http.client.HTTPResponse is not playing nice (Python 3.x)

[SimonSapin]

This gives the empty string on Python 3.x

from urllib.client import urlopen
html5lib.serialize(html5lib.parse(urlopen(
    'http://html5lib.readthedocs.org/en/latest/')))

The cause is a bug in Python: http://bugs.python.org/issue20007 , bug given CPython’s release cycle timeline I would like to have a work-around in html5lib.

The bug is triggered here:

html5lib-python/html5lib/inputstream.py

Line 122 in e269a2f

isUnicode = isinstance(source.read(0), text_type)

Unfortunately, the only work-around I can think of (adding a special case if isinstance(source, http.client.HTTPResponse): isUnicode = False ) is very ugly.

[gsnedders]

How important is this for WeasyPrint? Can push a new release if needs a timely turnaround.

[SimonSapin]

A PyPI release would help, thanks. WeasyPrint recently switched to default to html5lib, and this bug breaks HTTP URLs (such as the command-line example given on the home page) on Python 3.x. It just silently renders an empty PDF.