[ https://issues.apache.org/jira/browse/ARROW-4746?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782682#comment-16782682 ]
Uwe L. Korn commented on ARROW-4746: ------------------------------------ [~pitrou] When you have a chance to look at this, it would be nice to know why CPython can deal with this. > [C++/Python] PyDataTime_Date wrongly casted to PyDataTime_DateTime > ------------------------------------------------------------------ > > Key: ARROW-4746 > URL: https://issues.apache.org/jira/browse/ARROW-4746 > Project: Apache Arrow > Issue Type: Bug > Components: C++, Python > Reporter: Uwe L. Korn > Priority: Major > Labels: pypy > > As mentioned in > https://bitbucket.org/pypy/pypy/issues/2842/running-pyarrow-on-pypy-segfaults#comment-50670536, > we currently access a {{PyDataTime_Date}} object with a > {{PyDataTime_DateTime}} cast in {{PyDateTime_DATE_GET_SECOND}} in our code in > two instances. While CPython is able to deal with this wrong usage, PyPy is > not able to do so. We should separate the path here into one that deals with > dates and another that deals with datetimes. > Reproducible code: > {code:java} > pa.array([datetime.date(2018, 5, 10)], type=pa.date64()){code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)