| 11 May 2022 |
 Julio Sueiras | but need to rebuild the image with password in the configuration.nix to allow sshing | 23:59:37 |
| Julio Sueiras | (since the display have issue) | 23:59:41 |
| 12 May 2022 |
| Julio Sueiras | going to probably setup the device as a cachix agent | 00:00:02 |
| Julio Sueiras | for pushing cache for riscv64 | 00:00:10 |
| Julio Sueiras | the part I am confuse about is why is it building perl when is assume the unmatched.cachix.org should provide it | 01:24:07 |
| Julio Sueiras | (systemd also) | 01:30:37 |
| Julio Sueiras | Zhaofeng Li: quick question, is binary-cache/world.nix is what populate the unmatched.nixos.org? | 01:37:03 |
| Julio Sueiras | * Zhaofeng Li: quick question, is binary-cache/world.nix is what populate the unmatched.cachix.org? | 01:37:10 |
 Zhaofeng Li | Yes, on this (pretty old) commit: https://github.com/zhaofengli/nixpkgs/tree/riscv-cached | 01:38:57 |
| Zhaofeng Li | I haven't kept it up to date lately, so you may want to spend some time building against master instead | 01:39:38 |
| Julio Sueiras | k, since I was mostly scratching my head for the fact that it was building systemd(since just realized that unbound, perl, etc are for building systemd) | 01:40:17 |
| Julio Sueiras | I assume master in this context being riscv(non-cached) branch? | 01:42:14 |
| Julio Sueiras | k, going to let it compile overnight, since binfmt isn't the fastest thing in the world(and mostly doing it because need tweaking the config) | 01:43:19 |
| Zhaofeng Li | In reply to @juliosueiras:matrix.org
I assume master in this context being riscv(non-cached) branch? I mean the latest master in nixpkgs along with #168826 (and my commits in the riscv branch if you want native builds). I haven't kept my riscv branch up to date either. | 01:46:46 |
| Julio Sueiras | ah k | 02:17:30 |
 matthewcroughan - nix.how | Zhaofeng Li: What would be needed to improve the riscv cache or builders? | 02:39:32 |
| matthewcroughan - nix.how | is it hydra, baremetal, or is it VMs? | 02:39:42 |
| matthewcroughan - nix.how | I'd love to own whatever the best performing RISC-V hardware is, so I can develop on it. And would love to put a runner up that is contributing at the same time. | 02:42:26 |
| Zhaofeng Li | At this stage, I think doing cross from x86 may be more productive than native builds. For native it's painful sitting through stdenv rebuilds every couple of weeks. | 02:42:37 |
| matthewcroughan - nix.how | You can do cross? | 02:42:51 |
| matthewcroughan - nix.how | I thought that altered the nix store path, meaning it couldn't be used as expected | 02:43:04 |
| matthewcroughan - nix.how | when you use a cross-compiler, that's changing the inputs, therefore the output path right? | 02:43:16 |
| Zhaofeng Li | Yeah, that way you can't do stuff like nix-shell -p | 02:44:35 |
| matthewcroughan - nix.how | There's no way around that, that wouldn't break Nix, is that right? | 02:44:52 |
| matthewcroughan - nix.how | Even if you have a content addressed store, should we always expect the output of a cross-compiler to differ from native? | 02:45:11 |
| Zhaofeng Li | Not really sure for content-addressed store, maybe that's something that can be looked at. With the current setup the derivation names for native and cross are different. | 02:47:59 |
| matthewcroughan - nix.how | So does that mean we can compile images and boot them, but once booted we can't do anything with nix on the board? | 02:48:52 |
| matthewcroughan - nix.how | without waiting for the crawl of stdenv native | 02:49:05 |
| matthewcroughan - nix.how | I mean, the only way to really solve that is to have a proper build farm that just builds natively, right? | 02:49:47 |
| matthewcroughan - nix.how | If I can buy any hardware that helps with that, let me know. | 02:50:03 |
