Fixes devices vector alloc to fix seg fault, removes unused RAFT code…

… in PLC, re-enables full CI testing (#3167)

closes #3124 

* Adds check to avoid allocating and copying zero-length device vectors.  This prevents the seg fault shown below.
* Removes the special case to ignore seg faults in CI scripts
* Adds a test to reproduce seg fault locally (see output below).

This PR addresses the problem shown below:
```
================================= test session starts =================================
platform linux -- Python 3.8.15, pytest-7.2.0, pluggy-1.0.0 -- /opt/conda/envs/test/bin/python3.8
cachedir: .pytest_cache
rapids_pytest_benchmark: 0.0.14
benchmark: 3.2.3 (defaults: timer=time.perf_counter disable_gc=False min_rounds=5 min_time=0.000005 max_time=1.0 calibration_precision=10 warmup=False warmup_iterations=100000)
rootdir: /cugraph/python/pylibcugraph, configfile: pytest.ini
plugins: cov-4.0.0, rapids-pytest-benchmark-0.0.14, benchmark-3.2.3
collected 9 items / 8 deselected / 1 selected

python/pylibcugraph/pylibcugraph/tests/test_graph_sg.py::test_SGGraph_create_from_cudf
get edgelist...edgelist =     src  dst  wgt
0    0    1  0.0
1    1    2  0.1
2    2    4  0.2
done
create Graph...done
created SGGraph plc_graph=<pylibcugraph.graphs.SGGraph object at 0x7fb7e35f30f0>
PASSED

=========================== 1 passed, 8 deselected in 1.69s ===========================
Segmentation fault (core dumped)
```

@cjnolet found a work-around for us, so this should pass CI and can be merged after rapidsai/raft#1224

Authors:
  - Rick Ratzel (https://github.com/rlratzel)
  - Chuck Hastings (https://github.com/ChuckHastings)
  - Corey J. Nolet (https://github.com/cjnolet)
  - AJ Schmidt (https://github.com/ajschmidt8)

Approvers:
  - AJ Schmidt (https://github.com/ajschmidt8)
  - Chuck Hastings (https://github.com/ChuckHastings)
  - Brad Rees (https://github.com/BradReesWork)

URL: #3167

ci/test_notebooks.sh

@@ -32,10 +32,7 @@ rapids-mamba-retry install \
 NBTEST="$(realpath "$(dirname "$0")/utils/nbtest.sh")"
 NOTEBOOK_LIST="$(realpath "$(dirname "$0")/gpu/notebook_list.py")"
 EXITCODE=0
-# FIXME: This is temporary until a crash that occurs at cleanup is fixed. This
-# allows PRs that pass tests to pass even if they crash with a Seg Fault or
-# other error that results in 139. Remove this ASAP!
-# trap "EXITCODE=1" ERR
+trap "EXITCODE=1" ERR
 
 
 pushd notebooks
@@ -52,13 +49,6 @@ for folder in ${TOPLEVEL_NB_FOLDERS}; do
         pushd "$(dirname "${nb}")"
         nvidia-smi
         ${NBTEST} "${nbBasename}"
-            # FIXME: This is temporary until a crash that occurs at cleanup is fixed. This
-            # allows PRs that pass tests to pass even if they crash with a Seg Fault or
-            # other error that results in 139. Remove this ASAP!
-            exitcode=$?
-            if (( (${exitcode} != 0) && (${exitcode} != 139) )); then
-                EXITCODE=1
-            fi
         echo "Ran nbtest for $nb : return code was: $?, test script exit code is now: $EXITCODE"
         echo
         popd

ci/test_python.sh

@@ -62,11 +62,7 @@ pytest \
   tests
 exitcode=$?
 
-# FIXME: This is temporary until a crash that occurs at cleanup is fixed. This
-# allows PRs that pass tests to pass even if they crash with a Seg Fault or
-# other error that results in 139. Remove this ASAP!
-# if (( ${exitcode} != 0 )); then
-if (( (${exitcode} != 0) && (${exitcode} != 139) )); then
+if (( ${exitcode} != 0 )); then
     SUITEERROR=${exitcode}
     echo "FAILED: 1 or more tests in pylibcugraph"
 fi
@@ -85,11 +81,7 @@ pytest \
   tests
 exitcode=$?
 
-# FIXME: This is temporary until a crash that occurs at cleanup is fixed. This
-# allows PRs that pass tests to pass even if they crash with a Seg Fault or
-# other error that results in 139. Remove this ASAP!
-# if (( ${exitcode} != 0 )); then
-if (( (${exitcode} != 0) && (${exitcode} != 139) )); then
+if (( ${exitcode} != 0 )); then
     SUITEERROR=${exitcode}
     echo "FAILED: 1 or more tests in cugraph"
 fi
@@ -105,11 +97,7 @@ pytest \
   cugraph/pytest-based/bench_algos.py
 exitcode=$?
 
-# FIXME: This is temporary until a crash that occurs at cleanup is fixed. This
-# allows PRs that pass tests to pass even if they crash with a Seg Fault or
-# other error that results in 139. Remove this ASAP!
-# if (( ${exitcode} != 0 )); then
-if (( (${exitcode} != 0) && (${exitcode} != 139) )); then
+if (( ${exitcode} != 0 )); then
     SUITEERROR=${exitcode}
     echo "FAILED: 1 or more tests in cugraph benchmarks"
 fi
@@ -130,11 +118,7 @@ pytest \
   .
 exitcode=$?
 
-# FIXME: This is temporary until a crash that occurs at cleanup is fixed. This
-# allows PRs that pass tests to pass even if they crash with a Seg Fault or
-# other error that results in 139. Remove this ASAP!
-# if (( ${exitcode} != 0 )); then
-if (( (${exitcode} != 0) && (${exitcode} != 139) )); then
+if (( ${exitcode} != 0 )); then
     SUITEERROR=${exitcode}
     echo "FAILED: 1 or more tests in cugraph-pyg"
 fi
@@ -157,11 +141,7 @@ pytest \
   tests
 exitcode=$?
 
-# FIXME: This is temporary until a crash that occurs at cleanup is fixed. This
-# allows PRs that pass tests to pass even if they crash with a Seg Fault or
-# other error that results in 139. Remove this ASAP!
-# if (( ${exitcode} != 0 )); then
-if (( (${exitcode} != 0) && (${exitcode} != 139) )); then
+if (( ${exitcode} != 0 )); then
     SUITEERROR=${exitcode}
     echo "FAILED: 1 or more tests in cugraph-service"
 fi

cpp/include/cugraph/utilities/misc_utils.cuh

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2022, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2023, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -45,33 +45,39 @@ std::tuple<std::vector<vertex_t>, std::vector<edge_t>> compute_offset_aligned_ed
     thrust::make_counting_iterator(size_t{1}),
     [approx_edge_chunk_size] __device__(auto i) { return i * approx_edge_chunk_size; });
   auto num_chunks = (num_edges + approx_edge_chunk_size - 1) / approx_edge_chunk_size;
-  rmm::device_uvector<vertex_t> d_vertex_offsets(num_chunks - 1, handle.get_stream());
-  thrust::lower_bound(handle.get_thrust_policy(),
-                      offsets,
-                      offsets + num_vertices + 1,
-                      search_offset_first,
-                      search_offset_first + d_vertex_offsets.size(),
-                      d_vertex_offsets.begin());
-  rmm::device_uvector<edge_t> d_edge_offsets(d_vertex_offsets.size(), handle.get_stream());
-  thrust::gather(handle.get_thrust_policy(),
-                 d_vertex_offsets.begin(),
-                 d_vertex_offsets.end(),
-                 offsets,
-                 d_edge_offsets.begin());
-  std::vector<edge_t> h_edge_offsets(num_chunks + 1, edge_t{0});
-  h_edge_offsets.back() = num_edges;
-  raft::update_host(
-    h_edge_offsets.data() + 1, d_edge_offsets.data(), d_edge_offsets.size(), handle.get_stream());
-  std::vector<vertex_t> h_vertex_offsets(num_chunks + 1, vertex_t{0});
-  h_vertex_offsets.back() = num_vertices;
-  raft::update_host(h_vertex_offsets.data() + 1,
-                    d_vertex_offsets.data(),
-                    d_vertex_offsets.size(),
-                    handle.get_stream());
 
-  handle.sync_stream();
+  if (num_chunks > 1) {
+    rmm::device_uvector<vertex_t> d_vertex_offsets(num_chunks - 1, handle.get_stream());
+    thrust::lower_bound(handle.get_thrust_policy(),
+                        offsets,
+                        offsets + num_vertices + 1,
+                        search_offset_first,
+                        search_offset_first + d_vertex_offsets.size(),
+                        d_vertex_offsets.begin());
+    rmm::device_uvector<edge_t> d_edge_offsets(d_vertex_offsets.size(), handle.get_stream());
+    thrust::gather(handle.get_thrust_policy(),
+                   d_vertex_offsets.begin(),
+                   d_vertex_offsets.end(),
+                   offsets,
+                   d_edge_offsets.begin());
+    std::vector<edge_t> h_edge_offsets(num_chunks + 1, edge_t{0});
+    h_edge_offsets.back() = num_edges;
+    raft::update_host(
+      h_edge_offsets.data() + 1, d_edge_offsets.data(), d_edge_offsets.size(), handle.get_stream());
+    std::vector<vertex_t> h_vertex_offsets(num_chunks + 1, vertex_t{0});
+    h_vertex_offsets.back() = num_vertices;
+    raft::update_host(h_vertex_offsets.data() + 1,
+                      d_vertex_offsets.data(),
+                      d_vertex_offsets.size(),
+                      handle.get_stream());
 
-  return std::make_tuple(h_vertex_offsets, h_edge_offsets);
+    handle.sync_stream();
+
+    return std::make_tuple(h_vertex_offsets, h_edge_offsets);
+  } else {
+    return std::make_tuple(std::vector<vertex_t>{{0, num_vertices}},
+                           std::vector<edge_t>{{0, num_edges}});
+  }
 }
 
 template <typename T>

cpp/include/cugraph_c/sampling_algorithms.h

@@ -354,8 +354,11 @@ void cugraph_sample_result_free(cugraph_sample_result_t* result);
  * @param [in]   handle         Handle for accessing resources
  * @param [in]   srcs           Device array view to populate srcs
  * @param [in]   dsts           Device array view to populate dsts
- * @param [in]   weights        Device array view to populate weights
- * @param [in]   counts         Device array view to populate counts
+ * @param [in]   edge_id        Device array view to populate edge_id (can be NULL)
+ * @param [in]   edge_type      Device array view to populate edge_type (can be NULL)
+ * @param [in]   wgt            Device array view to populate wgt (can be NULL)
+ * @param [in]   hop            Device array view to populate hop
+ * @param [in]   label          Device array view to populate label (can be NULL)
  * @param [out]  result         Pointer to the location to store the
  *                              cugraph_sample_result_t*
  * @param [out]  error          Pointer to an error object storing details of
@@ -367,8 +370,11 @@ cugraph_error_code_t cugraph_test_sample_result_create(
   const cugraph_resource_handle_t* handle,
   const cugraph_type_erased_device_array_view_t* srcs,
   const cugraph_type_erased_device_array_view_t* dsts,
-  const cugraph_type_erased_device_array_view_t* weights,
-  const cugraph_type_erased_device_array_view_t* counts,
+  const cugraph_type_erased_device_array_view_t* edge_id,
+  const cugraph_type_erased_device_array_view_t* edge_type,
+  const cugraph_type_erased_device_array_view_t* wgt,
+  const cugraph_type_erased_device_array_view_t* hop,
+  const cugraph_type_erased_device_array_view_t* label,
   cugraph_sample_result_t** result,
   cugraph_error_t** error);
 

cpp/src/c_api/uniform_neighbor_sampling.cpp

@@ -526,8 +526,11 @@ extern "C" cugraph_error_code_t cugraph_test_sample_result_create(
   const cugraph_resource_handle_t* handle,
   const cugraph_type_erased_device_array_view_t* srcs,
   const cugraph_type_erased_device_array_view_t* dsts,
-  const cugraph_type_erased_device_array_view_t* weights,
-  const cugraph_type_erased_device_array_view_t* counts,
+  const cugraph_type_erased_device_array_view_t* edge_id,
+  const cugraph_type_erased_device_array_view_t* edge_type,
+  const cugraph_type_erased_device_array_view_t* wgt,
+  const cugraph_type_erased_device_array_view_t* hop,
+  const cugraph_type_erased_device_array_view_t* label,
   cugraph_sample_result_t** result,
   cugraph_error_t** error)
 {
@@ -567,36 +570,77 @@ extern "C" cugraph_error_code_t cugraph_test_sample_result_create(
   device_array_unique_ptr_t new_device_dsts(new_device_dsts_ptr,
                                             &cugraph_type_erased_device_array_free);
 
-  // copy weights to new device array
-  cugraph_type_erased_device_array_t* new_device_weights_ptr{nullptr};
-  error_code = cugraph_type_erased_device_array_create_from_view(
-    handle, weights, &new_device_weights_ptr, error);
-  if (error_code != CUGRAPH_SUCCESS) return error_code;
+  // copy edge_id to new device array
+  cugraph_type_erased_device_array_t* new_device_edge_id_ptr{nullptr};
+
+  if (edge_id != NULL) {
+    error_code = cugraph_type_erased_device_array_create_from_view(
+      handle, edge_id, &new_device_edge_id_ptr, error);
+    if (error_code != CUGRAPH_SUCCESS) return error_code;
+  }
 
-  device_array_unique_ptr_t new_device_weights(new_device_weights_ptr,
+  device_array_unique_ptr_t new_device_edge_id(new_device_edge_id_ptr,
                                                &cugraph_type_erased_device_array_free);
 
-  // copy counts to new device array
-  cugraph_type_erased_device_array_t* new_device_counts_ptr{nullptr};
-  error_code = cugraph_type_erased_device_array_create_from_view(
-    handle, counts, &new_device_counts_ptr, error);
+  // copy edge_type to new device array
+  cugraph_type_erased_device_array_t* new_device_edge_type_ptr{nullptr};
+
+  if (edge_type != NULL) {
+    error_code = cugraph_type_erased_device_array_create_from_view(
+      handle, edge_type, &new_device_edge_type_ptr, error);
+    if (error_code != CUGRAPH_SUCCESS) return error_code;
+  }
+
+  device_array_unique_ptr_t new_device_edge_type(new_device_edge_type_ptr,
+                                                 &cugraph_type_erased_device_array_free);
+
+  // copy wgt to new device array
+  cugraph_type_erased_device_array_t* new_device_wgt_ptr{nullptr};
+  if (wgt != NULL) {
+    error_code =
+      cugraph_type_erased_device_array_create_from_view(handle, wgt, &new_device_wgt_ptr, error);
+    if (error_code != CUGRAPH_SUCCESS) return error_code;
+  }
+
+  device_array_unique_ptr_t new_device_wgt(new_device_wgt_ptr,
+                                           &cugraph_type_erased_device_array_free);
+
+  // copy hop to new device array
+  cugraph_type_erased_device_array_t* new_device_hop_ptr{nullptr};
+  error_code =
+    cugraph_type_erased_device_array_create_from_view(handle, hop, &new_device_hop_ptr, error);
   if (error_code != CUGRAPH_SUCCESS) return error_code;
 
-  device_array_unique_ptr_t new_device_counts(new_device_counts_ptr,
-                                              &cugraph_type_erased_device_array_free);
+  device_array_unique_ptr_t new_device_hop(new_device_hop_ptr,
+                                           &cugraph_type_erased_device_array_free);
+
+  // copy label to new device array
+  cugraph_type_erased_device_array_t* new_device_label_ptr{nullptr};
+
+  if (label != NULL) {
+    error_code = cugraph_type_erased_device_array_create_from_view(
+      handle, label, &new_device_label_ptr, error);
+    if (error_code != CUGRAPH_SUCCESS) return error_code;
+  }
+
+  device_array_unique_ptr_t new_device_label(new_device_label_ptr,
+                                             &cugraph_type_erased_device_array_free);
 
   // create new cugraph_sample_result_t
   *result = reinterpret_cast<cugraph_sample_result_t*>(new cugraph::c_api::cugraph_sample_result_t{
     reinterpret_cast<cugraph::c_api::cugraph_type_erased_device_array_t*>(
       new_device_srcs.release()),
     reinterpret_cast<cugraph::c_api::cugraph_type_erased_device_array_t*>(
       new_device_dsts.release()),
-    nullptr,
     reinterpret_cast<cugraph::c_api::cugraph_type_erased_device_array_t*>(
-      new_device_weights.release()),
-    nullptr,
+      new_device_edge_id.release()),
     reinterpret_cast<cugraph::c_api::cugraph_type_erased_device_array_t*>(
-      new_device_counts.release())});
+      new_device_edge_type.release()),
+    reinterpret_cast<cugraph::c_api::cugraph_type_erased_device_array_t*>(new_device_wgt.release()),
+    reinterpret_cast<cugraph::c_api::cugraph_type_erased_device_array_t*>(
+      new_device_label.release()),
+    reinterpret_cast<cugraph::c_api::cugraph_type_erased_device_array_t*>(new_device_hop.release()),
+    nullptr});
 
   return CUGRAPH_SUCCESS;
 }

cpp/src/community/legacy/spectral_clustering.cu

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020-2022, NVIDIA CORPORATION.
+ * Copyright (c) 2020-2023, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -24,7 +24,9 @@
 #include <cugraph/legacy/graph.hpp>
 #include <cugraph/utilities/error.hpp>
 
+#if defined RAFT_DISTANCE_COMPILED
 #include <raft/distance/specializations.cuh>
+#endif
 #include <raft/spectral/modularity_maximization.cuh>
 #include <raft/spectral/partition.cuh>
 

python/pylibcugraph/CMakeLists.txt

@@ -1,5 +1,5 @@
 # =============================================================================
-# Copyright (c) 2022, NVIDIA CORPORATION.
+# Copyright (c) 2022-2023, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except
 # in compliance with the License. You may obtain a copy of the License at

python/pylibcugraph/pylibcugraph/CMakeLists.txt

@@ -13,7 +13,6 @@
 # =============================================================================
 
 add_subdirectory(components)
-add_subdirectory(raft/common)
 add_subdirectory(internal_types)
 add_subdirectory(testing)
 

python/pylibcugraph/pylibcugraph/_cugraph_c/algorithms.pxd

@@ -218,8 +218,11 @@ cdef extern from "cugraph_c/algorithms.h":
             const cugraph_resource_handle_t* handle,
             const cugraph_type_erased_device_array_view_t* srcs,
             const cugraph_type_erased_device_array_view_t* dsts,
-            const cugraph_type_erased_device_array_view_t* weights,
-            const cugraph_type_erased_device_array_view_t* counts,
+            const cugraph_type_erased_device_array_view_t* edge_id,
+            const cugraph_type_erased_device_array_view_t* edge_type,
+            const cugraph_type_erased_device_array_view_t* wgt,
+            const cugraph_type_erased_device_array_view_t* hop,
+            const cugraph_type_erased_device_array_view_t* label,
             cugraph_sample_result_t** result,
             cugraph_error_t** error
         )