================
@@ -8641,22 +8641,33 @@ enum class CountedByInvalidPointeeTypeKind {
VALID,
};
-static bool CheckCountedByAttrOnField(
- Sema &S, FieldDecl *FD, Expr *E,
- llvm::SmallVectorImpl<TypeCoupledDeclRefInfo> &Decls) {
+static bool
+CheckCountedByAttrOnField(Sema &S, FieldDecl *FD, Expr *E,
+ llvm::SmallVectorImpl<TypeCoupledDeclRefInfo> &Decls,
+ bool CountInBytes, bool OrNull) {
// Check the context the attribute is used in
+ unsigned Kind = CountInBytes;
+ if (OrNull)
+ Kind += 2;
+
if (FD->getParent()->isUnion()) {
S.Diag(FD->getBeginLoc(), diag::err_counted_by_attr_in_union)
- << FD->getSourceRange();
+ << Kind << FD->getSourceRange();
return true;
}
const auto FieldTy = FD->getType();
+ if (FieldTy->isArrayType() && (CountInBytes || OrNull)) {
+ S.Diag(FD->getBeginLoc(),
+ diag::err_counted_by_attr_not_on_ptr_or_flexible_array_member)
----------------
delcypher wrote:
> > The OrNull case probably deserves its own special diagnostic because in
> > that case the diagnostic should explain that they cannot use the attribute
> > on arrays and that they need to use __counted_by instead.
>
> That is also true for `CountInBytes`. But I agree that it would be good for
> the diagnostic to suggest `counted_by`.
Hmm looking at it more closely for the
`err_counted_by_attr_not_on_ptr_or_flexible_array_member` means the diagnostic
output for
```
struct Test {
int count;
int fma[] __counted_by_or_null(count)
}
```
would look something like
```
counted_by_or_null only applies to pointers
```
i.e. mentioning flexible array members is dropped (I initially thought it
wasn't). So actually I guess this is fine. Suggesting the right attribute would
be an extra bonus. If you don't want to do it now file an issue (tagged with
clang:bounds-safety).
https://github.com/llvm/llvm-project/pull/93231
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
