You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@arrow.apache.org by "Benjamin Kietzman (JIRA)" <ji...@apache.org> on 2018/12/12 19:25:00 UTC

[jira] [Comment Edited] (ARROW-4000) [Python] Error running CSV test_read_options on Windows

    [ https://issues.apache.org/jira/browse/ARROW-4000?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16719334#comment-16719334 ] 

Benjamin Kietzman edited comment on ARROW-4000 at 12/12/18 7:24 PM:
--------------------------------------------------------------------

For the next person who comes along, to attach to the test before it starts: (from the VS developer command prompt)

{{python -c "import os; import subprocess; subprocess.call(['VsJITDebugger.exe', '-p', str(os.getpid())]); import pytest; pytest.main(['pyarrow/tests/test_csv.py', '-v'])"}}

I had to rebuild _csv.pyd with debug symbols before this worked. This can be done without switching to a full debug build by opening {{python\build\temp.win-amd64-3.6\Release\pyarrow.sln}} using VisualStudio, rebuilding in RelWithDebInfo configuration, then copying the new pyd to {{python\pyarrow}}. (I'm looking to see if there's a way to do this from the command line, but I haven't found one yet)

I got a closer look at the exception, which originates in {{__pyx_pf_7pyarrow_4_csv_11ReadOptions___init__}} (cython-generated c++ corresponding to [_csv.pyx:62|https://github.com/apache/arrow/blob/master/python/pyarrow/_csv.pyx#L62]):

{{Exception thrown: read access violation.**__imp_arrow::csv::ReadOptions::Defaults**(...) returned 0xFFFFFFFFFFFFFFFF. occurred}}

... I'm not a cython expert, but this doesn't make sense; ReadOptions is just a POD struct so I don't see how ReadOptions::Defaults() is generating a segfault.


was (Author: bkietz):
For the next person who comes along, to attach to the test before it starts: (from the VS developer command prompt)

{{ python -c "import os; import subprocess; subprocess.call(['VsJITDebugger.exe', '-p', str(os.getpid())]); import pytest; pytest.main(['pyarrow/tests/test_csv.py', '-v'])" }}

I had to rebuild _csv.pyd with debug symbols before this worked. This can be done without switching to a full debug build by opening {{python\build\temp.win-amd64-3.6\Release\pyarrow.sln}} using VisualStudio, rebuilding in RelWithDebInfo configuration, then copying the new pyd to {{python\pyarrow}}. (I'm looking to see if there's a way to do this from the command line, but I haven't found one yet)

I got a closer look at the exception, which originates in {{__pyx_pf_7pyarrow_4_csv_11ReadOptions___init__}} (cython-generated c++ corresponding to [_csv.pyx:62|https://github.com/apache/arrow/blob/master/python/pyarrow/_csv.pyx#L62]):

{{ Exception thrown: read access violation.**__imp_arrow::csv::ReadOptions::Defaults**(...) returned 0xFFFFFFFFFFFFFFFF. occurred }}

... I'm not a cython expert, but this doesn't make sense; ReadOptions is just a POD struct so I don't see how ReadOptions::Defaults() is generating a segfault.

> [Python] Error running CSV test_read_options on Windows
> -------------------------------------------------------
>
>                 Key: ARROW-4000
>                 URL: https://issues.apache.org/jira/browse/ARROW-4000
>             Project: Apache Arrow
>          Issue Type: Bug
>          Components: C++, Python
>    Affects Versions: 0.11.1
>            Reporter: Benjamin Kietzman
>            Priority: Minor
>              Labels: csv, windows
>
> `py.test pyarrow -v` crashed at `pyarrow/tests/test_csv.py::test_read_options`.
> errorlevel was -1073741819, not sure what that means.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)