| 4 Sep 2025 |
 jaen | For AOSP. Basically there's this WIP thing - https://github.com/pdtpartners/nix-ninja - that lets you get nix-based incremental builds for projects using ninja with dynamic derivations. AOSP generates ninja files, but they are so humongous the Rust ninja implementation they use just chokes one them. I was trying to write one that doesn't choke to be able to finally update GrapheneOS in robotnix without going insane from time wasted on build failures. And since GrapheneOS is now up to date in robotnix, that particular reason is no longer there ¯\_(ツ)_/¯ | 20:55:13 |
| jaen | Anyway, I was probably going about it the wrong way, given that somebody has done it without that xD | 20:55:43 |
 Atemu | When I tried that one, that project choked hard on even lix | 20:56:15 |
| Atemu | As it turns out, dynamic derivation have quite a bit of overhead | 20:56:44 |
| Atemu | Like any derivation I suppose | 20:56:53 |
| Atemu | Ah disambiguation: on building the lix project | 20:57:39 |
| jaen | Part of it is probably because they spawn `nix` to write the dynamic derivation | 20:57:40 |
| jaen | Or maybe lix uses some feature they don't support yet | 20:57:58 |
| jaen | But for AOSP it's certainly the sheer scale | 20:58:09 |
| Atemu | Ah, hm it might not worked on Lix now that I think about it but I built some other medium sized project and it could barely handle that and took quite a lot of time to even begin building anything | 20:59:01 |
| Atemu | If we're going to make it incremental, I think the only way forward is the hard way; packaging every AOSP project as a drv and wiring them together ourselves | 21:01:42 |
| Atemu | If you have experience with ninja, make and the like, there's probably lots you could help us with here | 21:02:33 |
| jaen | In reply to @atemu12:matrix.org
If we're going to make it incremental, I think the only way forward is the hard way; packaging every AOSP project as a drv and wiring them together ourselves Maybe so, but I wouldn't discount it until it has a proper parallel parser (part I was working on) and doesn't have to spawn nix to do things with derivation. If both those are fixed and it still chokes - fair enough, I guess. | 21:06:48 |
| jaen | Not sure if I have that much experience with that, honestly I've only used it as an excuse to write some actual Rust after over a decade of stalking the language | 21:07:33 |
| jaen | If that ends up being helpful, then nice. If it doesn't, then oh well, either way glad to see someone managed to get it up to date ;' ) | 21:08:23 |
|  angerman set a profile picture. | 21:15:27 |
| 5 Sep 2025 |
|  David joined the room. | 01:08:48 |
| 6 Sep 2025 |
|  SomeoneSerge (back on matrix) changed their display name from SomeoneSerge (Ever OOMed by Element) to SomeoneSerge (@nixcon & back on matrix). | 09:22:19 |
|  matthewcroughan changed their display name from matthewcroughan @ nix.camp to matthewcroughan @ nixcon. | 09:39:20 |
| 8 Sep 2025 |
|  inayet set a profile picture. | 02:17:12 |
| 9 Sep 2025 |
|  mio joined the room. | 06:13:50 |
| mio | An incremental AOSP build system! I want to have that | 07:18:28 |
| Atemu | We all do ;D | 08:07:44 |
| Atemu | I suppose it is incremental OOTB but cache invalidation sucks because it can't have multiple versions of any given artifact | 08:08:41 |
|  aciceri changed their display name from zrsk to aciceri. | 15:03:04 |
| 10 Sep 2025 |
| SomeoneSerge (back on matrix) changed their display name from SomeoneSerge (@nixcon & back on matrix) to SomeoneSerge (back on matrix). | 00:35:47 |
| matthewcroughan changed their display name from matthewcroughan @ nixcon to matthewcroughan. | 14:42:07 |
| 11 Sep 2025 |
|  loudgolem joined the room. | 17:52:18 |
|  martijn removed their profile picture. | 18:33:41 |
| 13 Sep 2025 |
|  oak 🏳️🌈♥️ changed their profile picture. | 09:46:24 |
