[clang] [Clang] [C++26] Implement P2573R2: `= delete("should have a reason");` (PR #86526)

via cfe-commits cfe-commits at lists.llvm.org
Wed Apr 10 02:25:47 PDT 2024


================
@@ -1981,21 +1981,35 @@ class FunctionDecl : public DeclaratorDecl,
 
   };
 
-  /// Stashed information about a defaulted function definition whose body has
-  /// not yet been lazily generated.
-  class DefaultedFunctionInfo final
-      : llvm::TrailingObjects<DefaultedFunctionInfo, DeclAccessPair> {
+  /// Stashed information about a defaulted/deleted function body.
+  class DefaultedOrDeletedFunctionInfo final
+      : llvm::TrailingObjects<DefaultedOrDeletedFunctionInfo, DeclAccessPair,
+                              StringLiteral *> {
     friend TrailingObjects;
     unsigned NumLookups;
+    bool HasDeletedMessage;
----------------
Sirraide wrote:

> The alignment is likely to go from 32 to 64, isn't it?

It would normally, but from what I can tell, because it’s inheriting from `TrailingObjects`, the alignment is set to the maximum of the alignment of what the type would normally have and the alignment of each of the trailing objects (by an `alignas` on `TralingObjectsBase`), and since the maximum alignment of the latter is `64`, that ends up being the alignment of the overall struct as well and thus also its size.

Printing out `sizeof(DefaultedOrDeletedFunctionInfo)` shows `8` in both cases (w/ and w/o the bitfield). I also don’t have a strong opinion on this personally; I just didn’t want to make it a bitfield when there’s space available anyway because this way, it’s at least obvious that we have more space left if we need to add more flags later.



https://github.com/llvm/llvm-project/pull/86526


More information about the cfe-commits mailing list