LTO: Hash the set of imported symbols for each module.

[pcc]

Summary

This set may affect code generation and is sensitive to link order (and
possibly in the future to the linker's choice of prevailing symbol), so we
need to include it.

[tejohnson]

LGTM, thanks. Is a similar fix needed in the libLTO handling?

[pcc]

In D30586#691914, @tejohnson wrote:

LGTM, thanks. Is a similar fix needed in the libLTO handling?

It looks like we will need a similar fix here: http://llvm-cs.pcc.me.uk/lib/LTO/ThinLTOCodeGenerator.cpp#333

llvm/lib/LTO/LTO.cpp

#endif
   auto AddUnsigned = [&](unsigned I) {
     uint8_t Data[4];
     Data[0] = I;
     Data[1] = I >> 8;
     Data[2] = I >> 16;
     Data[3] = I >> 24;
     Hasher.update(ArrayRef<uint8_t>{Data, 4});
   };
+  auto AddUint64 = [&](uint64_t I) {
+    uint8_t Data[8];
+    Data[0] = I;
+    Data[1] = I >> 8;
+    Data[2] = I >> 16;
+    Data[3] = I >> 24;
+    Data[4] = I >> 32;
+    Data[5] = I >> 40;
+    Data[6] = I >> 48;
+    Data[7] = I >> 56;
+    Hasher.update(ArrayRef<uint8_t>{Data, 8});
+  };
   AddString(Conf.CPU);
   // FIXME: Hash more of Options. For now all clients initialize Options from
   // command-line flags (which is unsupported in production), but may set
   // RelaxELFRelocations. The clang driver can also pass FunctionSections,
   // DataSections and DebuggerTuning via command line flags.
   AddUnsigned(Conf.Options.RelaxELFRelocations);
   AddUnsigned(Conf.Options.FunctionSections);
   AddUnsigned(Conf.Options.DataSections);
#endif
 
   // Include the hash for the current module
   auto ModHash = Index.getModuleHash(ModuleID);
   Hasher.update(ArrayRef<uint8_t>((uint8_t *)&ModHash[0], sizeof(ModHash)));
   for (auto F : ExportList)
     // The export list can impact the internalization, be conservative here
     Hasher.update(ArrayRef<uint8_t>((uint8_t *)&F, sizeof(F)));
 
-  // Include the hash for every module we import functions from
+  // Include the hash for every module we import functions from. The set of
+  // imported symbols for each module may affect code generation and is
+  // sensitive to link order, so include that as well.
   for (auto &Entry : ImportList) {
     auto ModHash = Index.getModuleHash(Entry.first());
     Hasher.update(ArrayRef<uint8_t>((uint8_t *)&ModHash[0], sizeof(ModHash)));
+
+    AddUint64(Entry.second.size());
+    for (auto &Fn : Entry.second)
+      AddUint64(Fn.first);
   }
 
   // Include the hash for the resolved ODR.
   for (auto &Entry : ResolvedODR) {
     Hasher.update(ArrayRef<uint8_t>((const uint8_t *)&Entry.first,
                                     sizeof(GlobalValue::GUID)));
     Hasher.update(ArrayRef<uint8_t>((const uint8_t *)&Entry.second,
                                     sizeof(GlobalValue::LinkageTypes)));

llvm/test/ThinLTO/X86/Inputs/cache-import-lists1.ll

+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @f1() {
+  call void @linkonce_odr()
+  ret void
+}
+
+define linkonce_odr void @linkonce_odr() {
+  ret void
+}

llvm/test/ThinLTO/X86/Inputs/cache-import-lists2.ll

+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @f2() {
+  call void @linkonce_odr()
+  ret void
+}
+
+define linkonce_odr void @linkonce_odr() {
+  ret void
+}

llvm/test/ThinLTO/X86/cache-import-lists.ll

+; RUN: opt -module-hash -module-summary %s -o %t.bc
+; RUN: opt -module-hash -module-summary %S/Inputs/cache-import-lists1.ll -o %t1.bc
+; RUN: opt -module-hash -module-summary %S/Inputs/cache-import-lists2.ll -o %t2.bc
+
+; Tests that the hash for t is sensitive to the set of imported functions
+; for each module, which in this case depends on the link order (the function
+; linkonce_odr will be imported from either t1 or t2, whichever comes first).
+
+; RUN: rm -rf %t.cache
+; RUN: llvm-lto2 -cache-dir %t.cache -o %t.o %t.bc %t1.bc %t2.bc -r=%t.bc,main,plx -r=%t.bc,f1,lx -r=%t.bc,f2,lx -r=%t1.bc,f1,plx -r=%t1.bc,linkonce_odr,plx -r=%t2.bc,f2,plx -r=%t2.bc,linkonce_odr,lx
+; RUN: llvm-lto2 -cache-dir %t.cache -o %t.o %t.bc %t2.bc %t1.bc -r=%t.bc,main,plx -r=%t.bc,f1,lx -r=%t.bc,f2,lx -r=%t2.bc,f2,plx -r=%t2.bc,linkonce_odr,plx -r=%t1.bc,f1,plx -r=%t1.bc,linkonce_odr,lx
+; RUN: ls %t.cache | count 6
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @main() {
+  call void @f1()
+  call void @f2()
+  ret void
+}
+
+declare void @f1()
+declare void @f2()