In reply to @torelli:matrix.org
Hey, how did you find his contact? I only found him on GitHub

In reply to @sinan:sinanmohd.com

I'm interested in Nixpkgs analytics: nixpkgs-review with a time-budget and
adding nixpkgs library networking functions, it will directly help me as I'm
running NixOS on my PC and servers I maintain,

I'd also like to improve support for pppd and stalwart mail (something like
simple-nixos-mailserver, but fully upstreamed), as of now I'm using an out of
tree module for both

I'm comfortable with c, go, nix, /bin/sh, python, oop, and lua. I'm not as good
with python, but I've written python libraries for a networking project before.

In reply to @ss:someonex.net
Hi! Would you like to chat about the time-budget project in some more detail? We can do DM

In reply to @thatliuser:matrix.org

Hi all! My name is Eric and I'm a 3rd year undergraduate Computer Engineering student at UC Irvine. I have been daily driving NixOS for roughly two years now, and I love how reusable my config is across different machines, along with how simple it is to tinker with your system without having to worry about breaking anything you've already set up. I'm intrigued by the "Build GCC libraries and compiler separately" project but I believe I don't fully understand the premise of the project from the GSoC page. Initially I thought the proposal was referring to glibc, but from a preliminary search and what I've previously heard, glibc is already distributed separately from gcc.

• Is the proposal referring to compiling libgcc separately from the main compiler process? (https://github.com/NixOS/nixpkgs/blob/nixos-23.11/pkgs/development/compilers/gcc/common/libgcc.nix)? This feels like the wrong idea (since I don't see how this is more flexible), but I couldn't seem to find much else from looking around at nixpkgs and a brief look into GCC's components.

• It isn't immediately obvious to me why compiling the library separately from the compiler would be helpful for bootstrapping (I'm also not exactly sure what this term refers to. What's being bootstrapped?)

• I (briefly) looked into Exherbo to see what their approach to compiling GCC is, but I can't really tell what they're doing differently from a glance. I found this article (https://www.exherbolinux.org/docs/multiarch-pr.html) which seems to explain a similar concept, but I'm not sure if this is what's being referred to. It's also not a very detailed article about how Exherbo accomplishes this goal.

May I have some more clarification on the goals of this project? I tried to search for an issue mentioned in the nixpkgs repository as well, but I didn't find any relevant results.
I'm very excited by the opportunity to help out my favorite Linux distro and thank you in advance for the feedback!

Hi Eric! Glad you are interested:

1. Yes libgcc, but also the other gcc runtime libraries. (libssp, libstdc++ are two of them off hand)

2. Don't worry about "bootstrapping" too much, just consider that before we build libgcc we may not have a libgcc, etc. A lot of compiler devs are working on regular machines and doing builds where they already have 1 copy of everything, which allows being more sloppy about depedencies and breaking cycles. Building compiler and libraries together is bad for a few reasons, but those reasons might mainly be in my head haha. Examples:

   • cross compiling mean can mean the libraries are for one platform and the compiler itself is for another. But having the installed outputs of a package involve two platforms that way is really messy and breaks assumptions in other packaging infra
   • That fact that some of the libraries do need a libc, and the libc is a separate package, means we have zig-zag problem (finishing the compiler build depeneds on downstream package build which depends on the compielr build) we can get around this by building things twice but that is ugly and wasteful

Hello, my name is Chris and I'm an open source Rust developer. This is my third year writing software, and second year using Nix/NixOS for personal and work use. Until 2021 I was a full time music instructor, I learned entirely via online resources and have been working on language compilers and blockchain networks for about 2 years now. I'm interested in taking on a small to medium sized task for GSoC, and the proposals that interest me the most are nixpkgs library networking functions and Nix Internals: Use std::filesystem::path for Path. I also have a personal Nix project I'm working on, though I feel that it's more of a community project than something that would make for a good proposal. Nonetheless, it's called mynixui (My Nix UI), and it aims to give those developing their own desktop environments (or rices) robust scripts written in rust, and allow users to quickly and seamlessly switch between different desktop interfaces. It's currently a flake for convenience, though I could see it being integrated with home-manager later on. Writing Nix is extremely exciting to me, and I've also been using it to package and cross compile binaries for work using crane, lately.

I'm happy to take on any challenge, and I would greatly appreciate the mentorship.

In reply to @siddhant_codes:matrix.org
no problem. I have posted a draft proposal on github (https://github.com/siddhantk232/nix_gsoc) and created a post on discourse (https://discourse.nixos.org/t/request-for-review-of-my-gsoc24-proposal/41721).

In reply to @Ericson2314:matrix.org
it is normal to have e.g. a long-running PR, and then be splitting things out as they are ready for review, for example

In reply to @siddhant_codes:matrix.org
thanks! I'll mention this in my proposal to make it sound more concrete as I haven't written anything about contribution strategy.
Please let me know if there's anything else you'd like me to add/change in the proposal. I'll finish the remaining sections and submit it to google by this weekend.

In reply to @Ericson2314:matrix.org
the idea is that one one hand, we can't really be sure what approach is going to work best, so there just is some fundamental uncertainty. However, on the other hand, we can commit to getting smaller PRs merged throughout

In reply to @Ericson2314:matrix.org
indeed, some of then perhaps should stay std::string because they are for CanonPath

I'd have to do some more reading of the source to understand where and how CanonPath is used but I think I generally got the idea. CanonPath is for nix's view of paths and we want to use std::filesystem::path only where we interact with the real fs so that it's portable on unix and windows.

Thanks for clarifying