[Mlir-commits] [mlir] [MLIR] Implement remark emitting policies in MLIR (PR #161202)

[Guray Ozen]

================
@@ -406,9 +429,15 @@ class RemarkEngine {
   /// main remark streamer.
   ~RemarkEngine();
 
+  /// Shutdown the remark engine. This will finalize the remark engine and
+  /// close the output file.
+  void shutdown();
----------------
grypp wrote:

When the emissions policy is final, we now defer emitting remarks until the very end.
In my previous PR, postponed remarks were emitted during MLIRContext destruction. That turned out to be problematic, since it required accessing the MLIRContext while it was being destroyed.

In this PR, I introduced a finalize() function that must be called explicitly before destroying the MLIRContext.

I’m happy to adjust if there’s a better approach.

https://github.com/llvm/llvm-project/pull/161202