| 23 May 2021 |
 Sandro | maybe something for another PR but do we have written down what a mass rebuild is? The question comes up once in a while. | 15:38:02 |
 Synthetica | nixpkgs-update targets staging when it would cause >100 rebuilds | 15:38:24 |
 Alyssa Ross | for a human it's probabyl okay to use judgement to go a bit higher | 15:38:42 |
| Synthetica | Maybe we could adopt that? | 15:38:43 |
| Alyssa Ross | I think it's a bit more nuanced, because we can rebuild 1000 go programs when go updates really fast | 15:39:02 |
| Alyssa Ross | I don't believe we do have it written down, but it sounds like it would be a great addition to the Nixpkgs manual | 15:39:10 |
| Alyssa Ross | And I've seen sandro explain it well on IRC before | 15:39:19 |
| Alyssa Ross | I think | 15:39:24 |
| Sandro | In reply to @sandro:supersandro.de
5000+ is a must or if someone noticed it fast enough it is going to be reverted maybe it was this. not sure. | 15:41:27 |
|  cicucci set a profile picture. | 15:41:39 |
| Alyssa Ross | no, it was something giving Python as an example of where more rebuilds were okay | 15:41:49 |
| Alyssa Ross | I thought it was you but could have been someone lese | 15:41:55 |
| Synthetica | Of course we could just emphasize the fuzzy nature of it in the manual: "If your pr causes less than 100 packages to be rebuilt, you definitely don't need to target staging. When it's 100-x, it depends on the nature of those rebuilds. When it's more than x, you definitely should target staging. | 15:41:58 |
| Synthetica | * Of course we could just emphasize the fuzzy nature of it in the manual: "If your pr causes less than 100 packages to be rebuilt, you definitely don't need to target staging. When it's 100-x, it depends on the nature of those rebuilds. When it's more than x, you definitely should target staging." | 15:42:01 |
| Alyssa Ross | yeah, you could expand the middle sentence and say that interpreted languages and go can go a bit higher | 15:42:28 |
| Sandro | In reply to @sandro:supersandro.de
500+ depends on the packages. Python can go to master because it is fast and anything else that takes time to build should go to staging you mean this? | 15:42:29 |
| Alyssa Ross | sandro: that's the one! | 15:42:38 |
| Alyssa Ross | I liked that explanation | 15:42:43 |
| Sandro | I think we should work with examples to make it clear. | 15:43:08 |
| Alyssa Ross | yeah, agree | 15:43:20 |
| Alyssa Ross | we could also suggest that people look for past examples themselves | 15:43:37 |
| Alyssa Ross | that's what I usually do when I'm not sure | 15:43:41 |
| Alyssa Ross | "did the last person who updated this use staging or master?" | 15:43:50 |
 lukegb (he/him) | I wonder if we could use ofborg to compute a metric for the affected packages based on their total rebuild time | 15:44:50 |
| Alyssa Ross | I vaguely remember it having something like this that wasn't exposed very well | 15:45:09 |
| Alyssa Ross | I think I saw samueldr mention it once? | 15:45:13 |
 ris_ | could someone try restarting https://hydra.nixos.org/build/143224217 ? | 16:10:41 |
 andi- | ris_ done | 16:11:40 |
 toonn | AlyssaRoss: : O You lied! The hash totally did change! (No worries.) | 17:03:07 |
 Milan (they/them) 🏳️⚧️ | Mic92: The 248 update worked flawlessly by setting systemd.package to the systemd package from your branch. I couldn't do the full rebuild, because some haskell packages keep failing to build on my systems, regardless of the systemd version. | 17:14:37 |
