Skip to content

Comments on wheel building script #3

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from

Conversation

matthew-brett
Copy link

I think you actively don't want the pandas checkout to go into the host operating system, because you want the checkout to be a temporary copy that you will throw away with the building image.

The --no-deps flag avoids copying dependencies into the wheelhouse, such as numpy.

I think you actively don't want the pandas checkout to go into the host operating system, because you want the checkout to be a temporary copy that you will throw away with the building image.

The `--no-deps` flag avoids copying dependencies into the wheelhouse, such as numpy.
@TomAugspurger
Copy link
Owner

Hi Matthew, thanks for looking. --no-deps does help.

I've been tinkering with it a bit more in #2 and was able to get Travis to build and run the tests in a container without timing out. Would you be open to those uploading to wheels.scipy.org at the end of the build similar to https://github.com/MacPython/numpy-wheels? And then we can upload those to PyPI for each release?

@matthew-brett
Copy link
Author

Sure - no problem - I'll add the encrypted key to this PR.

@matthew-brett
Copy link
Author

Which repo will this be building from - TomAugspurger/pandas-manylinux ?

@TomAugspurger
Copy link
Owner

Ahh, I think I'll do it from pydata/pandas-manylinux. I'll create that quick.

@TomAugspurger
Copy link
Owner

Whoops, I don't have permission to create a new repo under pydata. @jreback I'm guessing you do? I think it make sense to have this live there, right? Could you make a blank pydata/pandas-manylinux repo and give the pandas devs access to it?

@jreback
Copy link

jreback commented Jun 11, 2016

https://github.com/pydata/pandas-manylinux

should be good

This should upload the wheels when build triggered on that repo (and
only that repo).
@matthew-brett
Copy link
Author

Sorry to be a bit NIH - but I was working on a scripting system for combined OSX / Manylinux wheels over at https://github.com/matthew-brett/multibuild - and it turned out to be pretty easy to update the current Pandas wheel-building repo to build 64- and 32-bit manylinux wheels :

There are some 32-bit test failures (see pandas-dev/pandas#13566), but the 64-bit wheels test OK, and are now up at http://wheels.scipy.org/ . See also updated info at https://github.com/MacPython/pandas-wheels .

Would you consider uploading these? Numpy / scipy / Cython already have Manylinux wheels, matplotlib and h5py wheels are imminent. Automated upload instructions at https://github.com/MacPython/pandas-wheels#uploading-the-built-wheels-to-pypi .

@TomAugspurger
Copy link
Owner

Saw your post yesterday on the manylinux ML, you beat me to it :) Thanks for solving this in a general way.

I think Jeff has the permissions to upload to PyPI.

@matthew-brett
Copy link
Author

@jreback - anything I can do to help here? Should I move the discussion over the pandas mailing list or issues?

@matthew-brett
Copy link
Author

@jreback - should I email the Pandas list? Or open a Pandas issue about this?

@stefanv
Copy link

stefanv commented Jul 19, 2016

@matthew-brett @jreback We'd very much appreciate if Pandas wheels can appear on PyPi. It's slowing down our test suites significantly to have to rebuild from source each time.

@jreback
Copy link

jreback commented Jul 19, 2016

@matthew-brett I just saw your comment.

what is the difference between m & mu? (need both)?

pandas-0.18.1-cp27-cp27m-macosx_10_6_intel.macosx_10_9_intel.macosx_10_9_x86_64.macosx_10_10_intel.macosx_10_10_x86_64.whl  10Mi    2016-07-05 12:36:32
pandas-0.18.1-cp27-cp27m-manylinux1_x86_64.whl  14Mi    2016-07-05 12:18:17
pandas-0.18.1-cp27-cp27mu-manylinux1_x86_64.whl 14Mi    2016-07-05 12:19:26
pandas-0.18.1-cp34-cp34m-macosx_10_6_intel.macosx_10_9_intel.macosx_10_9_x86_64.macosx_10_10_intel.macosx_10_10_x86_64.whl  9Mi 2016-07-05 12:36:57
pandas-0.18.1-cp34-cp34m-manylinux1_i686.whl    14Mi    2016-07-05 12:30:47
pandas-0.18.1-cp34-cp34m-manylinux1_x86_64.whl  15Mi    2016-07-05 12:18:24
pandas-0.18.1-cp35-cp35m-macosx_10_6_intel.macosx_10_9_intel.macosx_10_9_x86_64.macosx_10_10_intel.macosx_10_10_x86_64.whl  9Mi 2016-07-05 12:46:37
pandas-0.18.1-cp35-cp35m-manylinux1_i686.whl    14Mi    2016-07-05 12:30:30
pandas-0.18.1-cp35-cp35m-manylinux1_x86_64.whl  14Mi    

@jreback
Copy link

jreback commented Jul 19, 2016

ok they are all uploaded.

@matthew-brett
Copy link
Author

m is for Python with a narrow unicode build (16-bit unicode). These are a bit uncommon, but old pyenv builds were narrow, and in fact travis-ci trusty containers use older pyenv builds. mu is for the more standard 32-bit unicode build that most (?all) distributions ship. This distinction disappears with the new unicode representation in Python 3.3.

@matthew-brett
Copy link
Author

Great - thanks for uploading them. Please let me know if there's anything I can do to help with the next release. I think that it should be as simple as the instructions at https://github.com/macpython/pandas-wheels#triggering-a-build

@stefanv
Copy link

stefanv commented Jul 20, 2016 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants