Describe the bug
Build: spark-rapids-jni_submodule-sync-dev/6257
The spark-rapids-jni C++ build fails after the submodule-sync bot updated the cudf submodule to c209410f85e0d34f151abf741b8ab505cd5d4321. Compilation of SparkResourceAdaptorJni.cpp fails because the header rmm/mr/device_memory_resource.hpp no longer exists in the new RMM install tree. The installed RMM headers list many rmm/mr/*.hpp files (cuda_memory_resource.hpp, pool_memory_resource.hpp, etc.) but NOT device_memory_resource.hpp, indicating an upstream RMM API refactor that broke source compatibility with spark-rapids-jni.
Error logs:
[10/160] Building CXX object CMakeFiles/spark_rapids_jni.dir/src/SparkResourceAdaptorJni.cpp.o
FAILED: CMakeFiles/spark_rapids_jni.dir/src/SparkResourceAdaptorJni.cpp.o
src/main/cpp/src/SparkResourceAdaptorJni.cpp:17:10: fatal error: rmm/mr/device_memory_resource.hpp: No such file or directory
17 | #include <rmm/mr/device_memory_resource.hpp>
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
ninja: build stopped: subcommand failed.
[ERROR] Failed to execute goal org.apache.maven.plugins:maven-antrun-plugin:3.0.0:run (buildcpp) on project spark-rapids-jni: An Ant BuildException has occured: exec returned: 1
unzip: cannot find or open target/spark-rapids-jni-26.06.0-SNAPSHOT-cuda12.jar
Environment details
- Submodule-sync updating cudf from 15411d2e to c209410f85e0d34f151abf741b8ab505cd5d4321
- RockyLinux 8 + CUDA 12.9.1 build image
- GCC toolset 14, nvcc 12.9.86
- This is a source-compatibility break; retrying will not help. The SparkResourceAdaptorJni.cpp (and possibly other files) needs to be updated to use the new RMM API (likely
rmm/resource_ref.hpp / rmm::device_async_resource_ref).
Describe the bug
Build: spark-rapids-jni_submodule-sync-dev/6257
The spark-rapids-jni C++ build fails after the submodule-sync bot updated the cudf submodule to c209410f85e0d34f151abf741b8ab505cd5d4321. Compilation of SparkResourceAdaptorJni.cpp fails because the header
rmm/mr/device_memory_resource.hppno longer exists in the new RMM install tree. The installed RMM headers list many rmm/mr/*.hpp files (cuda_memory_resource.hpp, pool_memory_resource.hpp, etc.) but NOT device_memory_resource.hpp, indicating an upstream RMM API refactor that broke source compatibility with spark-rapids-jni.Error logs:
Environment details
rmm/resource_ref.hpp/rmm::device_async_resource_ref).