[WebAssemlby] Set IsUsedInRegularObj correctly for undefined data symbols

[ruiu]

Is this correct? I thought that if no one is referencing an undefined symbol from an object file, that symbol shouldn't be marked as "used".

[sbc100]

Yes, its correct, I believe. We were already doing it for undefined functions, just not data symbols. In the ELF loader we do it for all symbols in the ::insert function. I think I will do a followup to do it the same way here.

The example I have the I think shows it to be correct:

1. Regular obj inserted undefined symbol.
2. bitcode obj defines the symbols and replaces it.
3. LTO happens, at which point symbol can be discarded if !IsUsedInRegularObj

So (1) needs to set IsUsedInRegularObj otherwise the symbol can be discarded in (3) and the result will be an undefined symbol.

[sbc100]

Followup is here: https://reviews.llvm.org/D49147

[sbc100]

ping

[sbc100]

Adding @jgravelle-google. Maybe @ruiu is OOO. This change I pretty small an non-controversial I think, not that I've explained it. These two lines should have been in the original LTO change.

test/wasm/lto/Inputs/used.ll

+target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
+target triple = "wasm32-unknown-unknown"
+
+@foo = hidden global i32 1
+
+define hidden void @bar() {
+  ret void
+}

test/wasm/lto/used.ll

+; RUN: llc %s -o %t.o -filetype=obj
+; RUN: llvm-as %S/Inputs/used.ll -o %t1.o
+; RUN: wasm-ld %t.o %t1.o -o %t.wasm
+; RUN: obj2yaml %t.wasm | FileCheck %s
+
+; Verify that symbols references from regular objects are preserved by LTO
+
+target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
+target triple = "wasm32-unknown-unknown"
+
+declare void @bar()
+
+@foo = external global i32
+
+define void @_start() {
+  %val = load i32, i32* @foo, align 4
+  %tobool = icmp ne i32 %val, 0
+  br i1 %tobool, label %callbar, label %return
+
+callbar:
+  call void @bar()
+  br label %return
+
+return:
+  ret void
+}
+
+; CHECK:        - Type:            DATA
+; CHECK-NEXT:     Segments:
+; CHECK-NEXT:       - SectionOffset:   7
+; CHECK-NEXT:         MemoryIndex:     0
+; CHECK-NEXT:         Offset:
+; CHECK-NEXT:           Opcode:          I32_CONST
+; CHECK-NEXT:           Value:           1024
+; CHECK-NEXT:         Content:         '01000000'
+
+; CHECK:       - Type:            CUSTOM
+; CHECK-NEXT:    Name:            name
+; CHECK-NEXT:    FunctionNames:   
+; CHECK-NEXT:      - Index:           0
+; CHECK-NEXT:        Name:            __wasm_call_ctors
+; CHECK-NEXT:      - Index:           1
+; CHECK-NEXT:        Name:            _start
+; CHECK-NEXT:      - Index:           2
+; CHECK-NEXT:        Name:            bar

wasm/SymbolTable.cpp

 Symbol *SymbolTable::addUndefinedData(StringRef Name, uint32_t Flags,
                                       InputFile *File) {
   LLVM_DEBUG(dbgs() << "addUndefinedData: " << Name << "\n");
 
   Symbol *S;
   bool WasInserted;
   std::tie(S, WasInserted) = insert(Name);
 
+  if (!File || File->kind() == InputFile::ObjectKind)
+    S->IsUsedInRegularObj = true;
+
   if (WasInserted)
     replaceSymbol<UndefinedData>(S, Name, Flags, File);
   else if (auto *Lazy = dyn_cast<LazySymbol>(S))
     Lazy->fetch();
   else if (S->isDefined())
     checkDataType(S, File);
   return S;
 }