| 9 Oct 2023 |
 Alyssa Ross | ideally we'd bootstrap from mrustc, but people don't like it | 11:48:32 |
| Alyssa Ross | current policy is not to keep old versions of rust around | 11:48:42 |
| Alyssa Ross | aha https://github.com/rust-lang/rust/pull/113274 | 11:51:48 |
 Lotte (it/its)/Cinny (she/her) θΔ& | In reply to @qyliss:fairydust.space
ideally we'd bootstrap from mrustc, but people don't like it yeah it makes the bootstrap path extremely unpleasant | 11:52:59 |
| Lotte (it/its)/Cinny (she/her) θΔ& | tried it out and it took around a day to get it built on just x86 with two pcs which are decently powerful | 11:53:46 |
| Lotte (it/its)/Cinny (she/her) θΔ& | tried it out and it took around a day to get rust 1.69 built on just x86 with two pcs which are decently powerful | 11:54:21 |
| Lotte (it/its)/Cinny (she/her) θΔ& | imagine that times however often nixos has a mass rebuild lol | 11:54:51 |
| Lotte (it/its)/Cinny (she/her) θΔ& | i think it might be better with cargo2nix since you could relatively easily run that on a server farm | 11:55:34 |
| Lotte (it/its)/Cinny (she/her) θΔ& | vs standard rust bootstrap which can only run on one system period | 11:56:11 |
 raitobezarius | I would not be shocked to use mrustc for mips though | 12:06:55 |
| Alyssa Ross | mips doesn't seem like the best arch to start with given how little CPU power most MIPS machines have… | 12:08:07 |
| Lotte (it/its)/Cinny (she/her) θΔ& | yeah… | 12:11:00 |
| Lotte (it/its)/Cinny (she/her) θΔ& | building a server system on riscv already takes a week lol | 12:11:14 |
| raitobezarius | In reply to @qyliss:fairydust.space
mips doesn't seem like the best arch to start with given how little CPU power most MIPS machines have… right but unsure what would be the best way here | 12:20:38 |
| raitobezarius | I can see easily mrustc bootstrap, keeping old Rust versions around is meh but also second best I suppose, in all cases, it seems like a cross system is better for mips (?) | 12:21:09 |
| Lotte (it/its)/Cinny (she/her) θΔ& | although i haven't figured out a way to express "cross compile from x target to host target" | 12:22:44 |
| Lotte (it/its)/Cinny (she/her) θΔ& | although i haven't figured out a way to express "cross compile from x target to host target" in nixpkgs | 12:22:45 |
| Lotte (it/its)/Cinny (she/her) θΔ& | cross compilation maybe? | 12:22:47 |
| raitobezarius | In reply to @lotte:chir.rs
although i haven't figured out a way to express "cross compile from x target to host target" in nixpkgs in nixpkgs or nixos? | 12:26:21 |
| raitobezarius | you can set up your system as a cross system in nixos by instantiating nixpkgs with crossSystem | 12:26:32 |
| Lotte (it/its)/Cinny (she/her) θΔ& | In reply to @raitobezarius:matrix.org
in nixpkgs or nixos? nixpkgs | 12:27:00 |
| Lotte (it/its)/Cinny (she/her) θΔ& | i was trying to port the rustc bootstrap to aarch64 | 12:27:05 |
| raitobezarius | import <nixpkgs> { localSystem = "x target"; crossSystem = "host target"; } ? | 12:27:19 |
| Lotte (it/its)/Cinny (she/her) θΔ& | mrustc isn't supported iirc and so i just wanted to cross compile a bootstrap compiler from x86 to aarch64 | 12:27:35 |
| Lotte (it/its)/Cinny (she/her) θΔ& | In reply to @raitobezarius:matrix.org
import <nixpkgs> { localSystem = "x target"; crossSystem = "host target"; } ? no, inside of nixpkgs | 12:27:46 |
| raitobezarius | ah no inside of nixpkgs this is an open problem | 12:27:59 |
| Lotte (it/its)/Cinny (she/her) θΔ& | yeah | 12:28:04 |
| raitobezarius | but also I'm not sure this is something we want to do | 12:28:41 |
| Alyssa Ross | I don't think we should keep an old version around for it. I'm going to drop it, and if somebody wants Rust MIPS they should probably work with upstream to make that happen, given the reason it was dropped was LLVM codegen bugs. | 12:28:48 |
| raitobezarius | In reply to @qyliss:fairydust.space
I don't think we should keep an old version around for it. I'm going to drop it, and if somebody wants Rust MIPS they should probably work with upstream to make that happen, given the reason it was dropped was LLVM codegen bugs. Gotcha, seems reasonable | 12:29:11 |
