Test strategy?

[pmeier]

Although we needed #60 to move forward, we also lost significant test coverage. Since we no longer have the ability to test the individual parts (due to pip's internal changes), we can only perform end to end tests now. This means we basically need to run ltt install with varying arguments and check if the installed packages are correct. Of course that would mean we will download a humongous amount of data for very little information.

One possibility would be to patch the download logic of pip by injecting fake wheels. Furthermore we would also need to patch the installation logic to stop as soon as the candidates to install are determined and check those.

My guess is that will be quite brittle, but TBH, this whole project is 😇 When I find the time, I'll start experimenting with the idea. If someone wants to help out, feel free to reach out. This is quite a deep rabbit hole.