8)
+!30 = !DILocation(line: 9, column: 22, scope: !8)
+!31 = !DILocation(line: 9, column: 16, scope: !8)
+!32 = !DILocation(line: 10, column: 1, scope: !8)
+!33 = !DILocation(line: 9, column: 3, scope: !8)
+
+^0 = module: (path: "structfoo.bc", hash: (1764250837, 803149122, 2095734431, 310569936, 1025282646))
+^1 = gv: (name: "_Z1fv", summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 1), insts: 14))) ; guid = 2964830440387334114
+^2 = gv: (name: "llvm.dbg.declare") ; guid = 13513223491971101989

lld/test/ELF/lto/keepdbg.ll

+; RUN: llvm-as -o %t.o %s
+; RUN: llvm-as -o %t.structfoo.o %S/Inputs/structfoo.ll
+; RUN: ld.lld -plugin-opt=thinlto -save-temps -gc-sections -print-gc-sections %t.o %t.structfoo.o -o %t
+; RUN: llvm-dwarfdump -debug-info %t | FileCheck %s
+
+; CHECK: DW_TAG_structure_type
+; CHECK-NOT: DW_TAG_
+; CHECK:                 DW_AT_name      ("Foo")
+; CHECK: DW_TAG_member
+; CHECK:                 DW_AT_name    ("x")
+; CHECK:                 DW_AT_type    ({{[^ ]+}} "int")
+; CHECK: DW_TAG_member
+; CHECK:                 DW_AT_name    ("y")
+; CHECK:                 DW_AT_type    ({{[^ ]+}} "int")
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define dso_local i32 @main() local_unnamed_addr #0 !dbg !8 {
+  %call = tail call i32 @_Z1fv(), !dbg !12
+  ret i32 %call, !dbg !13
+}
+
+declare dso_local i32 @_Z1fv() local_unnamed_addr #0
+
+attributes #0 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5, !6}
+!llvm.ident = !{!7}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1, producer: "clang", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, nameTableKind: None)
+!1 = !DIFile(filename: "keepdbg.cpp", directory: "/home/tests")
+!2 = !{}
+!3 = !{i32 2, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{i32 1, !"EnableSplitLTOUnit", i32 0}
+!7 = !{!"clang"}
+!8 = distinct !DISubprogram(name: "main", scope: !1, file: !1, line: 3, type: !9, scopeLine: 3, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!9 = !DISubroutineType(types: !10)
+!10 = !{!11}
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !DILocation(line: 4, column: 10, scope: !8)
+!13 = !DILocation(line: 4, column: 3, scope: !8)
+
+^0 = module: (path: "keepdbg.bc", hash: (1783178388, 1195490640, 2048955326, 1635682736, 2020218366))
+^1 = gv: (name: "_Z1fv") ; guid = 2964830440387334114
+^2 = gv: (name: "main", summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 0, live: 0, dsoLocal: 1), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 0), calls: ((callee: ^1))))) ; guid = 15822663052811949562