| 22 May 2025 |
 emily | I guess the problem is that if you're using hooks they can't necessarily see cargoDeps until build | 12:30:37 |
 Toma | we wouldn't break hashes anyway, just APIs | 12:30:43 |
| emily | I guess I don't have a full picture of the problem | 12:31:01 |
| Toma | So the issue with the complete removal of the middle layer, is that then, we'll have fetchCargoVendor spit out our custom vendor-staging FOD, while other users may have assumed the cargoDeps is going to stay as something that is already the final vendor directory.
Now, if we had a marker in the FOD, we could just condition our logic to only trigger if we detect it.
In any case, the symlink solution doesn't sound that bad, if a little bit unnecessary...
| 12:35:50 |
| Toma | * So the issue with the complete removal of the middle layer, is that then we'll have fetchCargoVendor spit out our custom vendor-staging FOD, while other users may have assumed the cargoDeps is going to stay as something that is already the final vendor directory.
Now, if we had a marker in the FOD, we could just condition our logic to only trigger if we detect it.
In any case, the symlink solution doesn't sound that bad, if a little bit unnecessary...
| 12:36:21 |
 Adam Neverwas | Um...i have a kernel driver or what, patch, idk. I should have to add, because my notebook needs it. It's written in rust. | 17:08:15 |
| Adam Neverwas | Sec. | 17:09:38 |
| Adam Neverwas | However, don't you plan to add the asus-armory driver to the kernel? | 17:23:17 |
 K900 | Once mainline merges it, yes | 17:24:42 |
| K900 | Which so far is looking like it will take a while | 17:24:52 |
| Adam Neverwas | Thanks, asus-linux guys said it should reach somewhen, they dont know, didnt talk with main dev | 17:50:53 |
| K900 | Either way, the answer for "when will nixos have it" is "whenever mainline has it" | 17:51:34 |
| K900 | Which is right now 6.16 at the earliest | 17:51:40 |
| K900 | And I'm going to guess that they'll miss 6.16 | 17:51:46 |
| 23 May 2025 |
 John Ericson | In reply to @qyliss:fairydust.space
IIRC @Ericson2314:matrix.org didn't think it was possible some time in the past Hmm? | 03:21:26 |
| John Ericson | I definitely want to never build computers and standard libraries together | 03:21:39 |
| John Ericson | I would love to crate2nix rustc | 03:21:59 |
 Alyssa Ross | In reply to @Ericson2314:matrix.org
I definitely want to never build computers and standard libraries together I remember you thinking more work had to be done upstream in Rust to enable it | 08:35:10 |
| emily | I'm not sure why we can't just "fastCross all the things". my guess is because it will always build std for the host (and maybe build?) platform? | 08:36:42 |
| emily | what I wonder is whether, e.g., a build system hack to skip that, plus -Zbuild-std, would work. my guess is: maybe no because it assumes it can build build.rs scripts without needing to worry about std? | 08:37:33 |
| emily | (-Zbuild-std doesn't work well with our existing paradigm because we'd build it for every single application and is also unstable. but it would be a good proof-of-concept.) | 08:37:54 |
 rosssmyth | An alternative to -build-std is rustc_build_sysroot which miri uses
https://github.com/RalfJung/rustc-build-sysroot
But would require wiring it up for nixpkgs | 15:35:05 |
| rosssmyth | * An alternative to -build-std is rustc_build_sysroot which miri (and some others) uses
https://github.com/RalfJung/rustc-build-sysroot
But would require wiring it up for nixpkgs | 15:35:13 |
| John Ericson | So I think that upstream Cargo should be better and -Zbuild-std doing multiple solves is an ugly hack | 15:53:40 |
| John Ericson | but that is for "not having a sysroot" and "not treating the standard library as special" | 15:53:56 |
| John Ericson | for merely building things separately, we can still do that | 15:54:08 |
| John Ericson | I have recreated the -Zbuild-std hack by mashing together two crates2nix package sets | 15:54:22 |
| John Ericson | and even buildRustPackage could probably build a standard library sysroot more or less fine | 15:54:49 |
| John Ericson | so the purely downstream thing would be basically replacing that python script they have in the rust repo with a bunch of buildRustPackage invocations | 15:55:19 |
| John Ericson | according to the bootstrap process | 15:55:25 |
