Skip to content
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

cant install on python 3.8b1 on windows #13790

Closed
NewUserHa opened this issue Jun 15, 2019 · 4 comments · Fixed by #14453
Closed

cant install on python 3.8b1 on windows #13790

NewUserHa opened this issue Jun 15, 2019 · 4 comments · Fixed by #14453
Labels
57 - Close? Issues which may be closable unless discussion continued component: build

Comments

@NewUserHa
Copy link

pip install numpy

numpy\random\mtrand\mtrand.c(40364): error C2198: 'PyCode_New': too few arguments for call
@mattip
Copy link
Member

mattip commented Jun 15, 2019

Did you use cython 0.29.10 which should fix that?

@seberg seberg added the 57 - Close? Issues which may be closable unless discussion continued label Jun 18, 2019
@jdemeyer
Copy link
Contributor

Did you use cython 0.29.10 which should fix that?

Cython is run by the maintainer when packaging numpy, not when installing numpy. So the proper fix is making a new numpy release generated by the latest Cython.

There is some discussion happening at https://bugs.python.org/issue37221 and https://bugs.python.org/issue37250 which may lead to a fix in CPython such that older Cython-generated code will continue working with Python 3.8. I'm in favor of that, but it's not clear whether that will be accepted.

@NewUserHa
Copy link
Author

it is 0.29.10

@rgommers
Copy link
Member

Cython is run by the maintainer when packaging numpy, not when installing numpy. So the proper fix is making a new numpy release generated by the latest Cython.

This is happening a lot recently. I think the improvements to the packaging system make build-time requirements work in an acceptable fashion over the last year finally, so it may be time to consider shipping an sdist without generated sources. There are downsides too, but it's a discussion that would make sense to have.

flying-sheep added a commit to ivirshup/scanpy that referenced this issue Jun 27, 2019
rgommers added a commit to rgommers/numpy that referenced this issue Sep 8, 2019
Rationale: it has happened multiple times in the recent past that users
needed to use a more recent Cython than was used to generate the sdist
of the most recent release available on PyPI (e.g. for building with
a not-yet-released version of Python for which a Cython fix just
landed).  The dependency specification and packaging tools have matured
to the point where it should be fine to require Cython to be installed.
Also, we have wheels (and conda-forge packages) for Windows, Linux and
macOS, so this change won't be visible at all to the average user.

Closes numpygh-13790

Also adds pyproject.toml to MANIFEST.in, it was missing.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
57 - Close? Issues which may be closable unless discussion continued component: build
Projects
None yet
Development

Successfully merging a pull request may close this issue.

5 participants