Failing tests on windows

[amueller]

I'm trying to make a list, I'll keep editing this.

test_build_book is failing because of #583: <script src="_static/js/myjs.js"> is _static/js\myjs.js because the first half is added by sphinx which probably hardcoded the \ and the second part is what pathlib thinks paths should look like on windows.
I wonder if we can do a simple replace but that seems hacky.

'charmap' codec can't decode byte 0x9d in position 1589: character maps to <undefined> in test_build_docs0 from test_build_docs.

'charmap' codec can't encode character '\u03b8' in position 196: character maps to <undefined> in test_clean_html_latex

C:\Users\t3kci\checkout\jupyter-book\tests\books\toc\index.md:7:toctree contains reference to nonexisting document 'subfolder\\index' in test_toc_startwithlist and test_toc_withheaders and test_toc_urllink

[choldgraf]

Hmmm - I bet that this is where the problem is being introduced: https://github.com/executablebooks/jupyter-book/pull/583/files#diff-203e09245c463f220cf61faa8b789f18R17

I wonder if @phaustin has fixes for making Pathlib behave like *nix sometimes, and like windows other times? Maybe this is related to the posix stuff you were working on?

[amueller]

I thought there was more windows bugs but I think I get the same errors on linux (WSL). hrm.

[phaustin]

that should be fixed on my sphix_toc branch here:

e909222

@choldgraf I was holding off on testing on windows until all the sphix toc moves were completed. Is it time to retry now against current master?

[amueller]

@phaustin does that also fix the subfolder\\index issue?

[phaustin]

yes, and it passes all the tests except for toc_urllink, and headers don't show up in the rendered html. I wanted to let Chris finish turning over the toc creation to sphinx before chasing those last things down.

[phaustin]

I think rather than trying to toggle between windows and unix paths (which pathlib doesn't support) the better approach is to make every string represenation of a path posix (which is what as_posix is for)

[choldgraf]

@phaustin I believe that from a toctree standpoint, that should all be handled by sphinx functions etc now 👍

[phaustin]

great, will rebase and see where we're at tomorrow

[amueller]

the encoding issues are independent, though, right?

@phaustin did you work on adding windows to CI?

[phaustin]

Yes, the encoding is seperate, and we're not having any issues, so any more utf8 problems are new bugs.
I've just been running pytest on my laptop, so no on item 2.

[amueller]

Ok, I'm giving it a go at #807.

[chrisjsewell]

fixed in v0.8 😄 https://jupyterbook.org/reference/_changelog.html#v0-8-0-2020-09-01