[lld] [llvm] [dyndbg][LLD][ELF] Initial LLD support for dynamic debugging (PR #214188)

[Peter Smith via llvm-commits]

================
@@ -1914,6 +1918,41 @@ template <class ELFT> void Writer<ELFT>::finalizeSections() {
     // called after processSymbolAssignments() because it needs to know whether
     // a linker-script-defined symbol is absolute.
     scanRelocations<ELFT>(ctx);
+
+    // Process symbols referenced by the embedded unoptimized part of dynamic
+    // debugging.
+    if (ctx.hasDynDbg) {
+      bool ignoreUnresolved =
+          (ctx.arg.unresolvedSymbols == UnresolvedPolicy::Ignore);
+      bool warnOnly = (ctx.arg.unresolvedSymbols == UnresolvedPolicy::Warn);
+      for (Symbol *sym : ctx.symtab->getSymbols()) {
+        if (!sym->isDynDbgRef)
+          continue;
+
+        if (sym->isUndefined()) {
+          if (ignoreUnresolved || sym->isWeak())
+            continue;
+
+          static InputSection dummy(ctx.internalFile, dynDbgSecName, 0, 0, 0, 0,
----------------
smithp35 wrote:

For my understanding this presumably comes up when there's a reference from file without dynamic debugging, and a file with dynamic debugging. The symbol records the first so we don't have a dynDbgSec?

I couldn't spot any tests for this case. Would be good to add some in another test file?

https://github.com/llvm/llvm-project/pull/214188