navigation_with_keys = False

[gabalafou]

and update docs

Related to #1100.

Closes #1492.

[gabalafou]

There was talk on #1100 of adding a deprecation notice. Where would that go?

[drammock]

For changing the default value of a boolean, I typically use this pattern:

1. change the default to None
2. somewhere in the code, detect None, change None to the new default boolean, and warn that the default has changed.
3. after the release-after-next, change the default from None to the new default boolean, and remove the logic that triggers the warning.

In our case since we're changing a theme default (not a function or method param default) we'd probably need to put the "None->False & warning" logic in our __init__.py (in update_config function)

[gabalafou]

@drammock, is my previous commit what you had in mind? I inspected the value of theme_options within update_config when running the build command, sphinx-build docs docs/_build/html, and I noticed that the only keys in the dictionary were the same ones defined in conf.py under "html_theme_options", with none of the keys or values in theme.conf. Is there another point in the init and build process that I need to hook into or maybe a different object?

[drammock]

I noticed that the only keys in the dictionary were the same ones defined in conf.py under "html_theme_options", with none of the keys or values in theme.conf.

That is expected / not a problem. What we want to detect is whether the user overrode the default value of None, so it's fine that the values from theme.conf aren't accessible there.

[12rambau]

the failing CI is due to unrelated issues, I am confident enough that this is a working solution, are you ok if I merge it?

[drammock]

the failing CI is due to unrelated issues, I am confident enough that this is a working solution

in fact this broke all our tests. Running tests locally, the warning about not setting navigation_with_keys occurs on every test now. I will try to roll a fix for that into #1512

[12rambau]

Ok we can't continue this way tomorrow I'll try to spend some time to solve the existing crash of the CI and site build. I will may be forced to unplug some stuff on the way but tests are currently useless. I'll do it in multiple PR to keep track of the different steps starting by using nbsphinx to at least get the RDT build back

[gabalafou]

Ah sorry, I should have caught that but I stopped paying attention to CI because all of my PRs have been failing CI due to the Sphinx 5 issue that #1482 solves