| 9 Jul 2025 |
 K900 | That he forked | 17:03:36 |
| K900 | I know what it is | 17:03:40 |
 raitobezarius | what did you cook again | 17:03:46 |
| K900 | Nothing? | 17:03:51 |
| K900 | Well | 17:03:53 |
| K900 | I cooked that thing ages ago | 17:04:02 |
| raitobezarius | what's supposed to do | 17:04:08 |
| K900 | And Grimmauld cleaned it up considerably | 17:04:09 |
| K900 | It just builds an sqlite db of the dep graph | 17:04:24 |
| K900 | So you can compute rebuild heavy paths quickly | 17:04:35 |
| raitobezarius | https://git.grimmauld.de/Grimmauld/nix-closure-thingy/src/commit/87000c970a7428fee948b5046f9483beacf0626e/src/main.rs#L104-L106 oh boy this is going to be so slow | 17:04:45 |
| K900 | Without doing full evals | 17:04:45 |
| raitobezarius | anyway | 17:04:50 |
| raitobezarius | ❯ hyperfine "./2_93/bin/nix-eval-jobs --workers 6 /home/raito/dev/github.com/NixOS/nixpkgs --system x86_64-linux" -m 2
Benchmark 1: ./2_93/bin/nix-eval-jobs --workers 6 /home/raito/dev/github.com/NixOS/nixpkgs --system x86_64-linux
Time (mean ± σ): 105.262 s ± 1.510 s [User: 364.680 s, System: 75.009 s]
Range (min … max): 104.194 s … 106.330 s 2 runs
| 17:04:52 |
| K900 | I wonder if something broke inputDrvs or something | 17:05:05 |
| raitobezarius | now can HEAD beat all of that | 17:05:06 |
| K900 | And it's taking the slow path | 17:05:08 |
| K900 | Because the slow path is slow | 17:05:31 |
| K900 | Someone test this I am migraine | 17:06:31 |
| raitobezarius | spoiler yes | 17:08:42 |
| raitobezarius | ❯ hyperfine "./git/bin/nix-eval-jobs --workers 6 /home/raito/dev/github.com/NixOS/nixpkgs --system x86_64-linux" -m 2
Benchmark 1: ./git/bin/nix-eval-jobs --workers 6 /home/raito/dev/github.com/NixOS/nixpkgs --system x86_64-linux
Time (mean ± σ): 99.085 s ± 0.844 s [User: 364.338 s, System: 60.883 s]
Range (min … max): 98.489 s … 99.682 s 2 runs
| 17:08:44 |
| raitobezarius | and it's not even its final form | 17:08:47 |
 Grimmauld (any/all) | inputDrvs is not broken, it didn't trigger the eprintln | 17:14:07 |
| Grimmauld (any/all) | inputDrvs is broken on mainline nix-eval-jobs, but works fine on any lix-eval-jobs | 17:14:31 |
| K900 | OK so I guess the parsing is fucked | 17:14:49 |
| K900 | Maybe | 17:14:53 |
| Grimmauld (any/all) | that would surprise me | 17:14:59 |
| Grimmauld (any/all) | the rust code was exactly the same between both runs | 17:15:10 |
| K900 | Could be something about buffering or whatever | 17:15:26 |
| K900 | idk | 17:15:26 |
