Micha Reiser

With kill and reactivate, you mean you stop ty, then re-run the command?

Very interesting: * ty doesn't use any persistent caching. Each run is a fresh run * The `--watch` and non-watch use the exact same code paths. The only difference is...

> Could multiple events be fighting to apply changes at the same time? That should only be an issue for subsequent runs of the watch, but not the initial run....

I tried to reproduce the issue by running ty on `discord.py` in watch mode but it always reported the same number of diagnostics. But I put up https://github.com/astral-sh/ruff/pull/22078, which should...

I think all that is needed here is to have VIRTUAL_ENV set to an incorrect path I'm not sure if it's even worth erroring if the VIRTUAL_ENV path is incorrect....

> The URI is from a test framework (IntelliJ's). That ty can't resolve things is expected, but it shouldn't panic here. Fair. It's not entirely clear what ty should do...

Note, this is also an issue outside tests, e.g .when the `VIRTUALEN_ENV` variable points to an invalid directory as described here https://github.com/astral-sh/ty/issues/611#issuecomment-3123403717 Overall, my assumption that resolving the default `ProgramSettings`...

I'm surprised this is an issue, given that all Windows file systems are case-insensitive. But maybe the real issue is that we successfully resolve the environment with `Lib`, but then...

It might be worth improving our diagnostics if we dedect this very specific issue

> One issue I ran into when trying this with a project that is gradually adopting types is that tool.ty.src.exclude doesn't seem to be respected when running ty check against...