// RUN: tf-opt %s -tf-shape-inference=input-arg-shapes=1 -verify-diagnostics -split-input-file | FileCheck %s
// RUN: not tf-opt %s -tf-shape-inference=input-arg-shapes=* 2>&1 | FileCheck --check-prefix=INPUT_ARG_SHAPES_ERROR %s
// INPUT_ARG_SHAPES_ERROR: Missing input argument shapes

module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
    // CHECK-LABEL: func.func @main

    value {
      type: DT_INT32
    }
  }
  attr {
    key: "shape"
    value {
      shape {
        unknown_rank: true
      }
    }
  }
}
node {
  name: "x"
  op: "Placeholder"
  attr {
    key: "dtype"
    value {
      type: DT_VARIANT
    }
  }
  attr {
    key: "shape"
    value {
      shape {
        unknown_rank: true
      }
    }
  }
}
node {

// 7. Add necessary attributes to `node_def`, so we can replace it with a
//    XlaHostCompute node later. If all input shapes for XlaSendFromHost are
//    known, "shapes" attr will be set to the list of input shapes; otherwise
//    "shape_inference_graph" attr will be set to shape inference function name.
class RewriteOutsideCompilationSubgraphFn {
 public:
  RewriteOutsideCompilationSubgraphFn(

# RUN: tf-mlir-translate -graphdef-to-mlir -tf-enable-shape-inference-on-import=false %s -tf-input-arrays=input -tf-input-data-types=DT_FLOAT -tf-input-shapes='' -tf-output-arrays=out:1,out -o - | FileCheck %s

# Verify that we match correctly the input / output when they are scalar.

# CHECK-LABEL: func @main
# CHECK-SAME:  (%{{[a-z0-9]+}}: tensor<f32> {tf.device = "/device:CPU:0"})
# CHECK-SAME:  control_outputs = ""
# CHECK-SAME:  inputs = "input"

# RUN: tf-mlir-translate -graphdef-to-mlir -tf-enable-shape-inference-on-import=false %s -tf-input-arrays=input -tf-input-data-types=DT_FLOAT -tf-input-shapes='' -tf-output-arrays=input -tf-convert-legacy-fed-inputs -o - | FileCheck %s
# RUN: tf-mlir-translate -graphdef-to-mlir -tf-enable-shape-inference-on-import=false %s -tf-input-arrays=input -tf-input-shapes='' -tf-output-arrays=input -tf-convert-legacy-fed-inputs -o - | FileCheck --check-prefix=NODATATYPE %s

node {
  name: "Placeholder"
  op: "Placeholder"
  attr {
    key: "dtype"
    value {
      type: DT_FLOAT
    }
  }
  attr {
    key: "shape"
    value {
      shape {
        dim {
          size: 2
        }
        dim {

# CHECK: build_graph(out_dir)

# Checks the error message produced by tfcompile without mlir_component
# OLD: INVALID ARGUMENTS: Incompatible shapes: [2] vs. [3]
# OLD: x_y_sum

node: {
  name: "x"
  op: "Placeholder"
  attr: {
    key: "shape"
    value: {
      shape: {
        dim: {
          size: -1
        }
      }
    }
  }
  attr: {
    key: "dtype"
    value: {
      type: DT_INT32

Variable::Variable(ImmediateExecutionContext* ctx, DataType dtype,
                   TensorShape shape, absl::optional<std::string> name,
                   ImmediateTensorHandlePtr handle)
    : TensorHandleConvertible(std::move(handle)),
      name_(name.has_value() ? *name : "Variable"),
      dtype_(dtype),
      shape_(shape),
      ctx_(ctx) {}

Variable::~Variable() {

# RUN: not tf-mlir-translate -graphdef-to-mlir -tf-enable-shape-inference-on-import=false %s -tf-input-arrays=input -tf-input-data-types=DT_FLOAT -tf-input-shapes='' -tf-output-arrays=NotANodeInTheGraph -o - 2>&1 | FileCheck %s

# CHECK: Output NotANodeInTheGraph was not found in graph

node {
  name: "input"
  op: "Placeholder"
  device: "/device:CPU:0"
  attr {
    key: "dtype"
    value {
      type: DT_FLOAT
    }
  }
  attr {

// Check that attributes that define derived shapes are exported.

// CHECK: op: "PlaceholderWithDefault"
// CHECK: shape
// CHECK: unknown_rank: true
// CHECK: name: "static"
// CHECK: op: "PlaceholderWithDefault"
// CHECK: shape {
// CHECK-NEXT: }
// CHECK: name: "static_10"
// CHECK: op: "PlaceholderWithDefault"
// CHECK: shape
// CHECK: dim
// CHECK: size: 10

func.func @main() {