Bug#1004870: python-xarray: autopkgtest regression on s390x

[Graham Inggs]

Hi Rajendra

On Sun, 6 Mar 2022 at 12:42, Rajendra Kharat1 <Rajendra...@ibm.com> wrote:
> Could you please share the steps to reproduce the issue.
> As when I am trying to reproduce I ran into different problem of segmentation fault.
> Below are the steps that I followed.
> 1. Install python3-xarray on debian on s390x arch.
> 2. cd /usr/lib/python3/dist-packages/xarray
> 3. python3 -m pytest tests/ -vvv
>
> For now I am debugging the segmentation fault, however just wanted to validate the steps to reproduce.

Here's a quick and dirty way to run autopkgtests locally:

sudo apt install autopkgtest
apt source python-xarray
sudo autopkgtest -B python-xarray-0.17.0/ -- null

Regards
Graham

[Alastair McKinstry]

reopen 1004870

thanks


I've just tested on zelenka with python-xarray & pandas 1.4.3;

FAILED
xarray/tests/test_accessor_dt.py::test_calendar_cftime_2D[365_day] -
pandas._libs.tslibs.np_datetime.OutOfBoundsDatetime: Out of bounds
nanosecond timestamp: -809793270280017-0...
FAILED
xarray/tests/test_accessor_dt.py::test_calendar_cftime_2D[360_day] -
pandas._libs.tslibs.np_datetime.OutOfBoundsDatetime: Out of bounds
nanosecond timestamp: -809793270280017-0...
FAILED xarray/tests/test_accessor_dt.py::test_calendar_cftime_2D[julian]
- pandas._libs.tslibs.np_datetime.OutOfBoundsDatetime: Out of bounds
nanosecond timestamp: -809793270280017-04...
FAILED
xarray/tests/test_accessor_dt.py::test_calendar_cftime_2D[all_leap] -
pandas._libs.tslibs.np_datetime.OutOfBoundsDatetime: Out of bounds
nanosecond timestamp: -809793270280017-...
FAILED
xarray/tests/test_accessor_dt.py::test_calendar_cftime_2D[366_day] -
pandas._libs.tslibs.np_datetime.OutOfBoundsDatetime: Out of bounds
nanosecond timestamp: -809793270280017-0...
FAILED
xarray/tests/test_accessor_dt.py::test_calendar_cftime_2D[gregorian] -
pandas._libs.tslibs.np_datetime.OutOfBoundsDatetime: Out of bounds
nanosecond timestamp: -809793270280017...
FAILED
xarray/tests/test_accessor_dt.py::test_calendar_cftime_2D[proleptic_gregorian]
- pandas._libs.tslibs.np_datetime.OutOfBoundsDatetime: Out of bounds
nanosecond timestamp: -80979...
FAILED xarray/tests/test_indexes.py::TestPandasIndex::test_join -
AssertionError: assert dtype('>U4') == '<U4'
FAILED xarray/tests/test_indexes.py::TestPandasMultiIndex::test_concat -
AssertionError: assert {'bar': dtype...type('int32')} == {'bar':
'<U1'...numpy.int32'>}
FAILED xarray/tests/test_indexes.py::TestPandasMultiIndex::test_join -
AssertionError: assert {'one': dtype...type('int32')} == {'one': '<U2',
'two': 'i'}


On 27/08/2022 08:21, Rebecca N. Palmer wrote:
> pandas 1.4.3 in experimental has now been built for s390x - can
> someone try the python-xarray autopkgtest with that (i.e. test whether
> that actually fixes this bug)?
>
--
Alastair McKinstry,
GPG: 82383CE9165B347C787081A2CBE6BB4E5D9AD3A5
ph: +353 87 6847928 e: alas...@sceal.ie, im: @sceal.ie:mckinstry

[Graham Inggs]

Hi Rebecca, Alastair

I triggered python-xarray's autopkgtest on s390x against pandas from
experimental [1]. See the test results from 2022-08-29 12:40:12 UTC.

I think the intended packages were tested together, based on seeing
the following files fetched in the log:

Get:28 http://deb.debian.org/debian experimental/main s390x
python3-pandas-lib s390x 1.4.3+dfsg-6 [2,852 kB]
Get:29 http://deb.debian.org/debian experimental/main s390x
python3-pandas all 1.4.3+dfsg-6 [2,785 kB]

I still see the same failures that were in the original bug report.

Even if this issue eventually is fixed in pandas, I'd prefer this bug
be assigned back to src:python-xarray in order to track the
autopkgtest regression in testing.

Regards
Graham


[1] https://ci.debian.net/packages/p/python-xarray/unstable/s390x/

[Graham Inggs]

Control: reassign -1 src:python-xarray 0.21.0-1

Hopefully assigning back to src:python-xarray now.
If you have reason to believe this issue is caused by a bug in pandas,
then please file a bug against pandas with the relevant details and
mark that it blocks this bug.