| 16 Oct 2025 |
Mic92 | In reply to @brian:bmcgee.ie But wouldn't that mean more nars than narinfos? You can end up with more narinfos if you produce the very same build-output. | 15:13:26 |
Mic92 | Ah vcunat said this already | 15:13:48 |
dramforever | i am pretty sure this is wrong | 15:14:38 |
dramforever | narinfo is per path, nar is whatever | 15:14:44 |
dramforever | so you'll have one narinfo per output | 15:15:15 |
dramforever | protocol-wise the nars doesn't matter, you just need to point the URL in narinfo to something | 15:16:15 |
dramforever | i think it just happens that nix defaults to using a hash of the nar contents | 15:16:29 |
dramforever | or something like that | 15:16:33 |
dramforever | i wonder if we have a bunch of test derivations that just create an empty file at $out and that's why we have way more narinfo than nar | 15:17:37 |
Mic92 | Infra meeting in 7 minutes. | 15:54:00 |