Cody Allen

> Shouldn’t deleting a unique type fail if there are dependencies on that type? Yes. But there are currently some issues such as #5044 that can lead to this situation....

@aryairani I think that my honest answer is that saving a bit of time copy/pasting the prompt from the line above doesn't seem worth it to me. I think that...

@aryairani well I guess that it was concrete in the sense that we weren't able to test that `switch` properly returns to the previous branch in Mitchell's PR. But I'm...

Thanks @sellout that seems good enough to me and it seems like the approach that I suggested isn't palatable to others, so I'm going to go ahead and close this...

Thanks for the report. Strangely I'm on NixOS and I'm not seeing the same with ucm installed via home-manager in a flake. Are your dotfiles public somewhere that I could...

Looking at that diff it looks like you might have updated straight from 0.5.21 to 0.5.24. Does that sound right? I suspect that 0.5.23 is where the trouble started for...

The strange thing is that I am also on NixOS 24.05 with a similar flake setup. I'm struggling to think of why we are encountering different results. If only we...

Hmm here is something suspicious. I have a different hash for `unison-code-manager-0.5.24`: ``` ❯ ls /nix/store/l881g8z7az2n7d4k68q887pc0j9fabw0-unison-code-manager-0.5.24 bin/ lib/ share/ ui/ unison/ ``` I wonder if some derivation somewhere had that...

> I'm not sure I totally understand how that could happen in Nix, as I thought hashes were always checked deterministically. The scenario that I was thinking of (and have...

I just tried from macos with home-manager and I don't see an issue there either. Strange. I was wondering if it could be that an overlay is getting applied out...