| 5 Mar 2025 |
 flokli | Nix also has no concept of whether a build did fail or not | 08:50:50 |
| flokli | You can only guess by checking whether there's a build Log available locally, and the store path or not. Bus as soon as there's a GC in between that doesn't GC the logs you're screwed 💁♂️ | 08:51:53 |
| flokli | * You can only guess by checking whether there's a build Log available locally, and the store path or not. But as soon as there's a GC in between that doesn't GC the logs you're screwed 💁♂️ | 08:52:57 |
 Picnoir | Right! I found other cases where you don't have drv files: the builtins relying on addToStore, like builtins.toFile
For instance: https://pastebin.aquilenet.fr/?97621c09cc60ef1f#HASi9d2NbdGYa7ARzyP4eVZTAhVDcL7AsooCFFG4rqAk | 08:53:39 |
| Picnoir | * Right! I found other cases where you don't have drv files: the builtins relying on directly calling addToStore, like builtins.toFile
For instance: https://pastebin.aquilenet.fr/?97621c09cc60ef1f#HASi9d2NbdGYa7ARzyP4eVZTAhVDcL7AsooCFFG4rqAk | 08:53:57 |
| Picnoir | Well, complexity++ in my mental model :D Nice | 08:54:12 |
| flokli | Oh yes, of course, imports and fetches are no builds | 08:54:19 |
| Picnoir | * Well, complexity++ in my mental model :D Nice 😢 | 08:54:26 |
| flokli | (native fetches) | 08:54:37 |
| Picnoir | Yeah | 08:54:41 |
| Picnoir | but they still end up in the same validpaths table. | 08:54:50 |
| Picnoir | And there's no direct way to set them appart from the "regular" builds as far as I can tell. | 08:55:05 |
| Picnoir | By looking at the db I mean. | 08:55:15 |
| flokli | Yes of course not | 08:56:04 |
| flokli | Else or would be too simple :-) | 08:56:12 |
| flokli | You also can have non-native fetches and native fetches both being present in your build graph, producing the same store path. | 08:57:11 |
| flokli | So whether something is a build or not then depends on evaluation order ;-) | 08:57:57 |
 Robert Hensing (roberth) | Recent addition:
Besides functioning as a content-addressed store, the Nix store layer works as a build system.
https://hydra.nixos.org/build/291787002/download/1/manual/store/derivation/index.html
| 08:58:09 |
| Robert Hensing (roberth) | The store we have is not a fantastic CAS (more could be done), but it fulfills both tasks | 09:00:34 |
| flokli | 
Download 1741165269059.jpg | 09:01:21 |
| Robert Hensing (roberth) | tvix does a great job at this, layering the build system onto the castore layer, instead of just juxtaposing them into storedir | 09:01:31 |
| Robert Hensing (roberth) | you beat me to it :D | 09:01:57 |
| flokli | Hehe, I have that meme on quick dial | 09:02:21 |
 Ilan Joselevich (Kranzes) | im sitting here in front of flokli, i saw him press "send" and then your message came in, I knew your were writing it before he sent lmao | 09:03:00 |
| Picnoir | If I understand your categorization correctly, you'd put FODs in the "output" category? | 09:04:08 |
| Robert Hensing (roberth) | yes. They're also the current most frequent example where "first" is a thing, because many FODs can produce the same output path | 09:05:10 |
| Picnoir | Shouldn't there be a third category for FODs? They're resulting from the builder but are output-addressed. | 09:05:21 |
| Picnoir | Yeah | 09:05:23 |
| Picnoir | And most importantly: can be poisoned. | 09:05:29 |
| Robert Hensing (roberth) | It applies to all CA outputs, where FOD is a special case of CA as it can be implemented without rewriting any hashes | 09:06:14 |
