[Assembler] Allow assembling empty index with non-zero flags

[tejohnson]

What is the use case? Even with an empty module we at least get a module path entry. Hmm, we won't get a combined summary in that case either (with only entry being module path). Should this instead be changed to create the index if there is a module path entry (or in addition to checking flags)?

[evgeny777]

I was thinking of addressing 'skip-module-by-distributed-backend' case. For instance see writeEmptyDistributedBuildOutputs:

if (SkipModule) {
  ModuleSummaryIndex Index(/*HaveGVs*/ false);
  Index.setSkipModuleByDistributedBackend();
  WriteIndexToFile(Index, OS, nullptr);
}

[tejohnson]

LGTM. Ok makes sense, please update the comment to note that we check for non-empty flags too for that specific case.

llvm/test/Assembler/summary-flags2.ll

+; Check that we can assemble/disassemble empty index with non-trivial flags
+; RUN: llvm-as %s -o - | llvm-dis -o - | FileCheck %s
+
+; ModuleID = 'index.bc'
+source_filename = "index.bc"
+
+^0 = flags: 2
+; CHECK: ^0 = flags: 2

llvm/tools/llvm-as/llvm-as.cpp

if (EC) {
     errs() << EC.message() << '\n';
     exit(1);
   }
 
   if (Force || !CheckBitcodeOutputToConsole(Out->os(), true)) {
     const ModuleSummaryIndex *IndexToWrite = nullptr;
     // Don't attempt to write a summary index unless it contains any entries.
     // Otherwise we get an empty summary section.
-    if (Index && Index->begin() != Index->end())
+    if (Index && (Index->begin() != Index->end() || Index->getFlags()))
       IndexToWrite = Index;
     if (!IndexToWrite || (M && (!M->empty() || !M->global_empty())))
       // If we have a non-empty Module, then we write the Module plus
       // any non-null Index along with it as a per-module Index.
       // If both are empty, this will give an empty module block, which is
       // the expected behavior.
       WriteBitcodeToFile(*M, Out->os(), PreserveBitcodeUseListOrder,
                          IndexToWrite, EmitModuleHash);