[LLD] [COFF] Handle comdat sections without leader symbols

[mstorsjo]

Summary

Discard them unless they have been associated by other means (yet unimplemented).

According to MS link.exe, such sections are illegal, but MinGW setups use them in their take on associative comdats.

This avoids leaving references to the bogus SectionChunk* PendingComdat, which cannot be dereferenced.

This fixes PR38183.

[mstorsjo]

Ping @rnk on this one as well; this is a dependency to D49700.

[rnk]

COFF/InputFiles.cpp

289 ↗

(On Diff #156715)

I think this is probably the right place to do the check.

[mstorsjo]

COFF/InputFiles.cpp

289 ↗

(On Diff #156715)

Yes, this is the one place where createRegular is called without checking for PendingComdat before. I thought the check felt natural there, but I can move it here instead.

[mstorsjo]

Moved the check, as suggested by @rnk.

[mstorsjo]

Regardless of the status of the other patch and for how to handle the associativity, this patch probably is useful, to avoid crashes if faced with such object files.

[rnk]

lgtm

lld/trunk/COFF/InputFiles.cpp

for (uint32_t I = 0; I < NumSymbols; ++I) {
     I += COFFSym.getNumberOfAuxSymbols();
   }
 
   for (uint32_t I : PendingIndexes) {
     COFFSymbolRef Sym = check(COFFObj->getSymbol(I));
     if (auto *Def = Sym.getSectionDefinition())
       if (Def->Selection == IMAGE_COMDAT_SELECT_ASSOCIATIVE)
         readAssociativeDefinition(Sym, Def);
+    if (SparseChunks[Sym.getSectionNumber()] == PendingComdat) {
+      StringRef Name;
+      COFFObj->getSymbolName(Sym, Name);
+      log("comdat section " + Name +
+          " without leader and unassociated, discarding");
+      continue;
+    }
     Symbols[I] = createRegular(Sym);
   }
 
   for (auto &KV : WeakAliases) {
     Symbol *Sym = KV.first;
     uint32_t Idx = KV.second;
     checkAndSetWeakAlias(Symtab, this, Sym, Symbols[Idx]);
   }

lld/trunk/test/COFF/pending-comdat.s

+# REQUIRES: x86
+
+# RUN: llvm-mc -triple=x86_64-windows-gnu %s -filetype=obj -o %t.obj
+
+# RUN: not lld-link -lldmingw -out:%t.exe -entry:main -subsystem:console %t.obj 2>&1 | FileCheck %s
+
+# CHECK: error: undefined symbol: other
+
+# Check that the comdat section without a symbol isn't left pending once we iterate symbols
+# to print source of the undefined symbol.
+
+  .text
+  .globl main
+main:
+  call other
+  ret
+
+  .section  .data$pending,"w"
+  .linkonce discard
+.Llocal:
+  .byte 0