[DebugInfo] Fix potential CU mismatch for SubprogramScopeDIEs.

[JDevlieghere]

Summary

In constructAbstractSubprogramScopeDIE there can be a potential mismatch
between this and the CU of ContextDIE when a scope is shared between
two DISubprograms belonging to a different CU. In that case, this is
the CU that was specified in the IR, but the CU of ContextDIE is that of
the first subprogram that was emitted. This patch fixes the mismatch by
looking up the CU of ContextDIE, and switching to use that.

This fixes PR35212 (https://bugs.llvm.org/show_bug.cgi?id=35212)

Patch by Philip Craig!

[alexcrichton]

Thanks so much for helping us with this @JDevlieghere!

[dblaikie]

test/DebugInfo/cross-cu-scope.ll

3 ↗

(On Diff #122729)

Include the original source code here to make it clear what's being modeled?

Is this reproducible in C or C++, or only in Rust?

[aprantl]

test/DebugInfo/cross-cu-scope.ll

2 ↗

(On Diff #122729)

Can you add a comment that explains what is being tested here?

[JDevlieghere]

• Add source and description to test case

[JDevlieghere]

• Further reduced testcase
• Use %llc_dwarf

test/DebugInfo/cross-cu-scope.ll

3 ↗

(On Diff #122729)

My gut feeling says it might occur with C++ and LTO, but I couldn't come up with the actual code to make it happen.

[aprantl]

test/DebugInfo/cross-cu-scope.ll

11 ↗

(On Diff #122858)

can you add a check for there name or something that would make it obvious which of the two CUs we expect here?

llvm/trunk/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp

void DwarfCompileUnit::constructAbstractSubprogramScopeDIE(
     LexicalScope *Scope) {
   DIE *&AbsDef = getAbstractSPDies()[Scope->getScopeNode()];
   if (AbsDef)
     return;
 
   auto *SP = cast<DISubprogram>(Scope->getScopeNode());
 
   DIE *ContextDIE;
+  DwarfCompileUnit *ContextCU = this;
 
   if (includeMinimalInlineScopes())
     ContextDIE = &getUnitDie();
   // Some of this is duplicated from DwarfUnit::getOrCreateSubprogramDIE, with
   // the important distinction that the debug node is not associated with the
   // DIE (since the debug node will be associated with the concrete DIE, if
   // any). It could be refactored to some common utility function.
   else if (auto *SPDecl = SP->getDeclaration()) {
     ContextDIE = &getUnitDie();
     getOrCreateSubprogramDIE(SPDecl);
-  } else
+  } else {
     ContextDIE = getOrCreateContextDIE(resolve(SP->getScope()));
+    // The scope may be shared with a subprogram that has already been
+    // constructed in another CU, in which case we need to construct this
+    // subprogram in the same CU.
+    ContextCU = DD->lookupCU(ContextDIE->getUnitDie());
+  }
 
   // Passing null as the associated node because the abstract definition
   // shouldn't be found by lookup.
-  AbsDef = &createAndAddDIE(dwarf::DW_TAG_subprogram, *ContextDIE, nullptr);
-  applySubprogramAttributesToDefinition(SP, *AbsDef);
+  AbsDef = &ContextCU->createAndAddDIE(dwarf::DW_TAG_subprogram, *ContextDIE, nullptr);
+  ContextCU->applySubprogramAttributesToDefinition(SP, *AbsDef);
 
-  if (!includeMinimalInlineScopes())
-    addUInt(*AbsDef, dwarf::DW_AT_inline, None, dwarf::DW_INL_inlined);
-  if (DIE *ObjectPointer = createAndAddScopeChildren(Scope, *AbsDef))
-    addDIEEntry(*AbsDef, dwarf::DW_AT_object_pointer, *ObjectPointer);
+  if (!ContextCU->includeMinimalInlineScopes())
+    ContextCU->addUInt(*AbsDef, dwarf::DW_AT_inline, None, dwarf::DW_INL_inlined);
+  if (DIE *ObjectPointer = ContextCU->createAndAddScopeChildren(Scope, *AbsDef))
+    ContextCU->addDIEEntry(*AbsDef, dwarf::DW_AT_object_pointer, *ObjectPointer);
 }
 
 DIE *DwarfCompileUnit::constructImportedEntityDIE(
     const DIImportedEntity *Module) {
   DIE *IMDie = DIE::get(DIEValueAllocator, (dwarf::Tag)Module->getTag());
   insertDIE(Module, IMDie);
   DIE *EntityDie;
   auto *Entity = resolve(Module->getEntity());

llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h

class DwarfDebug : public DebugHandlerBase {
 
   /// Store file names for type units under fission in a line table
   /// header that will be emitted into debug_line.dwo.
   // FIXME: replace this with a map from comp_dir to table so that we
   // can emit multiple tables during LTO each of which uses directory
   // 0, referencing the comp_dir of all the type units that use it.
   MCDwarfDwoLineTable SplitTypeUnitFileTable;
   /// @}
-  
   /// True iff there are multiple CUs in this module.
   bool SingleCU;
   bool IsDarwin;
 
   AddressPool AddrPool;
 
   DwarfAccelTable AccelNames;
   DwarfAccelTable AccelObjC;
public:
   void addAccelType(StringRef Name, const DIE &Die, char Flags);
 
   const MachineFunction *getCurrentFunction() const { return CurFn; }
 
   /// A helper function to check whether the DIE for a given Scope is
   /// going to be null.
   bool isLexicalScopeDIENull(LexicalScope *Scope);
 
+  /// Find the matching DwarfCompileUnit for the given CU DIE.
+  DwarfCompileUnit *lookupCU(const DIE *Die) { return CUDieMap.lookup(Die); }
+
   /// \defgroup DebuggerTuning Predicates to tune DWARF for a given debugger.
   ///
   /// Returns whether we are "tuning" for a given debugger.
   /// @{
   bool tuneForGDB() const { return DebuggerTuning == DebuggerKind::GDB; }
   bool tuneForLLDB() const { return DebuggerTuning == DebuggerKind::LLDB; }
   bool tuneForSCE() const { return DebuggerTuning == DebuggerKind::SCE; }
   /// @}
 };
 
 } // end namespace llvm
 
 #endif // LLVM_LIB_CODEGEN_ASMPRINTER_DWARFDEBUG_H

llvm/trunk/test/DebugInfo/cross-cu-scope.ll

+; RUN: %llc_dwarf %s -filetype=obj -o %t
+; RUN: llvm-dwarfdump -debug-info %t | FileCheck %s
+
+; Reduced test case from PR35212. Two DISubprogram belong to a different CU but
+; share a scope. Both are declarations and end up in the scope's CU. We want to
+; check that the CU from the context DIE is used (rather than from the IR).
+; This manifests itself by the DW_base_type ending up in the second CU, rather
+; than in the first one as specified in the IR.
+
+; CHECK: DW_TAG_compile_unit
+; CHECK-NEXT: discriminator 0
+; CHECK: DW_TAG_compile_unit
+; CHECK-NEXT: discriminator 1
+; CHECK: DW_TAG_structure_type
+; CHECK-NOT: NULL
+; CHECK: DW_TAG_subprogram
+; CHECK-NOT: NULL
+; CHECK: DW_TAG_formal_parameter
+; CHECK-NOT: NULL
+; CHECK: DW_AT_type{{.*}}"usize"
+; CHECK: NULL
+; CHECK: DW_TAG_base_type
+; CHECK-NOT: NULL
+; CHECK: DW_AT_name{{.*}}"usize"
+
+define hidden void @foo() !dbg !4 {
+  ret void, !dbg !7
+}
+
+!llvm.dbg.cu = !{!0, !2}
+!llvm.module.flags = !{!3}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_Rust, file: !1, producer: "clang LLVM (rustc version 1.23.0-nightly (discriminator 0))", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug)
+!1 = !DIFile(filename: "../lib.rs", directory: "/home/alex/code/rust4/lol")
+!2 = distinct !DICompileUnit(language: DW_LANG_Rust, file: !1, producer: "clang LLVM (rustc version 1.23.0-nightly (discriminator 1))", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug)
+!3 = !{i32 2, !"Debug Info Version", i32 3}
+!4 = distinct !DISubprogram(name: "clone<alloc::string::String>", linkageName: "_ZN5alloc3vec8{{impl}}28clone<alloc::string::String>E", scope: null, file: !1, line: 1519, type: !5, isLocal: false, isDefinition: true, scopeLine: 1519, flags: DIFlagPrototyped, isOptimized: true, unit: !0, templateParams: !6, variables: !6)
+!5 = !DISubroutineType(types: !6)
+!6 = !{}
+!7 = !DILocation(line: 1612, scope: !8, inlinedAt: !11)
+!8 = distinct !DILexicalBlock(scope: !9, file: !1, line: 86, column: 12)
+!9 = distinct !DISubprogram(name: "allocate_in<alloc::string::String,alloc::heap::Heap>", linkageName: "_ZN5alloc7raw_vec8{{impl}}52allocate_in<alloc::string::String,alloc::heap::Heap>E", scope: !10, file: !1, line: 82, type: !5, isLocal: false, isDefinition: true, scopeLine: 82, flags: DIFlagPrototyped, isOptimized: true, unit: !2, templateParams: !6, variables: !6)
+!10 = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "RawVec<alloc::string::String, alloc::heap::Heap>", file: !1, size: 128, align: 64, elements: !6, identifier: "5c6e4db16d2c64555e40661d70c4d81e")
+!11 = distinct !DILocation(line: 86, scope: !8, inlinedAt: !12)
+!12 = distinct !DILocation(line: 141, scope: !13, inlinedAt: !17)
+!13 = distinct !DISubprogram(name: "with_capacity<alloc::string::String>", linkageName: "_ZN5alloc7raw_vec8{{impl}}36with_capacity<alloc::string::String>E", scope: !10, file: !1, line: 140, type: !5, isLocal: false, isDefinition: true, scopeLine: 140, flags: DIFlagPrototyped, isOptimized: true, unit: !0, templateParams: !6, variables: !14)
+!14 = !{!15}
+!15 = !DILocalVariable(name: "cap", arg: 1, scope: !13, file: !1, line: 1, type: !16)
+!16 = !DIBasicType(name: "usize", size: 64, encoding: DW_ATE_unsigned)
+!17 = !DILocation(line: 1521, scope: !4)