In chm(python363.chm) documents, some unicode chars (non ascii chars) can not display.
for example:

asyncio — Asynchronous I/O, event loop, coroutines and tasks

displayed as

asyncio � Asynchronous I/O, event loop, coroutines and tasks

and 

Asynchronous programming is more complex than classical “sequential” programming

display as

Asynchronous programming is more complex than classical 搒equential� programming


windows 10, simplified chinese language.
python3.6.3, python363.chm.

I'm not sure there will be any good fix for this. We might be able to coerce proper utf-8 output from Sphinx, and if it also adds the encoding tags required by whatever ancient version of Internet Explorer is used then it should be fine

It's likely just best to avoid special punctuation in doc source files though.

The doc source files do not contain smart quotes, and as far as I know, sphinx does produce correct utf-8.

Recently there was a bug where incorrect smart quotes were leaking out of the internationalization of the docs, so this might be a problem that is already fixed.  On the other hand, there might be something broken about the chm production process.  I have no idea who would be the right person to investigate that, since I think Steve just spins the wheel on existing tools to get them generated :)

On the gripping hand, could there be something broken about your local charset configuration?  Does anyone else see this problem?

I found the problem was not fixed on python364.chm but it show well on python362.chm, maybe the python.org official config was a change to let the coding error.

Here is a solution:
1, open a page(whatever) with Internet Explorer.
2, right click the page -> Encoding -> check "Auto-Select"
Then the wrong characters (�/抯) will disappear forever.

> Does anyone else see this problem?
Probably a lot of people have this problem.
I installed a clean Windows 10 recently, I believe it's the default visual effect of Python .chm document.
BTW my local is Simplified Chinese.

The source code of .chm changed between 3.6.2 and 3.6.3, the former uses escaped html entities.
I couldn't find out which commit caused this change.

3.6.2 chm: <h1>What&#8217;s New In Python 3.6</h1>
3.6.3 chm: <h1>What抯 New In Python 3.6</h1>

3.6.2 chm: <h2>Summary &#8211; Release highlights</h2>
3.6.3 chm: <h2>Summary ?Release highlights</h2>

Release date:
3.6.2 final: 2017-07-17
3.6.3 final: 2017-10-03

We should probably prefer to force ASCII with explicit escapes (ideally named escapes, rather than codepoints). I'm not sure how to make Sphinx/docutils do that, but presumably it could be our own extension that handles the problematic characters people add to our docs.

In python365.chm, it loss the style in any page but can show in IE.

And in python365.chm, also had some non ascii chars can not display.

This issue still persists in 3.6 and 3.7.

Until someone creates and enables a Sphinx extension/option to only generate ASCII output, it will remain. Volunteers are welcome

I will create a PR to fix this within a day.

It seems impossible to specify the encoding of .chm to ASCII [1], the available encodings of .chm are limited to a list [2].

So I wrote a Sphinx extension for .chm output, it escapes the characters which (codepoint > 0x7F) to 7-bit ASCII. Most escaped characters are: “”’–…—

[1] https://github.com/sphinx-doc/sphinx/blob/master/sphinx/builders/htmlhelp.py#L203-L206
[2] https://github.com/sphinx-doc/sphinx/blob/master/sphinx/builders/htmlhelp.py#L136-L170

New changeset 6261ae9b01fb8429b779169f8de37ff567c144e8 by Steve Dower (animalize) in branch 'master':
bpo-32174: Let .chm document display non-ASCII characters properly (GH-9758)
https://github.com/python/cpython/commit/6261ae9b01fb8429b779169f8de37ff567c144e8

Thanks, that looks perfect!

New changeset 64bcedce8d61e1daa9ff7980cc07988574049b1f by Miss Islington (bot) in branch '3.6':
bpo-32174: Let .chm document display non-ASCII characters properly (GH-9758)
https://github.com/python/cpython/commit/64bcedce8d61e1daa9ff7980cc07988574049b1f

New changeset c4c86fad8024dc91af8d785c33187c092b4e49d9 by Miss Islington (bot) in branch '3.7':
bpo-32174: Let .chm document display non-ASCII characters properly (GH-9758)
https://github.com/python/cpython/commit/c4c86fad8024dc91af8d785c33187c092b4e49d9

It seems to me that this adds escape4chm as unconditional dependency on all platforms. Which seems like a bad idea to me, I don't think users on Linux or Mac OS X are that keen on *.chm files.

I think this change broke my build of python3-doc package on openSUSE (which built until now absolutely perfectly).

I don't even know, where the escape4chm plugin comes from. ???

What version of Python are you running Sphinx with?  Your error is that `html.entities` does not exist, which makes it sound like Python 2; bump it to Python 3 and you'll be fine.

> It seems to me that this adds escape4chm as unconditional dependency on all platforms.

`Doc/Makefile` also includes `htmlhelp` command, it generates .chm materials.

BTW, the related navigation bar still has corrputed characters, see the attached file. This can't be fix via Sphinx extension.
I'm writing a Sphinx patch, if they accept it, we can revert this commit, a new option will sovle this perfectly: htmlhelp_ascii_output = True

Sorry, my mistake, it seems I was using python2 Sphinx even for building python3 documentation, which is a bad idea, I guess.