| 12 Dec 2024 |
 Pratham Patel | On the topic of a binary cache for riscv, other than "we don't have [funding for] powerful enough hardware available," is there anything else that needs to be ironed out before we start searching for powerful riscv hardware to put in hydra | 06:16:26 |
| Pratham Patel | * On the topic of a binary cache for riscv, other than "we don't have [funding for] powerful enough hardware available," is there anything else that needs to be ironed out before we start searching for powerful riscv hardware to put in hydra? | 06:16:34 |
 Alex | Importantly, doesn't the hardware also need to be reliable? | 06:17:21 |
| Pratham Patel | I have like 5 coupons for Oasis (when it gets out) which gives a 50% discount. I can give 4 or all of them to the nixos foundation. | 06:17:46 |
| Pratham Patel | Yes. But what I actually meant to ask is, "Are we at a stage where the only obstacle for building riscv binaries in hydra is the unavailability of powerful+reliable hardware?" | 06:19:05 |
| Alex | Right. We already know from experimenting that most of the important packages build. Is there precedent of running Hydra against a platform with no official support? (maybe aarch64 did this at some point?) | 06:21:05 |
| Pratham Patel | One thing that I know of the top of my head is that the entirety of stdenv doesn't feel ready. Like stdenv.isRiscV64 is missing but stdenv.targetPlatform.isRiscv64 and stdenv.hostPlatform.isRiscv64 exist. | 06:23:24 |
| * Pratham Patel is afk for a while | 06:23:27 |
| Pratham Patel | In reply to @alex:tunstall.xyz
Right. We already know from experimenting that most of the important packages build. Is there precedent of running Hydra against a platform with no official support? (maybe aarch64 did this at some point?) When we're ready, we need to bootstrap with devtools, binutils and kernel for hydra to run on top of. Basically a bootstrap from scratch because it's natively compiled this time. | 06:59:27 |
| Alex | In reply to @thefossguy:matrix.org
When we're ready, we need to bootstrap with devtools, binutils and kernel for hydra to run on top of. Basically a bootstrap from scratch because it's natively compiled this time. Does nix have a RISC-V target yet? Last I checked it didn't, and booting Nix natively from e.g. Debian seems like the right place to start a clean bootstrap. | 07:01:35 |
 Tristan Ross | Like the nix package itself? | 07:02:45 |
| Tristan Ross | Nixpkgs itself supports RISC-V | 07:02:59 |
| Tristan Ross | Afaict, the nix package doesn't have RISC-V on the search mentioned. | 07:03:20 |
| Tristan Ross | Oh, search filters out systems the search itself doesn't "recognize" it seems. The nix package has platforms.unix. | 07:04:58 |
| Alex | Yes. It's built from a flake IIRC and lacked riscv64 as a target when I checked over a year ago. | 07:09:58 |
| Pratham Patel | I'll try building nix with qemu-binfmt on my end and get back when it's done and/or fails. | 07:37:59 |
| Pratham Patel | This will take some days given that it's building gcc, linux-headers, curl, cmake, rustc, curl and what not :D | 07:39:13 |
| Pratham Patel | I wonder if we can bootstrap a minimal nix. | 07:39:31 |
| Alex | In reply to @thefossguy:matrix.org
I'll try building nix with qemu-binfmt on my end and get back when it's done and/or fails. If you mean native Nix via Nixpkgs, that works. NixOS requires it. | 07:59:14 |
| Pratham Patel | Yeah, that's what I started | 07:59:40 |
| Pratham Patel | Just cancelled the build on my m4 and pulled out the 8gb vf2 | 07:59:55 |
 doge.2 | Pratham Patel (you can mention me): thanks for the suggestions, i'll try switching those out and see if they help :) | 12:18:32 |
| doge.2 | hmm, seems like it still wants to build derivations after changing lib.mkForce to lib.mkBefore and nixpkgs to nick cao's fork :\ | 12:31:27 |
| Alex | Increase Nix's verbosity and it will print out information about substitution; this could help debug cache misconfiguration. | 12:34:03 |
| doge.2 | strange, i'm seeing absolutely nothing about substitution now after making said changes | 12:38:37 |
| doge.2 | tried max verbosity (-vvvv) and i saw nothing | 12:38:46 |
 @snaki:kescher.at | In reply to @thefossguy:matrix.org
This will take some days given that it's building gcc, linux-headers, curl, cmake, rustc, curl and what not :D i could offer some x86 build machines with binfmt if that would help | 12:39:00 |
| doge.2 | that's what i've been trying to do, i have a framework laptop 13 running nixos that i've been cross compiling with risc-v, i might be doing it wrong but it doesn't seem to work right for me, idk | 12:40:33 |
| doge.2 | (i spent almost all of yesterday working purely on this project with almost no success) | 12:40:51 |
| Alex | In reply to @doge.2:matrix.org
that's what i've been trying to do, i have a framework laptop 13 running nixos that i've been cross compiling with risc-v, i might be doing it wrong but it doesn't seem to work right for me, idk Do you have a Git repository with the latest changes?
We could take another look to try to figure out why you're not getting cache hits. | 12:53:13 |
