add Wmissing-prototypes to clang-tidy (pytorch#96805)

   This PR introduces **-Wmissing-prototypes** of clang-tidy to prevent further coding errors such as the one fixed by PR pytorch#96714.

<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at fd2cf2a</samp>

This pull request makes several internal functions static to improve performance and avoid name clashes. It also fixes some typos, formatting, and missing includes in various files. It adds a new .clang-tidy check to warn about missing prototypes for non-static functions.

Pull Request resolved: pytorch#96805
Approved by: https://github.com/malfet, https://github.com/albanD

.clang-tidy

@@ -9,6 +9,7 @@ bugprone-*,
 -bugprone-lambda-function-name,
 -bugprone-reserved-identifier,
 -bugprone-swapped-arguments,
+clang-diagnostic-missing-prototypes,
 cppcoreguidelines-*,
 -cppcoreguidelines-avoid-do-while,
 -cppcoreguidelines-avoid-magic-numbers,

aten/src/ATen/DLConvertor.cpp

@@ -221,7 +221,7 @@ struct ATenDLMTensor {
   DLManagedTensor tensor;
 };
 
-void deleter(DLManagedTensor* arg) {
+static void deleter(DLManagedTensor* arg) {
   delete static_cast<ATenDLMTensor*>(arg->manager_ctx);
 }
 

c10/core/TensorImpl.cpp

@@ -1108,7 +1108,7 @@ void TensorImpl::ShareExternalPointer(
   }
 }
 
-void clone_symvec(SymIntArrayRef src, SymDimVector& dst) {
+static void clone_symvec(SymIntArrayRef src, SymDimVector& dst) {
   dst.clear();
   dst.reserve(src.size());
   for (const auto& i : src) {

c10/test/core/SymInt_test.cpp

@@ -5,7 +5,7 @@
 
 using namespace c10;
 #ifndef C10_MOBILE
-void check(int64_t value) {
+static void check(int64_t value) {
   const auto i = SymInt(value);
   EXPECT_EQ(i.maybe_as_int(), c10::make_optional(value));
 }

torch/csrc/Module.cpp

@@ -78,6 +78,7 @@
 #include <torch/csrc/utils/tensor_new.h>
 #include <torch/csrc/utils/tensor_numpy.h>
 #include <torch/csrc/utils/tensor_qschemes.h>
+#include <torch/csrc/utils/verbose.h>
 
 #ifdef USE_DISTRIBUTED
 #ifdef USE_C10D
@@ -1229,10 +1230,6 @@ void initIttBindings(PyObject* module);
 } // namespace torch
 #endif
 
-namespace torch {
-void initVerboseBindings(PyObject* module);
-} // namespace torch
-
 static std::vector<PyMethodDef> methods;
 
 // In Python we can't use the trick of C10_LOG_API_USAGE_ONCE

torch/csrc/autograd/FunctionsManual.cpp

@@ -59,7 +59,7 @@ Tensor apply_loss_reduction(const Tensor& unreduced, int64_t reduction) {
   return unreduced;
 }
 
-bool isDefined(const c10::optional<Tensor>& t) {
+static bool isDefined(const c10::optional<Tensor>& t) {
   return t.has_value() && t->defined();
 }
 
@@ -145,7 +145,7 @@ int64_t _safe_size(IntArrayRef sizes, IntArrayRef dim) {
   return size;
 }
 
-c10::SymInt _safe_size(c10::SymIntArrayRef sizes, c10::IntArrayRef dim) {
+static c10::SymInt _safe_size(c10::SymIntArrayRef sizes, c10::IntArrayRef dim) {
   c10::SymInt size = 1;
   if (sizes.empty()) {
     return 1;
@@ -165,7 +165,7 @@ Tensor handle_r_to_c(ScalarType self_st, Tensor gradient_result) {
   return gradient_result;
 }
 
-Tensor handle_r_to_c(Tensor self, Tensor gradient_result) {
+static Tensor handle_r_to_c(Tensor self, Tensor gradient_result) {
   if (!self.is_complex() && gradient_result.is_complex()) {
     // R -> C
     return at::real(gradient_result);
@@ -4365,7 +4365,7 @@ Tensor fft_r2c_backward(
 }
 
 // Helper for batchnorm_double_backward
-Tensor sum_exclude_dim1(const Tensor& to_sum, bool keepdim = true) {
+static Tensor sum_exclude_dim1(const Tensor& to_sum, bool keepdim = true) {
   auto r = to_sum.sum(0, keepdim);
   int64_t start_point_exclusive = keepdim ? 1 : 0;
   for (int64_t dim = r.dim() - 1; dim > start_point_exclusive; dim--) {
@@ -4377,7 +4377,7 @@ Tensor sum_exclude_dim1(const Tensor& to_sum, bool keepdim = true) {
 // Helper for batchnorm_double_backward
 // similar to expand_as below, but doesn't do the expand_as; operates as if
 // reductions were done with keepdim=True
-Tensor unsqueeze_dim1(const Tensor& src, const Tensor& target) {
+static Tensor unsqueeze_dim1(const Tensor& src, const Tensor& target) {
   auto src_expanded = src;
   while (src_expanded.sizes().size() < target.sizes().size() - 1) {
     src_expanded = src_expanded.unsqueeze(1);
@@ -4391,7 +4391,7 @@ Tensor unsqueeze_dim1(const Tensor& src, const Tensor& target) {
 // Helper for batchnorm_double_backward
 // because gamma/ggG/ggB are 1-dimensional and represent dim==1, we can't
 // do a straight expansion because it won't follow the broadcasting rules.
-Tensor expand_as_dim1(const Tensor& src, const Tensor& target) {
+static Tensor expand_as_dim1(const Tensor& src, const Tensor& target) {
   auto src_expanded = src;
   while (src_expanded.sizes().size() < target.sizes().size() - 1) {
     src_expanded = src_expanded.unsqueeze(1);
@@ -4947,7 +4947,7 @@ bool any_variable_defined(const variable_list& variables) {
 // from the right.
 // Additionally, when the computation is done in-place, we exploit that the
 // first `k` coordinates of `u_full/v_full` are zeros.
-Tensor apply_simple_transformation(
+static Tensor apply_simple_transformation(
     int64_t m,
     int64_t k,
     const Tensor& u_full,

torch/csrc/autograd/TraceTypeManual.cpp

@@ -158,7 +158,9 @@ TORCH_LIBRARY_IMPL(aten, Tracer, m) {
 
 namespace torch {
 namespace jit {
-void general_trace_function(const c10::OperatorHandle& op, Stack* stack) {
+static void general_trace_function(
+    const c10::OperatorHandle& op,
+    Stack* stack) {
   const auto input_size = op.schema().arguments().size();
   const auto output_size = op.schema().returns().size();
 

torch/csrc/autograd/autograd.cpp

@@ -24,7 +24,7 @@ namespace autograd {
 // maintain the logic equality of this file and the python file together if one
 // changes.
 // TODO: Make the Python API above to just call this C++ API.
-variable_list _make_grads(
+static variable_list _make_grads(
     const variable_list& outputs,
     const variable_list& grad_outputs) {
   size_t num_tensors = outputs.size();
@@ -88,7 +88,7 @@ variable_list _make_grads(
   }
   return new_grads;
 }
-variable_list run_backward(
+static variable_list run_backward(
     const variable_list& outputs,
     const variable_list& grad_outputs,
     bool keep_graph,

torch/csrc/autograd/autograd_not_implemented_fallback.cpp

@@ -47,7 +47,7 @@ void _foreach_tensor(
 
 } // namespace
 
-void autogradNotImplementedFallbackImpl(
+static void autogradNotImplementedFallbackImpl(
     const c10::OperatorHandle& op,
     c10::DispatchKeySet dispatch_keys,
     torch::jit::Stack* stack) {
@@ -252,7 +252,7 @@ torch::CppFunction autogradNotImplementedFallback() {
       &autogradNotImplementedFallbackImpl>();
 }
 
-void autogradNotImplementedInplaceOrViewFallbackImpl(
+static void autogradNotImplementedInplaceOrViewFallbackImpl(
     const c10::OperatorHandle& op,
     c10::DispatchKeySet dispatch_keys,
     torch::jit::Stack* stack) {

torch/csrc/autograd/custom_function.cpp

@@ -46,7 +46,7 @@ Variable VariableInfo::zeros(at::OptionalDeviceGuard& device_guard) const {
 //  sure that its
 //    forward grad was also modified inplace and already present on the
 //    corresponding output.
-void _process_forward_mode_AD(
+static void _process_forward_mode_AD(
     const variable_list& inputs,
     std::unordered_map<at::TensorImpl*, size_t> inputs_mapping,
     const at::ArrayRef<c10::optional<Variable>> raw_outputs,
@@ -247,7 +247,7 @@ void _process_forward_mode_AD(
   }
 }
 
-at::Tensor _view_as_self_with_no_grad(at::Tensor self) {
+static at::Tensor _view_as_self_with_no_grad(at::Tensor self) {
   // This is called below in _process_backward_mode_ad in two places:
   //
   // (1) An input has been returned, but it wasn't modified. Return it as a view
@@ -268,7 +268,7 @@ at::Tensor _view_as_self_with_no_grad(at::Tensor self) {
   return self.view_as(self);
 }
 
-optional_variable_list _process_backward_mode_ad(
+static optional_variable_list _process_backward_mode_ad(
     const std::unordered_map<at::TensorImpl*, size_t>& inputs_mapping,
     const std::unordered_set<at::TensorImpl*>& non_differentiable,
     const std::unordered_set<at::TensorImpl*>& dirty_inputs,

torch/csrc/autograd/record_function_ops.cpp

@@ -18,7 +18,7 @@ namespace profiler {
 
 // Creates a new profiling scope using RecordFunction and invokes its starting
 // callbacks.
-void record_function_enter(
+static void record_function_enter(
     const std::string& name,
     const c10::optional<std::string>& args,
     at::RecordFunction& rec) {
@@ -33,7 +33,7 @@ void record_function_enter(
 }
 
 // Legacy signature using cpp_custom_type_hack
-at::Tensor record_function_enter_legacy(
+static at::Tensor record_function_enter_legacy(
     const std::string& name,
     const c10::optional<std::string>& args) {
   auto rec = std::make_unique<at::RecordFunction>(at::RecordScope::USER_SCOPE);
@@ -51,26 +51,27 @@ c10::intrusive_ptr<PythonRecordFunction> record_function_enter_new(
   return rec;
 }
 
-at::RecordFunction& getRecordFunctionFromTensor(const at::Tensor& handle) {
+static at::RecordFunction& getRecordFunctionFromTensor(
+    const at::Tensor& handle) {
   auto& rec = at::cpp_custom_type_hack::cast<at::RecordFunction>(handle);
   return rec;
 }
 
 // Ends the profiling scope created with record_function_enter.
-void record_function_exit(at::RecordFunction& rec) {
+static void record_function_exit(at::RecordFunction& rec) {
   rec.end();
 }
 
 // Legacy signature using cpp_custom_type_hack
-void record_function_exit_legacy(const at::Tensor& handle) {
+static void record_function_exit_legacy(const at::Tensor& handle) {
   // We don't actually need to do anything with handle just need to persist the
   // lifetime until now.
   auto& rec = getRecordFunctionFromTensor(handle);
   record_function_exit(rec);
 }
 
 // New signature using custom_class
-void record_function_exit_new(
+static void record_function_exit_new(
     const c10::intrusive_ptr<PythonRecordFunction>& record) {
   record_function_exit(record->record);
 }
@@ -100,7 +101,7 @@ c10::intrusive_ptr<c10::ivalue::Future> _call_end_callbacks_on_fut(
 }
 
 // Legacy signature using cpp_custom_type_hack
-c10::intrusive_ptr<c10::ivalue::Future> _call_end_callbacks_on_fut_legacy(
+static c10::intrusive_ptr<c10::ivalue::Future> _call_end_callbacks_on_fut_legacy(
     const at::Tensor& handle,
     const c10::intrusive_ptr<c10::ivalue::Future>& fut) {
   return _call_end_callbacks_on_fut(

torch/csrc/autograd/variable.cpp

@@ -157,7 +157,7 @@ AutogradMeta* materialize_autograd_meta(const at::TensorBase& self) {
   return get_autograd_meta(self);
 }
 
-void update_tensor_hooks_on_new_gradfn(
+static void update_tensor_hooks_on_new_gradfn(
     const at::TensorBase& self,
     const std::shared_ptr<torch::autograd::Node>& old_fn,
     const std::shared_ptr<torch::autograd::Node>& new_fn) {

torch/csrc/distributed/autograd/utils.cpp

@@ -76,7 +76,7 @@ ContextPtr addRecvRpcBackward(
   return autogradContext;
 }
 
-c10::intrusive_ptr<Message> getMessageWithProfiling(
+static c10::intrusive_ptr<Message> getMessageWithProfiling(
     c10::intrusive_ptr<torch::distributed::rpc::Message> wrappedRpcMessage,
     MessageType msgType,
     torch::autograd::profiler::ProfilerConfig&& profilerConfig) {

torch/csrc/jit/frontend/error_report.cpp

@@ -46,7 +46,7 @@ ErrorReport::CallStack::CallStack(
 ErrorReport::CallStack::~CallStack() {}
 #endif // C10_MOBILE
 
-std::string get_stacked_errors(const std::vector<Call>& error_stack) {
+static std::string get_stacked_errors(const std::vector<Call>& error_stack) {
   std::stringstream msg;
   if (!error_stack.empty()) {
     for (auto it = error_stack.rbegin(); it != error_stack.rend() - 1; ++it) {

torch/csrc/jit/passes/specialize_autogradzero.cpp

@@ -15,13 +15,13 @@ namespace jit {
 
 static const auto countsAttribute = Symbol::attr("none_counts");
 
-bool hasGradSumToSizeUses(Value* v) {
+static bool hasGradSumToSizeUses(Value* v) {
   return std::any_of(v->uses().begin(), v->uses().end(), [](const Use& use) {
     return use.user->kind() == aten::_grad_sum_to_size;
   });
 }
 
-void insertProfileNodesForSpecializeAutogradZero(
+static void insertProfileNodesForSpecializeAutogradZero(
     Block* block,
     ProfilingRecord* pr) {
   for (auto it = block->nodes().begin(); it != block->nodes().end(); ++it) {

torch/csrc/jit/runtime/profiling_record.cpp

@@ -206,7 +206,7 @@ void ProfilingRecord::insertShapeProfile(
   n->replaceInput(offset, pn->output());
 }
 
-bool needsProfiledInputs(Node* n) {
+static bool needsProfiledInputs(Node* n) {
   if (tensorexpr::isSupported(n) ||
 #ifndef C10_MOBILE
       (fuser::cuda::isEnabled() && fuser::cuda::profileNode(n))
@@ -243,7 +243,7 @@ bool needsProfiledInputs(Node* n) {
   }
 }
 
-bool needsProfiledOutput(Node* n) {
+static bool needsProfiledOutput(Node* n) {
   if (tensorexpr::isSupported(n) ||
 #ifndef C10_MOBILE
       (fuser::cuda::isEnabled() && fuser::cuda::profileNode(n))

torch/csrc/tensor/python_tensor.cpp

@@ -113,31 +113,31 @@ static PyObject* Tensor_instancecheck(PyObject* _self, PyObject* arg) {
   END_HANDLE_TH_ERRORS
 }
 
-PyObject* Tensor_dtype(PyTensorType* self, void* unused) {
+static PyObject* Tensor_dtype(PyTensorType* self, void* unused) {
   return torch::autograd::utils::wrap(self->dtype);
 }
 
-PyObject* Tensor_layout(PyTensorType* self, void* unused) {
+static PyObject* Tensor_layout(PyTensorType* self, void* unused) {
   return torch::autograd::utils::wrap(self->layout);
 }
 
-PyObject* Tensor_is_cuda(PyTensorType* self, void* unused) {
+static PyObject* Tensor_is_cuda(PyTensorType* self, void* unused) {
   if (self->is_cuda) {
     Py_RETURN_TRUE;
   } else {
     Py_RETURN_FALSE;
   }
 }
 
-PyObject* Tensor_is_sparse(PyTensorType* self, void* unused) {
+static PyObject* Tensor_is_sparse(PyTensorType* self, void* unused) {
   if (self->layout->layout == at::Layout::Strided) {
     Py_RETURN_FALSE;
   } else {
     Py_RETURN_TRUE;
   }
 }
 
-PyObject* Tensor_is_sparse_csr(PyTensorType* self, void* unused) {
+static PyObject* Tensor_is_sparse_csr(PyTensorType* self, void* unused) {
   if (self->layout->layout == at::Layout::SparseCsr) {
     Py_RETURN_TRUE;
   } else {
@@ -302,7 +302,7 @@ static THPObjectPtr get_tensor_dict() {
 // importing torch.
 static std::vector<PyTensorType*> tensor_types;
 
-void set_default_storage_type(Backend backend, ScalarType dtype) {
+static void set_default_storage_type(Backend backend, ScalarType dtype) {
   THPObjectPtr storage = get_storage_obj(backend, dtype);
 
   auto torch_module = THPObjectPtr(PyImport_ImportModule("torch"));
@@ -314,7 +314,7 @@ void set_default_storage_type(Backend backend, ScalarType dtype) {
   }
 }
 
-void set_default_tensor_type(
+static void set_default_tensor_type(
     c10::optional<Backend> backend,
     c10::optional<ScalarType> dtype) {
   if (backend.has_value()) {

torch/csrc/utils/byte_order.cpp

@@ -261,14 +261,15 @@ void THP_decodeComplexDoubleBuffer(
   for (const auto i : c10::irange(len)) {
     // NOLINTNEXTLINE(cppcoreguidelines-pro-type-member-init)
     union {
-      uint32_t x;
+      uint64_t x;
       double re;
     };
     // NOLINTNEXTLINE(cppcoreguidelines-pro-type-member-init)
     union {
-      uint32_t y;
+      uint64_t y;
       double im;
     };
+    static_assert(sizeof(uint64_t) == sizeof(double));
 
     x = (do_byte_swap ? decodeUInt64BE(src) : decodeUInt64LE(src));
     src += sizeof(double);
@@ -462,7 +463,7 @@ void THP_encodeComplexFloatBuffer(
   }
 }
 
-void THP_encodeCompelxDoubleBuffer(
+void THP_encodeComplexDoubleBuffer(
     uint8_t* dst,
     const c10::complex<double>* src,
     THPByteOrder order,