The first step in getting tests to pass was to fix up the Mercurial-related tests. It appears those were broken for me because I was doing something at least somewhat peculiar to have Mercurial available everywhere, but I realized, I don't really use it except for publishing this blog. (So, the fact that you're reading this means I got everything working well enough.) Now, Mercurial itself is not trapped on Python 2, but the latest PyPI release of the extension that is the other reason I ever use Mercurial nowadays is, so I undid the global-ish install. (To an extent.) I have to stop clinging obnoxiously to the past...
And start clinging obnoxiously to the future! The new Pijul alpha is super confusing after versions 0.12 and 0.11, but I don't care! You hear that, world‽ You can make me use git, but you can't make me like it, and I will gladly use experimental software for my own stuff!
Anyway, making it so the tests don't hit the wrong Mercurial script fixed some of the tests, but it looks like some tests failed for other reasons. Those other reasons include "I'm hitting the rate limits that broke pip's CI a few months ago. But this isn't CI. This is a laptop. Oh well."
If I can just use the same workaround, I'm happy.
So, I'm rerunning the tests with the "CI" logic activated, and we'll see how that does. The next step is getting it to run on every version of Python that isn't EOL.
Looks like all that is enough to get some of the Pythons to pass. I'm going to take it easy for the rest of the night while that runs.