[
https://issues.apache.org/jira/browse/ARROW-4747?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16789661#comment-16789661
]
Uwe L. Korn commented on ARROW-4747:
------------------------------------
Have a look at
[https://github.com/apache/arrow/blob/fd5e22158a7bf9f717db78b45579bc611374b007/docker-compose.yml#L146]
and
[https://github.com/apache/arrow/blob/fd5e22158a7bf9f717db78b45579bc611374b007/cpp/Dockerfile.ubuntu-xenial]
These are base dockerfiles that build against the system libraries. We would
need to additional flags to the CMake command to redirect it to using PyPy
nightlies instead of the system Python. Additionally, we also need to download
the PyPy nightly first in the Dockerfile.
Note that these files stem from the PR
[https://github.com/apache/arrow/pull/3688] and may not yet work on master.
> [C++/PyPy] Add docker image to test against PyPy nightlies
> ----------------------------------------------------------
>
> Key: ARROW-4747
> URL: https://issues.apache.org/jira/browse/ARROW-4747
> Project: Apache Arrow
> Issue Type: Task
> Components: C++, Python
> Reporter: Uwe L. Korn
> Priority: Major
> Labels: pypy
>
> It seems
> (https://bitbucket.org/pypy/pypy/issues/2842/running-pyarrow-on-pypy-segfaults#comment-50670536)
> that we are close to being able to run with PyPy. At the moment, we don't
> actively work on supporting PyPy but this would be a good start on providing
> feedback at what is still missing.
> To have such an image, one would need to fork one of the current test setups
> in {{docker-compose.yml}} that build with system libraries (e.g.
> ubuntu-xenial or debian-testing) and exchange the system Python with the PyPy
> nightly builds from http://buildbot.pypy.org/nightly/trunk/
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
