I’m not sure how to correctly frame it: I have a flake, and a lock file. I built it and used it and all is fine. Building made a ./result link. Then when I come back the next day and build that same flake again, I see it downloads data from cache nix or so.
With the same lock and all, should it not be super fast? What updates does it have to get from anywhere, should it not all be locally cached. I even kept the ./result link and was thinking that should prevent that derivation from being garbage collected.
Any pointers as to why, and/or how to prevent this, are appreciated
It might be that only inputs gets redownloaded, to be able to evaluate the drv, and only after evaluation it is seen that it is not required to actually build something.
Well I feel quite stupid right now. I waited a day on purpose before I nix build that flake again, and I only saw the evaluation, no download yet. I’ll report back when I “catch it”, but if memory serves, I would usually see it download some ~186mb from something like cache.nixos.org or so, but I think that is very non-specific