bdist_wheel misbehaves on Windows newlines

[boppreh]

Environment

Python 3.6 64 bits, Windows 10.

>>> setuptools.__version__
'36.2.6'

Steps to reproduce

setup.py

from setuptools import setup
setup(
    name='package',
    version='0.0.0',
    long_description='first line\r\nsecond line',
)

$ python setup.py bdist_wheel
running bdist_wheel
running build
installing to build\bdist.win-amd64\wheel
running install
running install_egg_info
running egg_info
writing package.egg-info\PKG-INFO
writing dependency_links to package.egg-info\dependency_links.txt
writing top-level names to package.egg-info\top_level.txt
reading manifest file 'package.egg-info\SOURCES.txt'
writing manifest file 'package.egg-info\SOURCES.txt'
Copying package.egg-info to build\bdist.win-amd64\wheel\.\package-0.0.0-py3.6.egg-info
running install_scripts
creating build\bdist.win-amd64\wheel\package-0.0.0.dist-info\WHEEL

Observed behavior

dist\package-0.0.0-py3-none-any.whl\package-0.0.0.dist-info\METADATA

Metadata-Version: 2.0
Name: package
Version: 0.0.0
Summary: UNKNOWN
Home-page: UNKNOWN
Author: UNKNOWN
Author-email: UNKNOWN
License: UNKNOWN

first line

Expected behavior

METADATA should include "second line", or an error raised during build.

Notes

python setup.py sdist generates a file with both lines correctly in the description field of PKG-INFO, but they maintain their '\r\n' line endings while the rest of the file uses only \n. This is very unusual and should probably be changed or warned against.

I'm not sure if this issue should be reported here or in https://github.com/pypa/wheel, so apologies if I'm in the wrong place.

I discovered this on msabramo/setuptools-markdown#11.

[maphew]

I have a similar problem, but only in py3. When I build with py27 the complete string is included. See JessicaTegner/pypandoc#147

[jaraco]

I worked on this a bit today, but ended up running up against #1196 and pypa/wheel#206.

The thing I'm after - where are the lines being stripped? It should be possible by tracing a dist_info invocation on the setup.py presented above to ascertain where the failure lies.

I traced this into wheel and then figured it would be easier to patch downstream.

[agronholm]

This can be closed now, yes?

[boppreh]

@agronholm I still experience the problem. Xoviat says the problem has been traced into "wheel", but I don't see an open issue, or reference to the offending piece of code. I would not rather close this until the problem is fixed, or a new issue is created.

[agronholm]

Why not open an issue against wheel then? Provide instructions to reproduce and I'll fix it.

[boppreh]

I haven't opened an issue against wheel because I don't know if it's a wheel problem or not. The only indication I have is the comment above, from a deleted account, made in 2017.

I've also implemented my own workaround (manually removing Windows linebreaks), so feel free to take ownership of the issue.

[agronholm]

I can't reproduce this on Linux. Will try again with Windows 10.