| 15 Jun 2026 |
|  Theodora changed their profile picture. | 14:48:41 |
| 17 Jun 2026 |
|  @braid:alsace.hair joined the room. | 22:24:56 |
| @braid:alsace.hair set their display name to The one with the braid (she/her) ⚡. | 22:25:33 |
| 18 Jun 2026 |
|  pentane banned @braid:alsace.hair. | 06:27:02 |
| pentane | sorry btw for not spending any time on robotnix in the past couple weeks, i've been unhealthily overworked from my full time job | 06:28:54 |
| pentane | autra: I think your PR looks like the right way to go as a hotfix. However I tbh think that we should take this as an opportunity to completely rewrite repo2nin | 06:30:26 |
| pentane | * autra: I think your PR looks like the right way to go as a hotfix. However I tbh think that we should take this as an opportunity to completely rewrite repo2nix | 06:30:27 |
| pentane | i'm not happy at all with the spaghetti-esque state the code is in right now | 06:32:33 |
| pentane | in particular, what needs to change is the git-repo manifest parsing. I'm currently modelling the manifest XML files essentially as key-value pairs of "relative path in source tree" => "git commit of source repository", but the way upstream git-repo operates makes them closer to "scripts" that are read from the top to the bottom | 06:34:42 |
| pentane | for instance, there's a "command" called <remove-project> which removes a project that was defined further up, which is for instance used by the Waydroid manifest snippets: https://github.com/waydroid/android_vendor_waydroid/blob/lineage-20/manifest_scripts/manifests-33/01-removes.xml | 06:35:45 |
| pentane | and yeah, if we want to sensibly support waydroid instead of hacking something together, repo2nix needs to know that "command" | 06:36:26 |
| pentane | also, on top of that, the current repo2nix architecture assumes that the update process is a straightforward "fetch manifest" -> "fetch lineage dependencies" -> "fetch git-repo dependencies" chain. With the addition of that new roomservice.py feature, a lineage-style dependency can now suddenly cause a manifest snippet to be parsed | 06:46:37 |
| pentane | the way my updating architecture does toposorting of dependency fetching graphs already is a big mess, and what we'd need is a unified way for the fetching process to generate the fetching graph on-the-fly. would also help for parallelism. | 06:48:15 |
 autra | no problem, you owe us nothing. Thanks for spending so much time on it! | 08:55:36 |
| autra | do you mean that order is significant? If remove-project happens before (on the order of reading) a project is added, will it remove that project or not? | 08:57:07 |
| autra | which roomservice.py are you referring? I've only found https://github.com/Lineage-OS/android_vendor_cm/blob/177ab036d0486e60b8d8000258069bed5dfc18e9/build/tools/roomservice.py, which seems to be very old. | 09:00:54 |
| pentane | Yes exactly | 09:01:37 |
| autra | omg but why? 😆 | 09:02:05 |
| pentane | the one in android_vendor_lineage | 09:02:10 |
| pentane | Because git-repo is fucking garbage, simple as :D | 09:02:23 |
| pentane | that's why LineageOS had to design a suite of hacky Python scripts around it to get any serious work done at all | 09:02:50 |
| autra | You'd think google would spend enough resources on a tool that is so used in their workflow.... :-/ | 09:03:39 |
| autra | Ok, I've searched in lineage-os github org, which is NOT the same as LineageOs org | 09:04:17 |
| pentane | yeah, but unfortunately i think that more often than not, "it works well enough" is a killer argument for never doing anything against tech debt in corporate environments | 09:04:55 |
| pentane | the AOSP signing scripts are similar man made horrors beyond comprehension | 09:05:21 |
| pentane | like. literally every single custom ROM has some sort of bespoke mechanism for wrapping git-repo because it's so useless | 09:06:48 |
| pentane | It's a failed experiment from the early days of git and was meant to provide for a generic way to do dependency management with git | 09:07:36 |
| pentane | (hence the generic name) | 09:07:43 |
 magic_rb | Maybe it should join the killed by google fsmily then | 09:11:27 |
|  piegames joined the room. | 09:22:12 |
