[clang] [Clang][attr] Add 'cfi_salt' attribute (PR #141846)

[Erich Keane via cfe-commits]

================
@@ -7941,6 +7942,32 @@ static bool handleFunctionTypeAttr(TypeProcessingState &state, ParsedAttr &attr,
     return true;
   }
 
+  if (attr.getKind() == ParsedAttr::AT_CFISalt) {
+    if (attr.getNumArgs() == 0)
+      return true;
+
+    // Delay if this is not a function type.
+    StringRef Argument;
+    if (!S.checkStringLiteralArgumentAttr(attr, 0, Argument))
+      return false;
+
+    // Delay if this is not a function type.
+    if (!unwrapped.isFunctionType())
+      return false;
+
+    const auto *FnTy = unwrapped.get()->getAs<FunctionProtoType>();
+    if (!FnTy)
----------------
erichkeane wrote:

Huh, interesting, I could have sworn we had a different location for no-prototype-attribute handling, but I don't think so anymore (after digging in for a few mins).

My bullets still are relevant.  I wouldn't expect an attribute that doesn't pass the builtin-TableGen'd stuff to reach this function (like for arg counts/appertainment/etc), so if this _IS_ getting here, it either needs to diagnose here, or figure out WHY we're continuing after failing tablegen.

https://github.com/llvm/llvm-project/pull/141846