| 22 Apr 2025 |
 dramforever | maybe there's some funny impurities? | 11:11:06 |
| dramforever | what's your nixpkgs commit, and what does which nix say? | 11:16:31 |
 jaen | In reply to @flokli:matrix.org
This is the dependencies / devDependencies problem, and whether to transitively propagate dev dependencies inputs Yes, but my understanding was you can work around this somewhat with flakes (either using getFlake one of the various flake-compats or https://flake.parts/options/flake-parts-partitions.html) - so my point there was that with automatic hosting as described with those "crystals", you would lose option to use those workarounds until a proper concept of "scoped" inputs is added. Unless I misunderstood how those workarounds work exactly and they don't really allow you to fetch those dev inputs lazily. Hope it's more clear now what I wanted to say? | 11:34:35 |
 flokli | In reply to @jaen:matrix.org
Yes, but my understanding was you can work around this somewhat with flakes (either using getFlake one of the various flake-compats or https://flake.parts/options/flake-parts-partitions.html) - so my point there was that with automatic hosting as described with those "crystals", you would lose option to use those workarounds until a proper concept of "scoped" inputs is added. Unless I misunderstood how those workarounds work exactly and they don't really allow you to fetch those dev inputs lazily. Hope it's more clear now what I wanted to say? You can use various hacks to workaround various problems, yes. But the way I understood it was that "crystals", or however we want to call a new schema / layout aim to solve this as a first-class citizen, not just another workaround. | 11:37:31 |
| flokli | There's too many workarounds for problems with the schema. | 11:37:52 |
| jaen | Yeah, that's true. I've just meant that "crystals" would have to come with input scopes from the get go, as it would make those workarounds (such as they are) not viable. | 11:39:52 |
| flokli | I'd be happy to read and give feedback on a design proposal. | 11:41:15 |
 aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | Idea: We should detect a "minimum syscall permission set" that allows Nix to work and make both hydra and ofborg follow this specification. Currently nested containers (for example, nix-build in systemd-nspawn) cannot pass the checkPhase of some packages, but we don't fix it because CI isn't set up as such. | 12:55:28 |
 Martin Schwaighofer | In reply to @puck:puck.moe
most of the logic in that file is to properly adjust stringLength output to the post-substitution paths, even, the actual logic there is like three lines Just asking out of curiosity: Why was string length important to you?
I don't see when that would matter with how they are usually used, so I figure maybe you are doing something unusual. 😅 | 15:21:26 |
|  Thomas joined the room. | 17:11:37 |
|  griff joined the room. | 19:07:55 |
| 23 Apr 2025 |
|  @slamphear:matrix.org joined the room. | 00:07:12 |
 puck | In reply to @mschwaig:matrix.org
Just asking out of curiosity: Why was string length important to you?
I don't see when that would matter with how they are usually used, so I figure maybe you are doing something unusual. 😅 it's for calculating the length of the final string, post-substitution, so it can be properly seated into the tarball | 12:39:20 |
| @slamphear:matrix.org left the room. | 12:40:10 |
| Martin Schwaighofer | In reply to @puck:puck.moe
it's for calculating the length of the final string, post-substitution, so it can be properly seated into the tarball Ok, I see. Thanks. 😊 | 13:04:10 |
 Mic92 | In reply to @aleksana:mozilla.org
Idea: We should detect a "minimum syscall permission set" that allows Nix to work and make both hydra and ofborg follow this specification. Currently nested containers (for example, nix-build in systemd-nspawn) cannot pass the checkPhase of some packages, but we don't fix it because CI isn't set up as such. The nix sandbox with seccomp defines the minimum set of syscalls already. | 13:24:18 |
| Mic92 | Issue is that your containers syscall set is smaller than set | 13:24:56 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @joerg:thalheim.io
Issue is that your containers syscall set is smaller than set We should at least throw a warning I guess? | 13:25:26 |
| Mic92 | How would you implement that? | 13:25:50 |
| Mic92 | This something you can do in a test derivation. But I don't think nix should test this | 13:26:55 |
| Mic92 | * This something you can do in a test derivation. But I don't think nix should test this on startup | 13:27:06 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @joerg:thalheim.io
This something you can do in a test derivation. But I don't think nix should test this Maybe test in Nix installer, and also add to like nix config check? | 13:28:27 |
|  jul1u5 changed their profile picture. | 18:31:40 |
| 24 Apr 2025 |
|  main.hose joined the room. | 18:39:23 |
| 25 Apr 2025 |
|  Benno Fünfstück joined the room. | 15:17:57 |
|  thirdofmay18081814goya joined the room. | 15:18:01 |
| Benno Fünfstück | is there some way to force a value in Nix but then later deallocate everything allocated by this forcing? Or is that fundamentally impossible due to how the evaluator works right now? | 15:23:57 |
| Benno Fünfstück | perhaps I could abuse fork for this... hmm | 15:24:40 |
 emily | well, you can drop all references to it and let it get GC'd | 15:29:27 |
| emily | but generally forced thunks don't go back to being thunks in lazy languages | 15:29:44 |
