| 21 Feb 2022 |
 Ryan Burns | ive been able to run nix built 32 bit executables on same-endian 64 bit machines but thats about as far as ive gone | 22:03:33 |
| 22 Feb 2022 |
 Linux Hackerman is moving: @linus:schreibt.jetzt | Hack of the night: https://github.com/lheckemann/uap-nix. Not particularly exotic, in the sense that everybody has MIPS hardware at home. But I bet not much of it is running nix-built firmware (unless you're telent I guess) :D | 00:01:11 |
 ius | In reply to @zhaofeng:zhaofeng.li
Anyone running powerpc 32-bit? What's the status of support in nixpkgs, and do you have any downstream patches? I was able to build/boot a tiny runit based closure a couple of months ago (i.e. just the bare minimum). I then accidentally plugged a 12 V adapter in the (unmarked) 5 V power jack of my board. | 06:51:30 |
 Zhaofeng Li | RIP, what kind of machine is/was that? | 06:52:25 |
| Ryan Burns | a 5V powerpc SOIC sounds tantalizing | 06:53:06 |
| Zhaofeng Li | Mine is a white-box switch (edge-core 5700-96x) | 06:53:28 |
| ius | It was a 2010'ish commercial board once used for IoT purposes (smart metering for my OEM'd board): http://www.3lectrik.com/greenwave/solutions/connected-lighting-solution/ | 06:56:30 |
| ius | Greenwave Reality Gateway - PPC405EX + 64MB ram | 06:57:04 |
| ius | Was surprised to discover the power people got the device tree thing sorted out before ARM, so PPC4XX 'just works' on a modern kernel even though it shipped with an ancient one. | 07:00:05 |
| Zhaofeng Li | Yeah, the other ppc device I have is an iBook G3 which still has a 3-year-old Gentoo install I haven't touched in a while, and that's my experience as well (OpenFirmware) | 07:02:08 |
| Zhaofeng Li | * Yeah, the other ppc device I have is an iBook G4 which still has a 3-year-old Gentoo install I haven't touched in a while, and that's my experience as well (OpenFirmware)
(edit: actually G4/A1054 - wow so long has passed) | 07:04:23 |
 Jamie | i’ve been working on pentesting a firewall device at work, used a toolchain from nixpkgs to cross-compile some ppd32 tools, worked fine | 07:04:49 |
| ius | I had been eyeing other hardware as a replacement, but nothing that can be had that satisfies at least 2 out of {dirt cheap, decent interfaces, okay'ish clocks} | 07:04:55 |
| Jamie | *ppc32 | 07:04:59 |
 Alyssa Ross | oh, I've just remembered I've built and run programs for powerpc 32-bit NetBSD with Nix | 11:26:09 |
| Alyssa Ross | worked fine apart from requiring a small tweak to nixpkgs to tell it which of NetBSD's multiple ppc32 targets to use | 11:26:44 |
|  Mika Feiler left the room. | 12:02:35 |
| 23 Feb 2022 |
| ius | colemickens 🏳️🌈: I assume you're going to build for RISCV/StarFive using qemu-user? We might want to sort out some way to divide the effort, considering it's really, really slow. | 19:49:56 |
| ius | I think Zhaofeng Li has a public cache up, but I haven't attempted to use it as I'm not quite sure our closures overlap all that much right now (due to qemu related patches). | 19:53:00 |
 [0x4A6F] | ius and colemickens 🏳️🌈, there are also packages inbound for Zhaofeng Li an me. | 19:53:18 |
| ius | Great 🙂 | 19:54:09 |
| Zhaofeng Li | We could have a shared hydra instance that we can all contribute some builders | 19:54:23 |
| ius | I'm building on a Hetzner box right now which I don't mind hooking up to whatever. It's not particularly fast though. | 19:55:11 |
| Zhaofeng Li | If we have a hydra, I guess it will have to be real hardware or qemu-system. qemu-user is pretty hit-and-miss and I don't think there's a good solution to make it automatically retry failing derivations on native builders. | 20:07:40 |
| Zhaofeng Li | On the other hand, qemu-system is indeed very slow (qemu-system on 3900X is barely as fast as U74 in some benchmarks) | 20:09:03 |
| Zhaofeng Li | * If we have a hydra, I guess it will have to be real hardware or qemu-system. qemu-user is pretty hit-and-miss and I don't think there's a good solution to make Hydra automatically retry failing derivations on native builders. | 20:10:02 |
| ius | I'm not sure how bad things are outside of the sdImage closure, but at least that one builds okay for me using qemu-user (and some patches). | 20:18:30 |
| ius | Here's my current tree, in case it's of use to anyone: https://github.com/ius/nixpkgs/commits/riscv64-20220219 | 20:20:04 |
| Zhaofeng Li | Yeah, I have a few "disable tests" commits on my branch as well (I just rebased a couple of days ago, haven't pushed yet). Those can make it hard to reuse stuff from binary caches when everyone is working on different trees | 20:27:51 |
 colemickens | [0x4A6F]: are you packaging the jh7100 tools by chance? That was the only thing I'd found that wasn't in one of the other two repos. | 22:00:23 |
