Fix CPP lint. (#8807)

src/data/sparse_page_dmatrix.cc

@@ -19,7 +19,7 @@ const MetaInfo &SparsePageDMatrix::Info() const { return info_; }
 
 namespace detail {
 // Use device dispatch
-std::size_t NSamplesDevice(DMatrixProxy *)
+std::size_t NSamplesDevice(DMatrixProxy *)  // NOLINT
 #if defined(XGBOOST_USE_CUDA)
 ;  // NOLINT
 #else
@@ -28,7 +28,7 @@ std::size_t NSamplesDevice(DMatrixProxy *)
   return 0;
 }
 #endif
-std::size_t NFeaturesDevice(DMatrixProxy *)
+std::size_t NFeaturesDevice(DMatrixProxy *)  // NOLINT
 #if defined(XGBOOST_USE_CUDA)
 ;  // NOLINT
 #else

src/metric/auc.cu

@@ -5,7 +5,7 @@
 
 #include <algorithm>
 #include <cassert>
-#include <cub/cub.cuh>
+#include <cub/cub.cuh>  // NOLINT
 #include <limits>
 #include <memory>
 #include <tuple>

src/metric/metric_common.h

@@ -6,6 +6,7 @@
 #define XGBOOST_METRIC_METRIC_COMMON_H_
 
 #include <limits>
+#include <memory>  // shared_ptr
 #include <string>
 
 #include "../common/common.h"

src/objective/adaptive.cu

@@ -3,8 +3,8 @@
  */
 #include <thrust/sort.h>
 
-#include <cstdint>  // std::int32_t
-#include <cub/cub.cuh>
+#include <cstdint>                     // std::int32_t
+#include <cub/cub.cuh>                 // NOLINT
 
 #include "../common/cuda_context.cuh"  // CUDAContext
 #include "../common/device_helpers.cuh"

src/predictor/cpu_treeshap.cc

@@ -3,10 +3,11 @@
  */
 #include "cpu_treeshap.h"
 
-#include <cinttypes>  // std::uint32_t
+#include <algorithm>             // copy
+#include <cinttypes>             // std::uint32_t
 
-#include "predict_fn.h"    // GetNextNode
-#include "xgboost/base.h"  // bst_node_t
+#include "predict_fn.h"          // GetNextNode
+#include "xgboost/base.h"        // bst_node_t
 #include "xgboost/logging.h"
 #include "xgboost/tree_model.h"  // RegTree
 

src/predictor/cpu_treeshap.h

@@ -1,6 +1,10 @@
+#ifndef XGBOOST_PREDICTOR_CPU_TREESHAP_H_
+#define XGBOOST_PREDICTOR_CPU_TREESHAP_H_
 /**
  * Copyright by XGBoost Contributors 2017-2022
  */
+#include <vector>                // vector
+
 #include "xgboost/tree_model.h"  // RegTree
 
 namespace xgboost {
@@ -15,3 +19,4 @@ void CalculateContributions(RegTree const &tree, const RegTree::FVec &feat,
                             std::vector<float> *mean_values, bst_float *out_contribs, int condition,
                             unsigned condition_feature);
 }  // namespace xgboost
+#endif  // XGBOOST_PREDICTOR_CPU_TREESHAP_H_

src/tree/gpu_hist/evaluate_splits.cu

@@ -359,8 +359,8 @@ void GPUHistEvaluator::LaunchEvaluateSplits(
 
   // One block for each feature
   uint32_t constexpr kBlockThreads = 32;
-  dh::LaunchKernel{static_cast<uint32_t>(combined_num_features), kBlockThreads,
-                   0}(
+  dh::LaunchKernel {static_cast<uint32_t>(combined_num_features), kBlockThreads,
+                    0}(
       EvaluateSplitsKernel<kBlockThreads>, max_active_features, d_inputs,
       shared_inputs,
       this->SortedIdx(d_inputs.size(), shared_inputs.feature_values.size()),

src/tree/gpu_hist/histogram.cu

@@ -299,7 +299,7 @@ void BuildGradientHistogram(CUDAContext const* ctx, EllpackDeviceAccessor const&
         min(grid_size,
             unsigned(common::DivRoundUp(items_per_group, kMinItemsPerBlock)));
 
-    dh::LaunchKernel{dim3(grid_size, num_groups), static_cast<uint32_t>(kBlockThreads), smem_size,
+    dh::LaunchKernel {dim3(grid_size, num_groups), static_cast<uint32_t>(kBlockThreads), smem_size,
                      ctx->Stream()} (kernel, matrix, feature_groups, d_ridx, histogram.data(),
                                      gpair.data(), rounding);
   };