How important is 3.6 support (on Windows?) at this time? I was looking into the build issues, but don't want to sink too much time if releasing 0.18 soon and don't care about this at all. I raised a draft PR that installs Rust toolchain for Windows builds, and it resolves the current issues. However, I'm still getting issues installing packages using pip in the CI, looks like for two reasons: 1. charmap can't deal with the explosion emoji coming from somewhere (click?) 2. Scarier thing is, getting stuff like

Ignoring ipython: markers 'python_version > "3.6"' don't match your environment
Ignoring pip-tools: markers 'python_version > "3.6"' don't match your environment
Ignoring pip-tools: markers 'python_version > "3.6"' don't match your environment

Guess the latter may explain the Jupyter timeouts? CI: https://app.circleci.com/pipelines/github/kedro-org/kedro/6002/workflows/a4c2f4b9-9745-4319-b8de-a5a898f40aca/jobs/92733

Unless you're written something that actually seems like it might realistically break something on Windows, Python 3.6 I wouldn't worry about it. We opted to make that job no longer required after the Rust toolchain errors started appearing (am curious to see how you fixed that :D), so if there's a whole load more difficulties that crop up after that's fixed then I wouldn't invest much time in it at all... I think chances are high that Windows, Python 3.6 e2e tests will not come back into force before 0.18 is released and they become irrelevant

I fixed it by using the installer from rustup.rs. I got the exact commands eventually from https://github.com/esp-rs/rust-build/blob/92da495f97b7d174f3a22a36627caf3fdc09d491/Install-RustToolchain.ps1#L32-L37, since it seemed easier than figuring out the flags myself.

3.6 is officially end of life so not allowed in enterprise contexts anyway - I’m looking forward to binning it

Fair enough. I won't burn more time on this.