HDF5 Version: 1.14.3 (develop branch as of this commit)
OS: Ubuntu 22.04
When using the Cache VOL without the async connector under it, writing to datasets with variable-length data leads to a segfault during cleanup of HDF5 objects.
This issue will occur roughly half the time when running test_dataset_vlen_io in hdf5/test/API/H5_api_dataset_test.c. The system environment is configured as in this github workflow in the case without the async connector.
I've tried refactoring the test to close all HDF5 objects before moving onto the next write, but the segfault still occurs when the function exits.
HDF5 Version: 1.14.3 (develop branch as of this commit)
OS: Ubuntu 22.04
When using the Cache VOL without the async connector under it, writing to datasets with variable-length data leads to a segfault during cleanup of HDF5 objects.
This issue will occur roughly half the time when running
test_dataset_vlen_ioin hdf5/test/API/H5_api_dataset_test.c. The system environment is configured as in this github workflow in the case without the async connector.I've tried refactoring the test to close all HDF5 objects before moving onto the next write, but the segfault still occurs when the function exits.