[clang-repl] Teach clang-repl how to load PCHs.

[vgvassilev]

No description provided.

[llvmbot]

@llvm/pr-subscribers-clang

@llvm/pr-subscribers-clang-codegen

Author: Vassil Vassilev (vgvassilev)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/94166.diff

6 Files Affected:

• (modified) clang/include/clang/CodeGen/ModuleBuilder.h (+6)
• (modified) clang/lib/CodeGen/BackendConsumer.h (-5)
• (modified) clang/lib/CodeGen/CodeGenAction.cpp (+1-5)
• (modified) clang/lib/CodeGen/ModuleBuilder.cpp (+8)
• (modified) clang/lib/Interpreter/IncrementalParser.cpp (+4)
• (added) clang/test/Interpreter/execute-pch.cpp (+14)

diff --git a/clang/include/clang/CodeGen/ModuleBuilder.h b/clang/include/clang/CodeGen/ModuleBuilder.h
index edacd82bf899d..cb5919d1c8af5 100644
--- a/clang/include/clang/CodeGen/ModuleBuilder.h
+++ b/clang/include/clang/CodeGen/ModuleBuilder.h
@@ -48,6 +48,12 @@ namespace CodeGen {
 class CodeGenerator : public ASTConsumer {
   virtual void anchor();
 
+protected:
+  /// True if we've finished generating IR. This prevents us from generating
+  /// additional LLVM IR after emitting output in HandleTranslationUnit. This
+  /// can happen when Clang plugins trigger additional AST deserialization.
+  bool IRGenFinished = false;
+
 public:
   /// Return an opaque reference to the CodeGenModule object, which can
   /// be used in various secondary APIs.  It is valid as long as the
diff --git a/clang/lib/CodeGen/BackendConsumer.h b/clang/lib/CodeGen/BackendConsumer.h
index 0fe9929dca2b3..76ab5add603b7 100644
--- a/clang/lib/CodeGen/BackendConsumer.h
+++ b/clang/lib/CodeGen/BackendConsumer.h
@@ -41,11 +41,6 @@ class BackendConsumer : public ASTConsumer {
   llvm::Timer LLVMIRGeneration;
   unsigned LLVMIRGenerationRefCount;
 
-  /// True if we've finished generating IR. This prevents us from generating
-  /// additional LLVM IR after emitting output in HandleTranslationUnit. This
-  /// can happen when Clang plugins trigger additional AST deserialization.
-  bool IRGenFinished = false;
-
   bool TimerIsEnabled = false;
 
   std::unique_ptr<CodeGenerator> Gen;
diff --git a/clang/lib/CodeGen/CodeGenAction.cpp b/clang/lib/CodeGen/CodeGenAction.cpp
index 6d3efdb5ffe34..6e2204d2dba0c 100644
--- a/clang/lib/CodeGen/CodeGenAction.cpp
+++ b/clang/lib/CodeGen/CodeGenAction.cpp
@@ -221,9 +221,7 @@ void BackendConsumer::HandleInlineFunctionDefinition(FunctionDecl *D) {
 }
 
 void BackendConsumer::HandleInterestingDecl(DeclGroupRef D) {
-  // Ignore interesting decls from the AST reader after IRGen is finished.
-  if (!IRGenFinished)
-    HandleTopLevelDecl(D);
+  HandleTopLevelDecl(D);
 }
 
 // Links each entry in LinkModules into our module.  Returns true on error.
@@ -285,8 +283,6 @@ void BackendConsumer::HandleTranslationUnit(ASTContext &C) {
       if (LLVMIRGenerationRefCount == 0)
         LLVMIRGeneration.stopTimer();
     }
-
-    IRGenFinished = true;
   }
 
   // Silently ignore if we weren't initialized for some reason.
diff --git a/clang/lib/CodeGen/ModuleBuilder.cpp b/clang/lib/CodeGen/ModuleBuilder.cpp
index df85295cfb2e2..768d2ffd2d8d9 100644
--- a/clang/lib/CodeGen/ModuleBuilder.cpp
+++ b/clang/lib/CodeGen/ModuleBuilder.cpp
@@ -138,6 +138,8 @@ namespace {
       assert(!M && "Replacing existing Module?");
       M.reset(new llvm::Module(ExpandModuleName(ModuleName, CodeGenOpts), C));
 
+      IRGenFinished = false;
+
       std::unique_ptr<CodeGenModule> OldBuilder = std::move(Builder);
 
       Initialize(*Ctx);
@@ -179,6 +181,10 @@ namespace {
     }
 
     bool HandleTopLevelDecl(DeclGroupRef DG) override {
+      // Ignore interesting decls from the AST reader after IRGen is finished.
+      if (IRGenFinished)
+        return true; // We can't CodeGen more but pass to other consumers.
+
       // FIXME: Why not return false and abort parsing?
       if (Diags.hasUnrecoverableErrorOccurred())
         return true;
@@ -282,6 +288,8 @@ namespace {
     }
 
     void HandleTranslationUnit(ASTContext &Ctx) override {
+      IRGenFinished = true;
+
       // Release the Builder when there is no error.
       if (!Diags.hasUnrecoverableErrorOccurred() && Builder)
         Builder->Release();
diff --git a/clang/lib/Interpreter/IncrementalParser.cpp b/clang/lib/Interpreter/IncrementalParser.cpp
index ef90fe9e6f545..eddd4f356c5af 100644
--- a/clang/lib/Interpreter/IncrementalParser.cpp
+++ b/clang/lib/Interpreter/IncrementalParser.cpp
@@ -219,6 +219,10 @@ IncrementalParser::IncrementalParser(Interpreter &Interp,
   Consumer = &CI->getASTConsumer();
   P.reset(
       new Parser(CI->getPreprocessor(), CI->getSema(), /*SkipBodies=*/false));
+
+  if (ExternalASTSource *External = CI->getASTContext().getExternalSource())
+    External->StartTranslationUnit(Consumer);
+
   P->Initialize();
 
   // An initial PTU is needed as CUDA includes some headers automatically
diff --git a/clang/test/Interpreter/execute-pch.cpp b/clang/test/Interpreter/execute-pch.cpp
new file mode 100644
index 0000000000000..30390d02f8c5c
--- /dev/null
+++ b/clang/test/Interpreter/execute-pch.cpp
@@ -0,0 +1,14 @@
+// REQUIRES: host-supports-jit
+// UNSUPPORTED: system-aix
+
+// RUN: rm -f %t.pch
+// RUN: %clang_cc1 -fmax-type-align=16 -pic-level 2 -fdeprecated-macro -stack-protector 1 -fblocks -fskip-odr-check-in-gmf -fexceptions -fcxx-exceptions -fgnuc-version=0 -triple=%target_triple -DPCH -fincremental-extensions -emit-pch -x c++-header -o %t.pch %s
+// RUN: clang-repl -Xcc -fgnuc-version=0 -Xcc -triple=%target_triple -Xcc -include-pch -Xcc %t.pch '#include "%s"' | FileCheck %s
+
+#ifdef PCH
+int f_pch() { return 5; }
+#endif // PCH
+
+extern "C" int printf(const char *, ...);
+auto r1 = printf("f_pch = %d\n", f_pch());
+// CHECK: f_pch = 5

[vgvassilev]

@weliveindetail, it seems that clang-repl prefers the process triple:

llvm-project/clang/lib/Interpreter/Interpreter.cpp

Line 188 in c26a993

std::string TT = TargetTriple ? *TargetTriple : llvm::sys::getProcessTriple();

What should we do here? Should we implement a lit version of host-supports-jit but host-jit-triple? Is there something smarter we could do?

[lhames]

When -triple is passed explicitly I'd expect it to override the process triple. Is there a reason not to do that?

[vgvassilev]

How we can get the “right” triple from clang? And yes, sometimes the build can be configured in an odd way (we have some bots) where the triple from clang is different from the triple of the JIT. Unfortunately, we did not find a way to suppress these cases with a good lit clause last time we dealt with them.

[AaronBallman]

Why is IR gen finished when we're just starting the module?

[vgvassilev]

It's a double negation: it is /not/ finished when we start a new module.

[AaronBallman]

shouldn't this be set at the end of HandleTranslationUnit rather than the beginning?

[vgvassilev]

Good point. I guess I was avoiding this early return which I have now removed. In fact this change fixed some of the failures we see in the tests.