ELF: Set symbol binding to STB_GLOBAL when undefining symbols during LTO.

[pcc]

Summary

If there is a bug in the LTO implementation that causes it to fail to provide
an expected symbol definition, the linker should report an undefined symbol
error. Unfortunately, we were failing to do so if the symbol definition
was weak, as the undefine() function was turning the definition into a weak
undefined symbol, which resolves to zero if the symbol remains undefined. This
patch causes us to set the binding to STB_GLOBAL when we undefine a symbol.

I can't see a good way to test this. The behaviour should only be observable
if there is a bug in the LTO implementation.

[ruiu]

LGTM

lld/ELF/LTO.cpp

108 ↗

(On Diff #97387)

Please add a comment to describe why you are doing this. Without comment it is hard to understand why because the linker works fine without this piece of code.

[pcc]

Actually I don't think this change is necessary, and it was masking a bug (PR32899). We should be able to revert once that is fixed.

lld/trunk/ELF/LTO.cpp

 
 BitcodeCompiler::BitcodeCompiler() : LTOObj(createLTO()) {}
 
 BitcodeCompiler::~BitcodeCompiler() = default;
 
 static void undefine(Symbol *S) {
   replaceBody<Undefined>(S, S->body()->getName(), /*IsLocal=*/false,
                          STV_DEFAULT, S->body()->Type, nullptr);
+  // It shouldn't normally matter what the binding is, but if a bug in the LTO
+  // implementation causes it to fail to provide a definition for a symbol,
+  // setting the binding to STB_GLOBAL will cause the linker to report an
+  // undefined symbol error, even if the definition was weak.
+  S->Binding = STB_GLOBAL;
 }
 
 void BitcodeCompiler::add(BitcodeFile &F) {
   lto::InputFile &Obj = *F.Obj;
   unsigned SymNum = 0;
   std::vector<Symbol *> Syms = F.getSymbols();
   std::vector<lto::SymbolResolution> Resols(Syms.size());