vector complex test failed on riscv64

[yuzibo]

Describe the issue or bug

I noticed some test cases failed on riscv64:

=========================== short test summary info ============================
FAILED rpy2/tests/rinterface/test_na.py::test_R_to_NAComplex - assert False
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_init_from_seqr - as...
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_getitem - assert (1...
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_setitem - assert (3...
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_getslice - assert (...
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_getslice_negative
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_setslice - assert (...
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_setslice_negative
FAILED rpy2/tests/rinterface/test_vector_complex.py::test_index - ValueError:...
FAILED rpy2/tests/robjects/test_conversion_numpy.py::TestNumpyConversions::test_vector_complex

To Reproduce
To see the log

Additional context
At first I thought there is no vector support on Debian for riscv64 now lead to the error. But from the #1050432 I think we have something are missing.

Could you have a look at this?

Thanks.

[lgautier]

If I understand the thread in the link provided, the issue might be on the Python / cffi library for Python or numpy library for Python.

(@eddelbuettel in case he knows more about this)

[eddelbuettel]

Could well be. I don't know anything else either -- as rpy2 maintainer for Debian this was just pointed out to me so I came to Laurent for help :) Maybe we just want to skip that test for now on that platform, and talk to the cffi maintainer and / or upstream?

PS Sorry I even confused this with a different (newer) bug report I recently forwarded: #1071362. Laurent, did you get the email about that or did the spam filters or gremlins eat it?

[yuzibo]

Maybe we just want to skip that test for now on that platform, and talk to the cffi maintainer and / or upstream?

Thanks. I will have a look at cffi library side. I had one patch based on skipping these test cases on riscv64, but before sending the patch, I think it would be better to let upstream know. I will send the debdiff if this is okay.

PS Sorry I even confused this with a different (newer) bug report I recently forwarded: #1071362. Laurent, did you get the email about that or did the spam filters or gremlins eat it?

hmm, I proposed the patch when I did not find the issue about #1071362. I will verity it again also.

[lgautier]

Could well be. I don't know anything else either -- as rpy2 maintainer for Debian this was just pointed out to me so I came to Laurent for help :) Maybe we just want to skip that test for now on that platform, and talk to the cffi maintainer and / or upstream?

PS Sorry I even confused this with a different (newer) bug report I recently forwarded: #1071362. Laurent, did you get the email about that or did the spam filters or gremlins eat it?

IRL time gremlins probably pilfered it. I just had a look at it and I can't reproduce the segfault with R 4.4.0.

However, I am also noting in the link you provide that Python 3.12 is used. I suspect that this is more Python 3.12-related than R 4.4.0-related. The CI test matrix segfaults with Python 3.12. Unfortunately I haven't had a chance to look at it yet.

[eddelbuettel]

Ah. Could well be. As I commented in the bug-system thread (ie not here but over at the link) it is unlikely that a change in R or rpy2 triggered that. I didn't think of Python 3.12 😜 In any event I have unplugged this for now in a brute-force way (but not running the test) to unplug the package. We'll see about 3.12 in due course.