[clang] cd64a42 - Reland "[CodeGen] Keep track info of lazy-emitted symbols in ModuleBuilder"

[Jun Zhang via cfe-commits]

Author: Jun Zhang
Date: 2022-06-18T20:27:21+08:00
New Revision: cd64a427efa0baaf1bb7ae624d4301908afc07f7

URL: https://github.com/llvm/llvm-project/commit/cd64a427efa0baaf1bb7ae624d4301908afc07f7
DIFF: https://github.com/llvm/llvm-project/commit/cd64a427efa0baaf1bb7ae624d4301908afc07f7.diff

LOG: Reland "[CodeGen] Keep track info of lazy-emitted symbols in ModuleBuilder"

This reverts commits:
d3ddc251acae631bf5ab4da13878f7e8b5b5a451
d90eecff5c9e7e9f8263de6cd72d70322400829f

It turned out there're some options turned on that leaks the memory
intentionally, which fires the asan builds after the patch being
applied. The issue has been fixed in
7bc00ce5cd41aad5fd0775f58c8e85a0a8d9ee56, so reland it.

Below is the original commit message:

The intent of this patch is to selectively carry some states over to
the Builder so we won't lose the information of the previous symbols.

This used to be several downstream patches of Cling, it aims to fix
errors in Clang Interpreter when trying to use inline functions.
Before this patch:

clang-repl> inline int foo() { return 42;}
clang-repl> int x = foo();

JIT session error: Symbols not found: [ _Z3foov ]
error: Failed to materialize symbols:
{ (main, { x, $.incr_module_1.__inits.0, __orc_init_func.incr_module_1 }) }

Co-authored-by: Axel Naumann <Axel.Naumann at cern.ch>
Signed-off-by: Jun Zhang <jun at junz.org>

Added: 
    

Modified: 
    clang/lib/CodeGen/CodeGenModule.h
    clang/lib/CodeGen/ModuleBuilder.cpp
    clang/test/Interpreter/execute.cpp

Removed: 
    


################################################################################
diff  --git a/clang/lib/CodeGen/CodeGenModule.h b/clang/lib/CodeGen/CodeGenModule.h
index f5ae83a2faca9..79e9a462a3d72 100644
--- a/clang/lib/CodeGen/CodeGenModule.h
+++ b/clang/lib/CodeGen/CodeGenModule.h
@@ -1486,6 +1486,33 @@ class CodeGenModule : public CodeGenTypeCache {
   void printPostfixForExternalizedDecl(llvm::raw_ostream &OS,
                                        const Decl *D) const;
 
+  /// Move some lazily-emitted states to the NewBuilder. This is especially
+  /// essential for the incremental parsing environment like Clang Interpreter,
+  /// because we'll lose all important information after each repl.
+  void moveLazyEmissionStates(CodeGenModule *NewBuilder) {
+    assert(DeferredDeclsToEmit.empty() &&
+           "Should have emitted all decls deferred to emit.");
+    assert(NewBuilder->DeferredDecls.empty() &&
+           "Newly created module should not have deferred decls");
+    NewBuilder->DeferredDecls = std::move(DeferredDecls);
+
+    assert(NewBuilder->DeferredVTables.empty() &&
+           "Newly created module should not have deferred vtables");
+    NewBuilder->DeferredVTables = std::move(DeferredVTables);
+
+    assert(NewBuilder->MangledDeclNames.empty() &&
+           "Newly created module should not have mangled decl names");
+    assert(NewBuilder->Manglings.empty() &&
+           "Newly created module should not have manglings");
+    NewBuilder->Manglings = std::move(Manglings);
+
+    assert(WeakRefReferences.empty() &&
+           "Not all WeakRefRefs have been applied");
+    NewBuilder->WeakRefReferences = std::move(WeakRefReferences);
+
+    NewBuilder->TBAA = std::move(TBAA);
+  }
+
 private:
   llvm::Constant *GetOrCreateLLVMFunction(
       StringRef MangledName, llvm::Type *Ty, GlobalDecl D, bool ForVTable,

diff  --git a/clang/lib/CodeGen/ModuleBuilder.cpp b/clang/lib/CodeGen/ModuleBuilder.cpp
index 50b7fd8eb993c..8e97a298ce7fa 100644
--- a/clang/lib/CodeGen/ModuleBuilder.cpp
+++ b/clang/lib/CodeGen/ModuleBuilder.cpp
@@ -134,7 +134,14 @@ namespace {
                               llvm::LLVMContext &C) {
       assert(!M && "Replacing existing Module?");
       M.reset(new llvm::Module(ExpandModuleName(ModuleName, CodeGenOpts), C));
+
+      std::unique_ptr<CodeGenModule> OldBuilder = std::move(Builder);
+
       Initialize(*Ctx);
+
+      if (OldBuilder)
+        OldBuilder->moveLazyEmissionStates(Builder.get());
+
       return M.get();
     }
 

diff  --git a/clang/test/Interpreter/execute.cpp b/clang/test/Interpreter/execute.cpp
index 298046c068c37..61e68990acf96 100644
--- a/clang/test/Interpreter/execute.cpp
+++ b/clang/test/Interpreter/execute.cpp
@@ -13,4 +13,8 @@ struct S { float f = 1.0; S *m = nullptr;} s;
 
 auto r2 = printf("S[f=%f, m=0x%llx]\n", s.f, reinterpret_cast<unsigned long long>(s.m));
 // CHECK-NEXT: S[f=1.000000, m=0x0]
+
+inline int foo() { return 42; }
+int r3 = foo();
+
 quit