| 7 May 2026 |
 siraben | going through 1.92 to 1.95 | 16:10:39 |
 K900 | It's not that it's hard to do | 16:11:15 |
| K900 | The problem is that it basically slows down the critical path on Hydra by like, days | 16:11:24 |
 whispers [& it/fae] | yeah, i did it out of tree a few versions ago and build times were rough. the numbers i had were:
reference for build perf on my laptop: nom-build -A rustc-unwrapped --check on nixpkgs master: 1h52m17s mrustc: 5m2s rust-1900: 1h54m18s (you can probably cut this one down but i didn't bother at the time) rust-1911: 17m33s rust-1920: 18m53s rust-1931: 18m1s rustc-1940: 1h48m52s (literally just pkgs.rustc-unwrapped.overrideAttrsed)
| 16:13:32 |
 Alyssa Ross | I would love to do mrustc | 16:13:39 |
| whispers [& it/fae] | yeah, i did it out of tree a few versions ago and build times were rough. the numbers i had were:
reference for build perf on my laptop: nom-build -A rustc-unwrapped --check on nixpkgs master: 1h52m17s mrustc: 5m2s rust-1900: 1h54m18s (you can probably cut this one down but i didn't bother at the time) rust-1911: 17m33s rust-1920: 18m53s rust-1931: 18m1s rustc-1940: 1h48m52s (literally just pkgs.rustc-unwrapped.overrideAttrsed)
and that's uh. really rough | 16:13:44 |
| whispers [& it/fae] | yeah, i did it out of tree a few versions ago and build times were rough. the numbers i had were:
reference for build perf on my laptop: nom-build -A rustc-unwrapped --check on nixpkgs master: 1h52m17s mrustc: 5m2s rust-1900: 1h54m18s (you can probably cut this one down but i didn't bother at the time) rust-1911: 17m33s rust-1920: 18m53s rust-1931: 18m1s rustc-1940: 1h48m52s (literally just pkgs.rustc-unwrapped.overrideAttrsed)
and that's uh. really rough when the chain gets big | 16:13:51 |
| whispers [& it/fae] | yeah, i did it out of tree a few versions ago and build times were rough. the numbers i had were:
reference for build perf on my laptop: nom-build -A rustc-unwrapped --check on nixpkgs master: 1h52m17s mrustc: 5m2s
rust-1_90_0: 1h54m18s (you can probably cut this one down but i didn't bother at the time)
rust-1_91_1: 17m33s
rust-1_92_0: 18m53s
rust-1_93_1: 18m1s
rustc-1_94_0: 1h48m52s (literally just pkgs.rustc-unwrapped.overrideAttrsed)
and that's uh. really rough when the chain gets big | 16:14:28 |
| whispers [& it/fae] | * yeah, i did it out of tree a few versions ago and build times were rough. the numbers i had were:
- reference for build perf on my laptop:
nom-build -A rustc-unwrapped --check on nixpkgs master: 1h52m17s
mrustc: 5m2srust-1_90_0: 1h54m18s (you can probably cut this one down but i didn't bother at the time)rust-1_91_1: 17m33srust-1_92_0: 18m53srust-1_93_1: 18m1srustc-1_94_0: 1h48m52s (literally just pkgs.rustc-unwrapped.overrideAttrsed)
and that's uh. really rough when the chain gets big (especially for local builds on staging that involve rust)
| 16:15:12 |
| whispers [& it/fae] | * yeah, i did it out of tree a few versions ago and build times were rough. the numbers i had were:
- reference for build perf on my laptop:
nom-build -A rustc-unwrapped --check on nixpkgs master: 1h52m17s
mrustc: 5m2srust-1_90_0: 1h54m18s (you can probably cut this one down but i didn't bother at the time)rust-1_91_1: 17m33srust-1_92_0: 18m53srust-1_93_1: 18m1srustc-1_94_0: 1h48m52s (literally just pkgs.rustc-unwrapped.overrideAttrsed)
and that's uh. really rough when the chain gets big (especially for local builds on staging that involve rust)
| 16:16:25 |
 dish [Fox/It/She] | mrustc would be good only if they could do smaller releases that keep up with rust more closely, but i think that requires a higher bus factor which just isn't a thing atm | 16:19:10 |
| siraben | we could have a non-blocking path for it? | 16:22:13 |
| siraben | similar to how minimal-bootstrap used to do that before the switchover | 16:22:28 |
| K900 | We could, and then it'll bitrot immediately\ | 16:22:39 |
| K900 | * We could, and then it'll bitrot immediately | 16:22:40 |
| K900 | Just like minimal-bootstrap did | 16:22:49 |
| siraben | i see | 16:22:55 |
| K900 | Like, if you really want to do it, I can't stop you | 16:23:08 |
| siraben | i'll just wait to see what comes out of the slop machine and keep or discard | 16:23:30 |
| K900 | But long term I don't think it's worth doing for nixpkgs unless we can have reasonable confidence that the bootstrap will not take months | 16:23:34 |
| siraben | 
Download Screenshot 2026-05-07 at 09.23.35.png | 16:23:42 |
| K900 | We are already way slower than where we'd like to be | 16:23:55 |
| K900 | For end to end world rebuilds | 16:24:02 |
| siraben | what areas need most attention that people don't want to really touch? | 16:24:21 |
| K900 | Those areas are also areas where I'd like to see the least LLM use | 16:24:50 |
| siraben | yeah of course, im limiting it to very routine things atm that would take hours to measure and configure | 16:26:26 |
| siraben | Ok I can see why people don't want rustc bootstrap lol | 23:08:24 |
| siraben | it's taking forever to build, finally on the last release | 23:08:31 |
| 8 May 2026 |
|  jopejoe1 changed their display name from jopejoe1 (4094@epvpn) to jopejoe1. | 08:42:36 |
| siraben | openjdk 17 bootstrap via iced-tea works | 15:06:59 |
