[clang] [C++20] [Modules] Instantiate pending instantiations when GMF ends (PR #126842)

[Chuanqi Xu via cfe-commits]

================
@@ -1104,9 +1104,13 @@ void Sema::ActOnStartOfTranslationUnit() {
 }
 
 void Sema::ActOnEndOfTranslationUnitFragment(TUFragmentKind Kind) {
-  // No explicit actions are required at the end of the global module fragment.
-  if (Kind == TUFragmentKind::Global)
+  if (Kind == TUFragmentKind::Global) {
+    // Perform Pending Instantiations at the end of global module fragment so
+    // that the module ownership of TU-level decls won't get messed.
+    llvm::TimeTraceScope TimeScope("PerformPendingInstantiations");
+    PerformPendingInstantiations();
----------------
ChuanqiXu9 wrote:

Are you saying the lexical decl context of an instantiation should be the same with the template? If yes, I feel it is incorrect. I think the lexical decl context should show the lexical context of a decl literally.

---

BTW, I do think the modules ownership and decl context ownership are two distinct problems. And if you can solve two of them in a consistent manner, it is good. But it is good too to solve one standalone.

My attitude to such "fundamental" problems is: yes, it is always good to fix them fundamentally. But it is also good to do some adhoc mix to stop bleeding first. So the ordering in my mind can be:

landing this patch as a fix -> fix the fundamental problems you described -> revert this patch

I don't feel it will be problematic.

https://github.com/llvm/llvm-project/pull/126842