It's always caching. Or DNS. This time it was caching.
A while back, we changed the default cache key prefix from empty-string to `semver` so that our cache key names don't start with a dash, since users found that confusing.
We should have updated these tests accordingly at the time, but we did not. So how did the PR with the change pass?
It used the caches from older runs of the same job which used the old names. When those caches expired and were GC'd by GitHub, the jobs started failing.
This is what broke CI for #77. This PR will fix it.
* Add exclude, allow lists in package
* Add test for input exclude
* Second dummy crate in test workspace
* New tests
* Fix cache key
* Forgot about flat()
* Reflect changes in docs
* New key in cache test
* Shorter job name
* Allowed to fail -> expected to fail
* Update README.md
Co-authored-by: Predrag Gruevski <2348618+obi1kenobi@users.noreply.github.com>
* Hash inputs in cache key
* Update cache test
* Update README
---------
Co-authored-by: Predrag Gruevski <2348618+obi1kenobi@users.noreply.github.com>
* Add inputs for controlling rust toolchain
* Add tests for new inputs
* Negated ifs!
* Fix action input name
* Better description of rust-override
* New logic of choosing toolchain
* Add explaining comment.
* Add rust-toolchain.toml test
