SciPy-bundle-2023.07-gfbf-2023a.eb fails to build (self test crashes)

[schiotz]

I am trying to build SciPy-bundle-2023.07-gfbf-2023a.eb, it works on the Intel Cascade_Lake Xeon_Gold_6242R but fails on the IceLake Xeon_Gold_6348 with this error in the self-test:

scipy/integrate/tests/test__quad_vec.py::test_quad_vec_pool_args[1-extra_args1] PASSED                                           [ 11%]
scipy/integrate/tests/test__quad_vec.py::test_quad_vec_pool_args[10-2] Fatal Python error: Aborted

[boegel]

Hmm, this seems similar to the hard crash in #17539, but it's not a segmentation fault.

Maybe the test suite is consuming a significant amount of memory, and that's triggering the kernel to abort the process because it's running in a Slurm job where only a specific amount of memory is available?

[schiotz]

I am compiling on a login node, where memory is limited to 50GB per process. Total memory is 500 GB. I really hope that is not too little to build SciPy-bundle. :-)

[schiotz]

Follow-up: It could very well be related to ressources - building on a compute node in a slurm job worked. Thank you very much, indeed, for pointing me in the right direction.