[Mlir-commits] [mlir] [mlir] Convert TensorType and BaseMemRefType to interfaces (PR #133053)

[Matthias Springer]

================
@@ -41,7 +41,7 @@ TensorType inferReshapeInputType(TypedValue<TensorType> input,
   // 0D tensor. While such construct is not incorrect on its own, bufferization
   // cannot properly handle it at the moment, so we avoid it.
   SmallVector<int64_t> shape(input.getType().getRank(), 1);
-  return input.getType().clone(shape);
+  return mlir::cast<TensorType>(input.getType().clone(shape));
----------------
matthias-springer wrote:

I think that's acceptable. We have a similar issue with `Operation::clone`.

https://github.com/llvm/llvm-project/pull/133053