[BUG FIX] Fix tensor_or_memref.h build error

[i-chaochen]

We found the build error on tensor_or_memref as the following, this PR has fixed it.

opt/bin/tensorflow/compiler/xla/mlir_hlo/_virtual_includes/mlir_interpreter_framework/tools/mlir_interpreter/framework/tensor_or_memref.h:243:29: error: 'isnan' was not declared in this scope; did you mean 'std::isnan'?
  243 |         bool thisnan = isnan(at(indices));
      |                        ~~~~~^~~~~~~~~~~~~
      |                        std::isnan
In file included from /usr/include/c++/9/complex:44,
                 from bazel-out/k8-opt/bin/tensorflow/compiler/xla/mlir_hlo/_virtual_includes/mlir_interpreter_framework/tools/mlir_interpreter/framework/interpreter_value.h:19,
                 from tensorflow/compiler/xla/mlir_hlo/tools/mlir_interpreter/framework/tests/interpreter_value_test.cc:16:
/usr/include/c++/9/cmath:632:5: note: 'std::isnan' declared here
  632 |     isnan(_Tp __x)
      |     ^~~~~


opt/bin/tensorflow/compiler/xla/mlir_hlo/_virtual_includes/mlir_interpreter_framework/tools/mlir_interpreter/framework/interpreter_value.h:108:31:   required from here
bazel-out/k8-opt/bin/tensorflow/compiler/xla/mlir_hlo/_virtual_includes/mlir_interpreter_framework/tools/mlir_interpreter/framework/tensor_or_memref.h:243:29: error: 'isnan' was not declared in this scope; did you mean 'std::isnan'?
  243 |         bool thisnan = isnan(at(indices));
      |                        ~~~~~^~~~~~~~~~~~~
      |                        std::isnan
In file included from /usr/include/c++/9/complex:44,
                 from bazel-out/k8-opt/bin/tensorflow/compiler/xla/mlir_hlo/_virtual_includes/mlir_interpreter_framework/tools/mlir_interpreter/framework/interpreter_value.h:19,
                 from tensorflow/compiler/xla/mlir_hlo/tools/mlir_interpreter/framework/tests/interpreter_value_test.cc:16:
/usr/include/c++/9/cmath:632:5: note: 'std::isnan' declared here
  632 |     isnan(_Tp __x)
      |     ^~~~~
In file included from bazel-out/k8-opt/bin/tensorflow/compiler/xla/mlir_hlo/_virtual_includes/mlir_interpreter_framework/tools/mlir_interpreter/framework/interpreter_value.h:34,
                 from tensorflow/compiler/xla/mlir_hlo/tools/mlir_interpreter/framework/tests/interpreter_value_test.cc:16:
bazel-out/k8-opt/bin/tensorflow/compiler/xla/mlir_hlo/_virtual_includes/mlir_interpreter_framework/tools/mlir_interpreter/framework/tensor_or_memref.h:244:30: error: 'isnan' was not declared in this scope, and no declarations were found by argument-dependent lookup at the point of instantiation [-fpermissive]
  244 |         bool othernan = isnan(other.at(indices));

[reedwm]

Unfortunately, I rolled back this PR in b3cc517 because it broke the Mac build with the error

error: no member named 'isnan' in namespace 'std'; did you mean simply 'isnan'?

According to this StackOverflow question, including math.h will put isnan in the global namespace, and optionally may also put it in the std namespace. Since only the Mac build was broken, presumably this means on Mac it was only put in the global namespace but on Windows/Linux, it was also put in the std namespace.

I think just including math.h without adding the std:: prefix to isnan will fix the issue. Can you create a new PR that just includes math.h?

[i-chaochen]

Unfortunately, I rolled back this PR in b3cc517 because it broke the Mac build with the error

error: no member named 'isnan' in namespace 'std'; did you mean simply 'isnan'?

According to this StackOverflow question, including math.h will put isnan in the global namespace, and optionally may also put it in the std namespace. Since only the Mac build was broken, presumably this means on Mac it was only put in the global namespace but on Windows/Linux, it was also put in the std namespace.

I think just including math.h without adding the std:: prefix to isnan will fix the issue. Can you create a new PR that just includes math.h?

Thanks a lot for this reference and explaination!

I have created a new PR based on the suggestion and it does resolve the build error on my side.

#59775