BUG: test_table_method_rolling_methods segfault on Mac

[jbrockmendel]

i think it isnt getting run because it is "slow"

works fine on Ubuntu.

Might be a 3.9 thing; until fairly recently i couldnt install numba on 3.9

cc @mroeschke

[mroeschke]

Hmm this when you run it locally?

(pandas-dev-39) me % pytest pandas/tests/window/test_numba.py::TestTableMethod::test_table_method_rolling_methods
============================================================================ test session starts ============================================================================
platform darwin -- Python 3.9.2, pytest-6.2.2, py-1.10.0, pluggy-0.13.1
rootdir: /Users/me, configfile: setup.cfg
plugins: cov-2.11.1, hypothesis-6.8.1, xdist-2.2.1, asyncio-0.14.0, instafail-0.4.1, forked-1.3.0
collected 160 items

pandas/tests/window/test_numba.py ................................................................................................................................... [ 81%]
.............................                                                                                                                                         [100%]

=========================================================================== 160 passed in 27.00s ============================================================================

[jbrockmendel]

Hmm this when you run it locally?

Yes. Tried updating versions to better match what you listed and still segfaulted.

platform darwin -- Python 3.9.2, pytest-6.2.2, py-1.10.0, pluggy-0.13.1
rootdir: ~/pd/pandas, configfile: setup.cfg
plugins: vcr-1.0.2, cov-2.11.1, hypothesis-6.8.1, xdist-2.2.1, asyncio-0.14.0, repeat-0.9.1, forked-1.3.0

[mroeschke]

Here's my show versions if it helps spot any other differences in setup

INSTALLED VERSIONS
------------------
commit           : 7af47c999dc3d2782caf63fe8c42f0858937867b
python           : 3.9.2.final.0
python-bits      : 64
OS               : Darwin
OS-release       : 19.6.0
Version          : Darwin Kernel Version 19.6.0: Tue Nov 10 00:10:30 PST 2020; root:xnu-6153.141.10~1/RELEASE_X86_64
machine          : x86_64
processor        : i386
byteorder        : little
LC_ALL           : None
LANG             : en_US.UTF-8
LOCALE           : en_US.UTF-8

pandas           : 1.3.0.dev0+1065.g7af47c999d.dirty
numpy            : 1.20.1
pytz             : 2021.1
dateutil         : 2.8.1
pip              : 21.0.1
setuptools       : 49.6.0.post20210108
Cython           : 0.29.22
pytest           : 6.2.2
hypothesis       : 6.8.1
sphinx           : None
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : None
html5lib         : None
pymysql          : None
psycopg2         : None
jinja2           : None
IPython          : None
pandas_datareader: None
bs4              : None
bottleneck       : None
fsspec           : None
fastparquet      : None
gcsfs            : None
matplotlib       : None
numexpr          : None
odfpy            : None
openpyxl         : None
pandas_gbq       : None
pyarrow          : None
pyxlsb           : None
s3fs             : None
scipy            : None
sqlalchemy       : None
tables           : None
tabulate         : None
xarray           : None
xlrd             : None
xlwt             : None
numba            : 0.53.0

[jbrockmendel]

@WillAyd can you reproduce this? lldb isnt giving me anything useful

[WillAyd]

I unfortunately could not reproduce a segfault. Assuming you went through this guide:

https://pandas.pydata.org/pandas-docs/stable/development/debugging_extensions.html

Just wanted to double check you did a full clean of the extensions and then compiled --with-debugging-symbols before trying to run pytest through lldb, right?

[jbrockmendel]

i hadnt done the debugging symbols part until just now, no evident change. ive opened an issue about this on the numba tracker

[WillAyd]

Out of the two debuggers gcc I’ve personally found to be a little better at providing a useful stack trace on segfaults. If you have time and interest could also try that tool chain instead on macOS, assuming it still generates the fault

…

Sent from my iPad

On Mar 18, 2021, at 7:27 PM, jbrockmendel ***@***.***> wrote:  i hadnt done the debugging symbols part until just now, no evident change. ive opened an issue about this on the numba tracker — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[jbrockmendel]

Did some digging w/r/t to which fixture combinations cause the segfault. It occurs if and only if axis == 0 and not nogil and parallel and nopython and arithmetic_numba_supported_operators[0] in ["sum", "mean", "median", "max", "min"]

[WillAyd]

Is there a shareable core dump or is it too large? Alternately what backtrace are you getting on the segfault?

…

Sent from my iPhone

On Dec 2, 2021, at 5:53 PM, jbrockmendel ***@***.***> wrote:  Did some digging w/r/t to which fixture combinations cause the segfault. It occurs if and only if axis == 0 and not nogil and parallel and nopython and arithmetic_numba_supported_operators[0] in ["sum", "mean", "median", "max", "min"] — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.