Forbid trailing whitespace (pytorch#53406)

Summary:
Context: pytorch#53299 (comment)

These are the only hand-written parts of this diff:
- the addition to `.github/workflows/lint.yml`
- the file endings changed in these four files (to appease FB-internal land-blocking lints):
  - `GLOSSARY.md`
  - `aten/src/ATen/core/op_registration/README.md`
  - `scripts/README.md`
  - `torch/csrc/jit/codegen/fuser/README.md`

The rest was generated by running this command (on macOS):
```
git grep -I -l ' $' -- . ':(exclude)**/contrib/**' ':(exclude)third_party' | xargs gsed -i 's/ *$//'
```

I looked over the auto-generated changes and didn't see anything that looked problematic.

Pull Request resolved: pytorch#53406

Test Plan:
This run (after adding the lint but before removing existing trailing spaces) failed:
- https://github.com/pytorch/pytorch/runs/2043032377

This run (on the tip of this PR) succeeded:
- https://github.com/pytorch/pytorch/runs/2043296348

Reviewed By: walterddr, seemethere

Differential Revision: D26856620

Pulled By: samestep

fbshipit-source-id: 3f0de7f7c2e4b0f1c089eac9b5085a58dd7e0d97

Author: samestep

.circleci/scripts/binary_ios_test.sh

@@ -24,6 +24,6 @@ rm cert.txt
 if ! [ -x "$(command -v xcodebuild)" ]; then
     echo 'Error: xcodebuild is not installed.'
     exit 1
-fi 
+fi
 PROFILE=PyTorch_CI_2021
 ruby ${PROJ_ROOT}/scripts/xcode_build.rb -i ${PROJ_ROOT}/build_ios/install -x ${PROJ_ROOT}/ios/TestApp/TestApp.xcodeproj -p ${IOS_PLATFORM} -c ${PROFILE} -t ${IOS_DEV_TEAM_ID}

.github/workflows/lint.yml

@@ -40,6 +40,9 @@ jobs:
           rm -r "shellcheck-${scversion}"
           shellcheck --version
           .jenkins/run-shellcheck.sh
+      - name: Ensure no trailing spaces
+        run: |
+          (! git grep -I -l ' $' -- . ':(exclude)**/contrib/**' ':(exclude)third_party' || (echo "The above files have trailing spaces; please remove them"; false))
       - name: Ensure no tabs
         run: |
           (! git grep -I -l $'\t' -- . ':(exclude)*.svg' ':(exclude)**Makefile' ':(exclude)**/contrib/**' ':(exclude)third_party' ':(exclude).gitattributes' ':(exclude).gitmodules' || (echo "The above files have tabs; please convert them to spaces"; false))

.jenkins/caffe2/bench.sh

@@ -21,7 +21,7 @@ if (( $num_gpus == 0 )); then
 fi
 if (( $num_gpus >= 1 )); then
     "$PYTHON" "$caffe2_pypath/python/examples/imagenet_trainer.py" --train_data null --batch_size 128 --epoch_size 12800 --num_epochs 2 --num_gpus 1
-    # Let's skip the fp16 bench runs for now, as it recompiles the miopen kernels and can take 10+min to run. 
+    # Let's skip the fp16 bench runs for now, as it recompiles the miopen kernels and can take 10+min to run.
     # We can resume when we (1) bindmount the miopen cache folder in jenkins; (2) install the pre-compiled miopen kernel library in the docker
     # "$PYTHON" "$caffe2_pypath/python/examples/imagenet_trainer.py" --train_data null --batch_size 256 --epoch_size 25600 --num_epochs 2 --num_gpus 1 --float16_compute --dtype float16
 fi

CONTRIBUTING.md

@@ -159,7 +159,7 @@ with `brew install cmake` if you are developing on MacOS or Linux system.
     check whether your Git local or global config file contains any `submodule.*` settings. If yes, remove them and try again.
     (please reference [this doc](https://git-scm.com/docs/git-config#Documentation/git-config.txt-submoduleltnamegturl) for more info).
 
-  - If you encountered error such as 
+  - If you encountered error such as
     ```
     fatal: unable to access 'https://github.com/pybind11/pybind11.git': could not load PEM client certificate ...
     ```
@@ -169,11 +169,11 @@ with `brew install cmake` if you are developing on MacOS or Linux system.
     openssl x509 -noout -in <cert_file> -dates
     ```
 
-  - If you encountered error that some third_party modules are not checkout correctly, such as 
+  - If you encountered error that some third_party modules are not checkout correctly, such as
     ```
     Could not find .../pytorch/third_party/pybind11/CMakeLists.txt
     ```
-    remove any `submodule.*` settings in your local git config (`.git/config` of your pytorch repo) and try again.  
+    remove any `submodule.*` settings in your local git config (`.git/config` of your pytorch repo) and try again.
 
 ## Nightly Checkout & Pull
 

GLOSSARY.md

@@ -1,4 +1,4 @@
-# PyTorch Glossary 
+# PyTorch Glossary
 
 - [PyTorch Glossary](#pytorch-glossary)
 - [Operation and Kernel](#operation-and-kernel)
@@ -39,7 +39,7 @@ For example, this
 to create Custom Operations.
 
 ## Kernel
-Implementation of a PyTorch operation, specifying what should be done when an 
+Implementation of a PyTorch operation, specifying what should be done when an
 operation executes.
 
 ## Compound Operation
@@ -57,7 +57,7 @@ Same as Compound Operation.
 ## Leaf Operation
 An operation that's considered a basic operation, as opposed to a Compound
 Operation. Leaf Operation always has dispatch functions defined, usually has a
-derivative function defined as well. 
+derivative function defined as well.
 
 ## Device Kernel
 Device-specific kernel of a leaf operation.
@@ -79,4 +79,4 @@ using just-in-time compilation.
 
 ## Scripting
 Using `torch.jit.script` on a function to inspect source code and compile it as
-TorchScript code.
+TorchScript code.

aten/src/ATen/BatchingRegistrations.cpp

@@ -300,7 +300,7 @@ Tensor trace_backward_batching_rule(const Tensor& grad, IntArrayRef input_sizes)
   auto grad_input = at::zeros(grad_physical.getPhysicalShape(input_sizes), grad.options());
   // Batched Diagonal View
   auto grad_input_diag = at::diagonal(grad_input, /*offset*/0, /*dim1*/-2, /*dim2*/-1);
-  // Append a dimension of size one to the grad output 
+  // Append a dimension of size one to the grad output
   auto grad_physical_tensor = grad_physical.tensor().unsqueeze(-1);
   grad_input_diag.copy_(grad_physical_tensor);
   return grad_physical.getPhysicalToLogicalMap().apply(grad_input);

aten/src/ATen/CPUGeneratorImpl.cpp

@@ -38,7 +38,7 @@ struct CPUGeneratorImplStateLegacy {
  * new data introduced in at::CPUGeneratorImpl and the legacy state. It is used
  * as a helper for torch.get_rng_state() and torch.set_rng_state()
  * functions.
- */ 
+ */
 struct CPUGeneratorImplState {
   CPUGeneratorImplStateLegacy legacy_pod;
   float next_float_normal_sample;
@@ -119,7 +119,7 @@ uint64_t CPUGeneratorImpl::seed() {
  * must be a strided CPU byte tensor and of the same size as either
  * CPUGeneratorImplStateLegacy (for legacy CPU generator state) or
  * CPUGeneratorImplState (for new state).
- * 
+ *
  * FIXME: Remove support of the legacy state in the future?
  */
 void CPUGeneratorImpl::set_state(const c10::TensorImpl& new_state) {

aten/src/ATen/SparseTensorUtils.h

@@ -94,7 +94,7 @@ TORCH_API Tensor flatten_indices(const Tensor& indices, IntArrayRef full_size, b
 //   new_indices = [ 3, 1, 3 ]  # uncoalesced
 TORCH_API Tensor flatten_indices_by_dims(const Tensor& indices, const IntArrayRef& sizes, const IntArrayRef& dims_to_flatten);
 
-// Find the CSR representation for a row `indices` from the COO format 
+// Find the CSR representation for a row `indices` from the COO format
 TORCH_API Tensor coo_to_csr(const int64_t* indices, int64_t dim, int64_t nnz);
 
 }} // namespace at::sparse

aten/src/ATen/Version.cpp

@@ -114,7 +114,7 @@ std::string used_cpu_capability() {
     case native::CPUCapability::AVX2:
       ss << "AVX2";
       break;
-#endif      
+#endif
     default:
       break;
   }

aten/src/ATen/VmapTransforms.h

@@ -47,7 +47,7 @@ using VmapDimVector = SmallVector<int64_t, kVmapStaticDimVecSize>;
 // argument.
 
 // VmapTransform for operators that take tensors with multiple batch dims.
-// Given one or more logical views on Tensors, `logicalToPhysical` 
+// Given one or more logical views on Tensors, `logicalToPhysical`
 // permutes all of the batch dims to the front of the tensor, aligns
 // and expands the batch dims to match each other (according to their `level`),
 // and returns a VmapPhysicalView on the tensor(s).

aten/src/ATen/core/Generator.h

@@ -143,7 +143,7 @@ namespace detail {
 /**
  * Helper function for checking the validity of new random generator
  * state. Right now following conditions are checked:
- * 
+ *
  * - The new state tensor must be a torch.ByteTensor
  * - Data of the new state tensor must be contiguous
  */

aten/src/ATen/core/PhiloxRNGEngine.h

@@ -40,13 +40,13 @@ typedef at::detail::Array<float, 2> FLOAT2;
  * Note that currently this implementation of the philox engine is not used
  * anywhere except for tests in cpu_generator_test.cpp. However, this engine
  * will replace curandStatePhilox4_32_10_t in the future.
- * 
+ *
  * The philox engine takes a seed value, a subsequeunce
  * for starting the generation and an offset for the subsequence.
- * Think of this engine as an algorithm producing a huge array. We are 
- * parallelizing this array by partitioning the huge array and assigning 
- * a thread index to each partition. In other words, each seed value 
- * (there are 2^64 possible seed values) gives a sub array of size 
+ * Think of this engine as an algorithm producing a huge array. We are
+ * parallelizing this array by partitioning the huge array and assigning
+ * a thread index to each partition. In other words, each seed value
+ * (there are 2^64 possible seed values) gives a sub array of size
  * 2^128 (each element in that array is a 128 bit number). Reasoning
  * behind the array being of size 2^128 is, there are 2^64 possible
  * thread index value and there is an array of size 2^64 for each of
@@ -59,9 +59,9 @@ typedef at::detail::Array<float, 2> FLOAT2;
  * seed:        Seed values could be any number from 0 to 2^64-1.
  * subsequence: Subsequence is just the cuda thread indexing with:
  *              - blockIdx.x * blockDim.x + threadIdx.x
- * offset:      The offset variable in PhiloxEngine  decides how many 128-bit 
+ * offset:      The offset variable in PhiloxEngine  decides how many 128-bit
  *              random numbers to skip (i.e. how many groups of 4, 32-bit numbers to skip)
- *              and hence really decides the total number of randoms that can be achieved 
+ *              and hence really decides the total number of randoms that can be achieved
  *              for the given subsequence.
  */
 

aten/src/ATen/core/op_registration/README.md

@@ -254,5 +254,3 @@ Also, there's some requirements on the operator schema for it to be callable fro
 * Except for `Tensor` or `Tensor[]`, only arguments of type `int`, `double` and `bool` are supported. These can be in any position in the argument list and will be read from the caffe2 operator arguments, based on the argument name in the operator schema.
 * We do not support lists (`int[]`, `double[]` or `bool[]`) or optionals (`int?`, `double?`, `bool?`) yet.
 * The operator must return a single `Tensor` or multiple tensors as in `(Tensor, Tensor, Tensor)`. It cannot return a list `Tensor[]`, optional `Tensor?` or any primitive types.
-    
-

aten/src/ATen/core/type.cpp

@@ -1124,12 +1124,12 @@ std::string ClassType::getForwardPreHookErrorMessage(int pre_hook_idx) const {
   const FunctionSchema& forward_schema = getMethod("forward").getSchema();
   std::string input_types = getSchemaInputTypesString(forward_schema);
   const std::vector<Argument>& forward_args = forward_schema.arguments();
-   
+
   std::string single_output = "";
   if (forward_args.size() == 2 &&
       forward_args[1].type()->cast<TupleType>() == nullptr) {
     // if the output type is a single tuple, it needs to be wrapped in an outer tuple
-    // to match eager's behavior 
+    // to match eager's behavior
     single_output = ", '" + forward_args[1].type()->annotation_str() + "',";
   }
   std::string pre_hook_schema =
@@ -1138,17 +1138,17 @@ std::string ClassType::getForwardPreHookErrorMessage(int pre_hook_idx) const {
       "This error occured while scripting the forward pre-hook '" +
       pre_hook_name + "' on module '" + name()->name() +
       "'. If you did not want to script this pre-hook remove it from the "
-      "original NN module before scripting. Pre-hooks for module '" + 
-      name()->name() + "' are expected to have the following signature: " 
-      + pre_hook_schema + " with a return type of either 'None'" + 
+      "original NN module before scripting. Pre-hooks for module '" +
+      name()->name() + "' are expected to have the following signature: "
+      + pre_hook_schema + " with a return type of either 'None'" +
       single_output + " or 'Tuple[" + input_types + "]'.";
   return return_string;
 }
 
 std::string ClassType::getForwardHookErrorMessage(int hook_idx) const {
   const std::string& hook_name = forward_hooks_[hook_idx]->name();
   const FunctionSchema& forward_schema = getMethod("forward").getSchema();
-  std::string input_types = getSchemaInputTypesString(forward_schema); 
+  std::string input_types = getSchemaInputTypesString(forward_schema);
 
   // create expected output types string
   const Argument& pre_output =
@@ -1160,33 +1160,33 @@ std::string ClassType::getForwardHookErrorMessage(int hook_idx) const {
   std::string hook_schema = hook_name + "(self, input: Tuple[" +
                             input_types + "], output: " + output_types + ")";
   std::string return_string =
-      "This error occured while scripting the forward hook '" 
+      "This error occured while scripting the forward hook '"
       + hook_name + "' on module " + name()->name() +
       ". If you did not want to script this hook remove it from" +
       " the original NN module before scripting. This hook was" +
       " expected to have the following signature: " + hook_schema +
-      ". The type of the output arg is the returned type from" + 
-      " either the forward method or the previous hook if it exists. " + 
-      "Note that hooks can return anything, but if the hook is " + 
+      ". The type of the output arg is the returned type from" +
+      " either the forward method or the previous hook if it exists. " +
+      "Note that hooks can return anything, but if the hook is " +
       "on a submodule the outer module is expecting" +
       " the same return type as the submodule's forward.";
   return return_string;
 }
 
 void checkForwardHookInputArguments(
-    const FunctionSchema& forward_schema, 
-    const FunctionSchema& hook_schema, 
-    const std::string& hook_id, 
+    const FunctionSchema& forward_schema,
+    const FunctionSchema& hook_schema,
+    const std::string& hook_id,
     const std::string& hook_err_msg) {
   // check for proper tuple input types
   const std::vector<Argument>& forward_args = forward_schema.arguments();
   const Argument input_arg = hook_schema.arguments()[1];
   TORCH_CHECK(
-      input_arg.type()->cast<TupleType>() != nullptr, 
+      input_arg.type()->cast<TupleType>() != nullptr,
       hook_id,
       "expected the input argument to be typed as a Tuple but found type: '",
-      input_arg.type()->annotation_str(), 
-      "' instead.\n", 
+      input_arg.type()->annotation_str(),
+      "' instead.\n",
       hook_err_msg
    );
 
@@ -1229,7 +1229,7 @@ void checkForwardHookInputArguments(
 }
 
 void ClassType::checkForwardPreHookSchema(
-    int pre_hook_idx, 
+    int pre_hook_idx,
     const FunctionSchema& pre_hook_schema) const {
   const torch::jit::Function* pre_hook = forward_pre_hooks_[pre_hook_idx];
   std::string hook_id =
@@ -1261,17 +1261,17 @@ void ClassType::checkForwardPreHookSchema(
             pre_hook_err_msg
   );
   const Argument return_arg = pre_hook_schema.returns()[0];
-  std::string wrong_type_returned_err_msg = hook_id + 
+  std::string wrong_type_returned_err_msg = hook_id +
       "returned the wrong type of: '" +
       return_arg.type()->annotation_str() + "'.";
 
   if (return_arg.type()->kind() == NoneType::get()->kind()) {
     return;
   }
   if (forward_args.size() == 2 && *forward_args[1].type() == *return_arg.type()) {
-    // TORCH_CHECK below is for the edge case where forward's input is a tuple and the 
+    // TORCH_CHECK below is for the edge case where forward's input is a tuple and the
     // pre-hook returns a matching tuple. Eager doesn't support this- the working eager return
-    // for a tuple type is the forward's input tuple wrapped inside of another tuple. 
+    // for a tuple type is the forward's input tuple wrapped inside of another tuple.
     TORCH_CHECK(
         return_arg.type()->cast<TupleType>() == nullptr,
         wrong_type_returned_err_msg,
@@ -1316,7 +1316,7 @@ void ClassType::checkForwardPreHookSchema(
   for (int i = 1; i < forward_args.size(); ++i) {
     if (*forward_args[i].type() != *return_tuple_types[i - 1]) {
       TORCH_CHECK(
-          false, 
+          false,
           wrong_type_returned_err_msg,
           " The returned tuple contains the wrong inner types.\n",
           pre_hook_err_msg);
@@ -1325,7 +1325,7 @@ void ClassType::checkForwardPreHookSchema(
 }
 
 void ClassType::checkForwardHookSchema(
-      int hook_idx, 
+      int hook_idx,
       const FunctionSchema& hook_schema) const {
   const torch::jit::Function* hook = forward_hooks_[hook_idx];
   std::string hook_id =
@@ -1388,8 +1388,8 @@ torch::jit::Function& ClassType::getMethod(const std::string& name) const {
 torch::jit::Function* ClassType::findHook(const std::string& name) const {
   auto hook = findForwardHook(name);
   if (hook == nullptr) {
-    hook = findForwardPreHook(name); 
-  } 
+    hook = findForwardPreHook(name);
+  }
   return hook;
 }
 

aten/src/ATen/cpu/vec256/vec256_double.h

@@ -113,7 +113,7 @@ template <> class Vec256<double> {
     const auto not_nan_mask = _mm256_cmp_pd(values, values, _CMP_EQ_OQ);
     const auto nan_mask = _mm256_cmp_pd(not_nan_mask, zero_vec, _CMP_EQ_OQ);
     const auto pi = _mm256_set1_pd(c10::pi<double>);
-    
+
     const auto neg_mask = _mm256_cmp_pd(values, zero_vec, _CMP_LT_OQ);
     auto angle = _mm256_blendv_pd(zero_vec, pi, neg_mask);
     angle = _mm256_blendv_pd(angle, nan_vec, nan_mask);

aten/src/ATen/cpu/vec256/vec256_float.h

@@ -120,7 +120,7 @@ template <> class Vec256<float> {
     const auto not_nan_mask = _mm256_cmp_ps(values, values, _CMP_EQ_OQ);
     const auto nan_mask = _mm256_cmp_ps(not_nan_mask, zero_vec, _CMP_EQ_OQ);
     const auto pi = _mm256_set1_ps(c10::pi<float>);
-    
+
     const auto neg_mask = _mm256_cmp_ps(values, zero_vec, _CMP_LT_OQ);
     auto angle = _mm256_blendv_ps(zero_vec, pi, neg_mask);
     angle = _mm256_blendv_ps(angle, nan_vec, nan_mask);

aten/src/ATen/cpu/vec256/vsx/vec256_complex_double_vsx.h

@@ -364,7 +364,7 @@ class Vec256<ComplexDbl> {
   }
 
   Vec256<ComplexDbl> sqrt() const {
-    return map(std::sqrt);  
+    return map(std::sqrt);
   }
 
   Vec256<ComplexDbl> reciprocal() const {

aten/src/ATen/cpu/vec256/vsx/vec256_complex_float_vsx.h

@@ -417,7 +417,7 @@ class Vec256<ComplexFlt> {
   }
 
   Vec256<ComplexFlt> sqrt() const {
-    return map(std::sqrt);  
+    return map(std::sqrt);
   }
 
   Vec256<ComplexFlt> reciprocal() const {

aten/src/ATen/cpu/vec256/vsx/vec256_double_vsx.h

@@ -82,7 +82,7 @@ class Vec256<double> {
       blend(const Vec256<double>& a, const Vec256<double>& b) {
       return { a._vec0, b._vec1 };
   }
- 
+
 
   template <int64_t mask>
   static std::enable_if_t<blendChoiceDbl(mask) == 4, Vec256<double>> C10_ALWAYS_INLINE
@@ -206,7 +206,7 @@ class Vec256<double> {
     for (int i = 0; i < size()/2; i++) {
         ret._vec0[i] = f(_vec0[i], other._vec0[i]);
     }
-    for (int i = 0; i < size()/2; i++) {   
+    for (int i = 0; i < size()/2; i++) {
         ret._vec1[i] = f(_vec1[i], other._vec1[i]);
     }
     return ret;
@@ -314,7 +314,7 @@ class Vec256<double> {
   Vec256<double> C10_ALWAYS_INLINE sqrt() const {
     return {vec_sqrt(_vec0), vec_sqrt(_vec1)};
   }
-  Vec256<double> C10_ALWAYS_INLINE reciprocal() const { 
+  Vec256<double> C10_ALWAYS_INLINE reciprocal() const {
     return {
         vec_div(vd_one, _vec0), // vec_re(_vec0) is estimated one.
         vec_div(vd_one, _vec1)};