Build issues with latest cuGraph main

[sg0]

I am using 522055b and trying to build libcugraph on a cluster using instructions in README (https://github.com/rapidsai/cugraph/tree/main/cpp/tests) and encountering issues during configuration/build.

My environment:

Currently Loaded Modulefiles:
  1) gcc/10.3.0               2) cuda/12.1                3) openmpi/4.1.4            4) cmake/3.28.1             5) python/miniconda24.4.0

>> ./build.sh libcugraph cpp-mgtests 
 
 Cloning into 'cugraph-ops-src'...
 Permission denied (publickey).
 fatal: Could not read from remote repository.

 Please make sure you have the correct access rights
 and the repository exists.
 Cloning into 'cugraph-ops-src'...
 Permission denied (publickey).
 fatal: Could not read from remote repository.

 Please make sure you have the correct access rights
 and the repository exists.
 Cloning into 'cugraph-ops-src'...
 Permission denied (publickey).
 fatal: Could not read from remote repository.

 Please make sure you have the correct access rights
 and the repository exists.
 -- Had to git clone more than once: 3 times.
 CMake Error at cugraph-ops-subbuild/cugraph-ops-populate-prefix/tmp/cugraph-ops-populate-gitclone.cmake:39 (message):
   Failed to clone repository: '[email protected]:rapidsai/cugraph-ops.git'

I tried to exclude cugraphops, then configuration passes, but fails during building.

>> ./build.sh --without_cugraphops libcugraph cpp-mgtests
  /people/ghos167/sources/cugraph-latest/cpp/src/prims/detail/sample_and_compute_local_nbr_indices.cuh(1681): error: identifier "edge_t" is undefined
        rmm::device_uvector<edge_t>(0, handle.get_stream()), std::nullopt, std::vector<size_t>());
                            ^

  1 error detected in the compilation of "/people/ghos167/sources/cugraph-latest/cpp/src/sampling/random_walks_mg_v64_e64.cu".

[ChuckHastings]

@seunghwak - Looks like PR #4443 introduced a compilation error with NO_CUGRAPH_OPS. The definition of edge_t is inside the #ifdef block so can't be used in the return.

@sg0 - as a workaround, since the line before this is a CUGRAPH_FAIL, which throws an exception, I suspect you can just comment out the two-line call to std::make_tuple.

[sg0]

I tried with the suggested modification, but the build still fails.

 [ghos167@deception04 cugraph-latest]$ git diff
 diff --git a/cpp/src/prims/detail/sample_and_compute_local_nbr_indices.cuh b/cpp/src/prims/detail/sample_and_compute_local_nbr_indices.cuh
 index 64b6aab..75a4671 100644
 --- a/cpp/src/prims/detail/sample_and_compute_local_nbr_indices.cuh
 +++ b/cpp/src/prims/detail/sample_and_compute_local_nbr_indices.cuh
 @@ -1677,8 +1677,8 @@ uniform_sample_and_compute_local_nbr_indices(
      std::move(local_nbr_indices), std::move(key_indices), std::move(local_frontier_sample_offsets));
  #else
    CUGRAPH_FAIL("unimplemented.");
 -  return std::make_tuple(
 -    rmm::device_uvector<edge_t>(0, handle.get_stream()), std::nullopt, std::vector<size_t>());
 +  // return std::make_tuple(
 +  //  rmm::device_uvector<edge_t>(0, handle.get_stream()), std::nullopt, std::vector<size_t>());
  #endif
  }

[347/1050] Building CXX object CMakeFiles/cugraph_c.dir/src/c_api/pagerank.cpp.o
FAILED: CMakeFiles/cugraph_c.dir/src/c_api/pagerank.cpp.o
/share/apps/gcc/10.3.0/bin/g++ -DCUDA_API_PER_THREAD_DEFAULT_STREAM -DCUTLASS_ENABLE_CUDNN=1 -DCUTLASS_NAMESPACE=raft_cutlass -DFMT_HEADER_ONLY=1 -DLIBCUDACXX_ENABLE_EXPERIMENTAL_MEMORY_RESOURCE -DRAFT_SYSTEM_LITTLE_ENDIAN=1 -DSPDLOG_FMT_EXTERNAL -DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_CUDA -DTHRUST_DISABLE_ABI_NAMESPACE -DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -DTHRUST_IGNORE_ABI_NAMESPACE_ERROR -Dcugraph_c_EXPORTS -I/people/ghos167/sources/cugraph-latest/cpp/src -I/people/ghos167/sources/cugraph-latest/cpp/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/rmm-src/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/cccl-src/thrust/thrust/cmake/../.. -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/cccl-src/libcudacxx/lib/cmake/libcudacxx/../../../include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/cccl-src/cub/cub/cmake/../.. -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/fmt-src/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/spdlog-src/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/nvtx3-src/c/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/cuco-src/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/nvidiacutlass-src/include -I/people/ghos167/sources/cugraph-latest/cpp/build/_deps/nvidiacutlass-build/include -I/share/apps/cuda/12.3/include -isystem /share/apps/cuda/12.3/targets/x86_64-linux/include -O3 -DNDEBUG -std=gnu++17 -fPIC -Werror -Wno-error=deprecated-declarations -Wno-deprecated-declarations -DRAFT_HIDE_DEPRECATION_WARNINGS -DNO_CUGRAPH_OPS -MD -MT CMakeFiles/cugraph_c.dir/src/c_api/pagerank.cpp.o -MF CMakeFiles/cugraph_c.dir/src/c_api/pagerank.cpp.o.d -o CMakeFiles/cugraph_c.dir/src/c_api/pagerank.cpp.o -c /people/ghos167/sources/cugraph-latest/cpp/src/c_api/pagerank.cpp
In file included from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/fmt-src/include/fmt/core.h:2920,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/spdlog-src/include/spdlog/fmt/fmt.h:31,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/spdlog-src/include/spdlog/common.h:50,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include/raft/core/detail/callback_sink.hpp:22,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include/raft/core/logger-inl.hpp:34,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include/raft/core/logger.hpp:22,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include/raft/core/resources.hpp:21,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include/raft/core/resource/comms.hpp:20,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include/raft/core/device_resources.hpp:23,
                 from /people/ghos167/sources/cugraph-latest/cpp/build/_deps/raft-src/cpp/include/raft/core/handle.hpp:19,
                 from /people/ghos167/sources/cugraph-latest/cpp/include/cugraph/utilities/dataframe_buffer.hpp:20,
                 from /people/ghos167/sources/cugraph-latest/cpp/include/cugraph/edge_property.hpp:19,
                 from /people/ghos167/sources/cugraph-latest/cpp/include/cugraph/graph_view.hpp:20,
                 from /people/ghos167/sources/cugraph-latest/cpp/include/cugraph/graph.hpp:18,
                 from /people/ghos167/sources/cugraph-latest/cpp/src/c_api/graph.hpp:23,
                 from /people/ghos167/sources/cugraph-latest/cpp/src/c_api/pagerank.cpp:19:
In function ‘fmt::v10::detail::format_decimal_result<Char*> fmt::v10::detail::format_decimal(Char*, UInt, int) [with Char = char; UInt = unsigned int]’,
    inlined from ‘constexpr fmt::v10::detail::format_decimal_result<Iterator> fmt::v10::detail::format_decimal(Iterator, UInt, int) [with Char = char; UInt = unsigned int; Iterator = fmt::v10::appender; typename std::enable_if<(! std::is_pointer<typename std::remove_cv<typename std::remove_reference<_Arg>::type>::type>::value), int>::type <anonymous> = 0]’ at /people/ghos167/sources/cugraph-latest/cpp/build/_deps/fmt-src/include/fmt/format.h:1371:28,
    inlined from ‘void fmt::v10::detail::format_hexfloat(Float, int, fmt::v10::detail::float_specs, fmt::v10::detail::buffer<char>&) [with Float = double; typename std::enable_if<(! std::integral_constant<bool, (std::numeric_limits<_Tp>::digits == 106)>::value), int>::type <anonymous> = 0]’ at /people/ghos167/sources/cugraph-latest/cpp/build/_deps/fmt-src/include/fmt/format.h:3270:23:
/people/ghos167/sources/cugraph-latest/cpp/build/_deps/fmt-src/include/fmt/format.h:1327:11: error: writing 2 bytes into a region of size 0 [-Werror=stringop-overflow=]
 1327 |     memcpy(dst, src, 2);
      |     ~~~~~~^~~~~~~~~~~~~
/people/ghos167/sources/cugraph-latest/cpp/build/_deps/fmt-src/include/fmt/format.h: In function ‘void fmt::v10::detail::format_hexfloat(Float, int, fmt::v10::detail::float_specs, fmt::v10::detail::buffer<char>&) [with Float = double; typename std::enable_if<(! std::integral_constant<bool, (std::numeric_limits<_Tp>::digits == 106)>::value), int>::type <anonymous> = 0]’:
/people/ghos167/sources/cugraph-latest/cpp/build/_deps/fmt-src/include/fmt/format.h:1370:8: note: at offset -2 to object ‘buffer’ with size 10 declared here
 1370 |   Char buffer[digits10<UInt>() + 1] = {};
      |        ^~~~~~
cc1plus: all warnings being treated as errors
[378/1050] Building CUDA object CMakeFiles/cugraph.dir/src/structure/create_graph_from_edgelist_mg_v32_e64.cu.o
ninja: build stopped: subcommand failed.

[ChuckHastings]

We believe these errors are due to your build environment. Our build documentation is out-of-date. All of RAPIDS now requires gcc 11.4 or greater, and nvcc 11.8 or greater. We are using compiler features that are not supported in the older compilers.

I have created #4564 to address the documentation issue.

[ChuckHastings]

This got closed by the PR that fixed the bug that your test revealed.

If you continue to have issues building from source after upgrading to a newer set of compilers (gcc 11.4+ and nvcc 11.8+), please open a new issue and we will investigate further.