| 3 Dec 2025 |
 raitobezarius | The pipeline for the job is a simple nix run lix#ci-config which calls a calculate-buildkite-steps.py script which emits dynamic JSON that expands the graph of builds as we n-e-j them | 21:17:38 |
| raitobezarius | Theoretically yes, there was an AFNix infra task to get this working, but I didn't have to finish it for lanzaboote in the end | 21:18:19 |
| raitobezarius | 
Download clipboard.png | 21:18:38 |
| raitobezarius | that's the exact Lix plan | 21:18:41 |
| raitobezarius | ah Forgejo actions you said | 21:19:09 |
| raitobezarius | that's a pretty weird thing to do to call Buildkite from there, but I don't see any reason why not | 21:19:25 |
 piegames | Then what is the business model of Buildkite? Because gerrit users probably don't abound | 21:32:04 |
| raitobezarius | the business model of buildkite is selling you hosted CI minutes and various features surrounding the control plane | 21:32:29 |
| raitobezarius | what is your fundamental question? | 21:32:57 |
| piegames | I'm just curious | 21:35:03 |
| raitobezarius | I see | 21:35:13 |
| raitobezarius | TLDR Buildkite is the most advanced public CI system that exist and is probably earning a shit ton of money via enterprise contracts with very large monorepos out there IMHO | 21:35:33 |
 Qyriad | it's a shame to not use our own infra but man does it work a hell of a lot better than buildbot ever did | 22:36:50 |
 just1602 | Yeah, buildbot looks nice in theory, but in practice it looked rough | 23:27:25 |
| 4 Dec 2025 |
| raitobezarius | this is being fixed | 01:32:51 |
| raitobezarius | but slowly :D | 01:33:04 |
 aloisw | In reply to @raitobezarius:matrix.org
yeah that's good except that the lifetime of my SQLite DB is usually <1y before it corrupts itself: D Is it because of SQLite bugs, or because you're doing something weird to it? | 05:49:14 |
| raitobezarius | the bar to clear to say something is a SQLite bug is alas very high | 11:38:12 |
| raitobezarius | i don't trust that highly SQLite though | 11:38:17 |
| raitobezarius | but | 11:38:18 |
| raitobezarius | I trust even less our code given prior experience with the architecture | 11:38:28 |
| raitobezarius | so I lean on "SQLite is fine, we are holding it wrong, it's very hard to know why without redoing it entirely" | 11:38:41 |
| raitobezarius | (I don't do anything wrong) | 11:38:55 |
| raitobezarius | * (I don't do anything weird) | 12:42:46 |
 toonn | And there's little hope of a reproducer, I assume? | 12:44:21 |
| raitobezarius | if you torture the DB layer via the state layer, perhaps | 12:46:19 |
| raitobezarius | but idk if it's worth spending that amount of time | 12:46:24 |
 antifuchs | Anything that gets the “error: (not an actual error) SQLite db is locked” spew to stop would be fine by me (: | 13:17:07 |
| antifuchs | Just wanted to do a quick vibe check: what’s the most modern way to build aarch64-linux things without a dedicated aarch64 Linux machine these days? I was trying binfmt_misc but doing stuff like nixos images hangs for a half hour and then times out (not to mention the enormous amount of “cannot allocate memory” spew) | 13:20:09 |
 K900 | Remote builder | 13:25:07 |
