[llvm-branch-commits] [llvm] release/22.x: [DTLTO] Fix handling of multi-module bitcode inputs (#174624) (PR #176066)

[via llvm-branch-commits]

https://github.com/llvmbot created https://github.com/llvm/llvm-project/pull/176066

Backport 663647f1b2befbb4eda52369438febf42e528921

Requested by: @bd1976bris

>From df78006a37c03ac927cc6707ef2fb906a258b8e1 Mon Sep 17 00:00:00 2001
From: Ben Dunbobbin <Ben.Dunbobbin at sony.com>
Date: Thu, 15 Jan 2026 00:21:01 +0000
Subject: [PATCH] [DTLTO] Fix handling of multi-module bitcode inputs (#174624)

This change fixes two issues when processing multi-module bitcode files
in DTLTO:

1. The DTLTO archive handling code incorrectly uses
getSingleBitcodeModule(), which asserts when the bitcode file contains
more than one module.
2. The temporary file containing the contents of an input archive member
was not emitted for multi-module bitcode files. This was due to
incorrect logic for recording whether a bitcode input contains any
ThinLTO modules. In a typical multi-module bitcode file, the first
module is a ThinLTO module while a subsequent auxiliary module is
non-ThinLTO. When modules are processed in order, the auxiliary module
causes the entire bitcode file to be classified as non-ThinLTO, and the
archive-member emission logic then incorrectly skips it.

In addition, this patch adds a test that verifies that multi-module
bitcode files can be successfully linked with DTLTO. The test reproduces
both issues as they existed prior to this change.

SIE Tracker: TOOLCHAIN-21008

(cherry picked from commit 663647f1b2befbb4eda52369438febf42e528921)
---
 cross-project-tests/dtlto/multimodule.test | 42 ++++++++++++++++++++++
 llvm/include/llvm/LTO/LTO.h                |  2 ++
 llvm/lib/DTLTO/DTLTO.cpp                   |  2 +-
 llvm/lib/LTO/LTO.cpp                       |  4 ++-
 4 files changed, 48 insertions(+), 2 deletions(-)
 create mode 100644 cross-project-tests/dtlto/multimodule.test

diff --git a/cross-project-tests/dtlto/multimodule.test b/cross-project-tests/dtlto/multimodule.test
new file mode 100644
index 0000000000000..b371a26ace399
--- /dev/null
+++ b/cross-project-tests/dtlto/multimodule.test
@@ -0,0 +1,42 @@
+REQUIRES: x86-registered-target,ld.lld,llvm-ar
+
+# Test that a DTLTO link succeeds with a multi-module (via -fsplit-lto-unit)
+# bitcode file. We use an archive, as archive member inputs exercise more of
+# the DTLTO specific code than other input file types.
+
+RUN: rm -rf %t && split-file %s %t && cd %t
+
+RUN: %clang -O2 --target=x86_64-linux-gnu -flto=thin -c usebar.cc \
+RUN:   -fno-rtti -fno-exceptions -fsplit-lto-unit 
+
+# Sanity check that multi-module bitcode was produced.
+RUN: not llvm-modextract -n 2 usebar.o -o - 2>&1 \
+RUN:   | FileCheck %s --check-prefix=TWO
+TWO: bitcode file contains 2 module(s)
+
+# Create an archive.
+RUN: llvm-ar rcs usebar.a usebar.o
+
+# Build with DTLTO.
+RUN: %clang -O2 --target=x86_64-linux-gnu -flto=thin -fuse-ld=lld \
+RUN:   -nostdlib -shared -Wl,--whole-archive,--allow-shlib-undefined usebar.a \
+RUN:   -fthinlto-distributor=%python \
+RUN:   -Xthinlto-distributor=%llvm_src_root/utils/dtlto/local.py \
+RUN:   -Wl,--save-temps
+
+RUN: ls | sort | FileCheck %s
+
+# DTLTO JSON file - confirms DTLTO occurred.
+CHECK: .dist-file.json
+
+# .native.o exists - confirms archive member usebar.o participated in DTLTO.
+CHECK: {{^}}usebar.a(usebar.o
+CHECK-SAME: .native.o
+
+#--- usebar.cc
+// Minimal C++ input to exercise multi-module emission with -fsplit-lto-unit.
+struct A { virtual int foo(); };
+int bar(A *a);
+
+struct B : A { int foo() { return 2; } };
+int use() { static B b; return bar(&b); }
diff --git a/llvm/include/llvm/LTO/LTO.h b/llvm/include/llvm/LTO/LTO.h
index 819be1909ec12..ea80a25ac4ed0 100644
--- a/llvm/include/llvm/LTO/LTO.h
+++ b/llvm/include/llvm/LTO/LTO.h
@@ -194,6 +194,8 @@ class InputFile {
 
   // Returns the only BitcodeModule from InputFile.
   LLVM_ABI BitcodeModule &getSingleBitcodeModule();
+  // Returns the primary BitcodeModule from InputFile.
+  LLVM_ABI BitcodeModule &getPrimaryBitcodeModule();
   // Returns the memory buffer reference for this input file.
   MemoryBufferRef getFileBuffer() const { return MbRef; }
   // Returns true if this input file is a member of an archive.
diff --git a/llvm/lib/DTLTO/DTLTO.cpp b/llvm/lib/DTLTO/DTLTO.cpp
index 7ba4bfd80b6ab..d41544277dd45 100644
--- a/llvm/lib/DTLTO/DTLTO.cpp
+++ b/llvm/lib/DTLTO/DTLTO.cpp
@@ -146,7 +146,7 @@ lto::DTLTO::addInput(std::unique_ptr<lto::InputFile> InputPtr) {
     return Input;
 
   SmallString<64> NewModuleId;
-  BitcodeModule &BM = Input->getSingleBitcodeModule();
+  BitcodeModule &BM = Input->getPrimaryBitcodeModule();
 
   // Check if the archive is a thin archive.
   Expected<bool> IsThin = isThinArchive(ArchivePath);
diff --git a/llvm/lib/LTO/LTO.cpp b/llvm/lib/LTO/LTO.cpp
index ff6762ebb59be..4d69ffeda1280 100644
--- a/llvm/lib/LTO/LTO.cpp
+++ b/llvm/lib/LTO/LTO.cpp
@@ -606,6 +606,8 @@ BitcodeModule &InputFile::getSingleBitcodeModule() {
   return Mods[0];
 }
 
+BitcodeModule &InputFile::getPrimaryBitcodeModule() { return Mods[0]; }
+
 LTO::RegularLTOState::RegularLTOState(unsigned ParallelCodeGenParallelismLevel,
                                       const Config &Conf)
     : ParallelCodeGenParallelismLevel(ParallelCodeGenParallelismLevel),
@@ -804,7 +806,7 @@ LTO::addModule(InputFile &Input, ArrayRef<SymbolResolution> InputRes,
   // If any of the modules inside of a input bitcode file was compiled with
   // ThinLTO, we assume that the whole input file also was compiled with
   // ThinLTO.
-  Input.IsThinLTO = IsThinLTO;
+  Input.IsThinLTO |= IsThinLTO;
 
   auto ModSyms = Input.module_symbols(ModI);
   addModuleToGlobalRes(ModSyms, Res,