Lix contributors

Information you might want if you're thinking about contributing to Lix

Running Lix main

Thank you for choosing to help us dogfood in our daily development! If you run into any issues, you can reach out for quick support in #running-main:lix.systems on Matrix.

There is a lot of work-in-progress documentation and a lot of it is work in progress or awaiting move to the wiki. Our apologies for this state, let us know if there is something you need.

If you run into any friction, please let us know. We would like to hear all your complaints.

Getting yourself set up with an account (if desired)

Sign in with GitHub on https://identity.lix.systems.

Note that your email will be visible on Gerrit if you use it, so change it on https://identity.lix.systems if necessary.

A brief tour of the Lix systems

See Information Organisation for where information is.

The Lix sources are developed on Gerrit, built with Buildbot, and released on a Forgejo repo.

Contributor documentation for the project is maintained on this wiki. FIXME(jade): a lot of it is awaiting migration onto the wiki from the private pad system, see tracking issue.

On NixOS/nix-darwin

Use the overlay: https://git.lix.systems/lix-project/nixos-module

Please file bugs if this explodes the build of tooling you use, we can fix it in the overlay.

Flakes

Add Lix to your system configuration like so:

{
  inputs = {
    lix = {
      url = "https://git.lix.systems/lix-project/lix/archive/main.tar.gz";
      flake = false;
    };

    lix-module = {
      url = "https://git.lix.systems/lix-project/nixos-module/archive/main.tar.gz";
      inputs.nixpkgs.follows = "nixpkgs";
      inputs.lix.follows = "lix";
    };
  };
  
  outputs = {nixpkgs, lix-module, lix, ...}: {
    # or equivalent for darwin
    nixosConfigurations.your-box = nixpkgs.lib.nixosSystem {
      system = "x86_64-linux";
      modules = [
        ./machines/your-box
        lix-module.nixosModules.default
      ];
    };
  };
}

You can then update it with nix flake update lix; nix flake update lix-module.

Not flakes

Also supported.

Add inputs for git+https://git.lix.systems/lix-project/lix and git+https://git.lix.systems/lix-project/nixos-module to your preferred pinning tool.

Use in a NixOS module: e.g. imports = [ (import "${your-pinning-thingy.lix-nixos-module}/module.nix" { lix = your-pinning-thingy.lix; }) ];

Niv

Add the sources for the module and Lix itself, using ssh:// after registering your keys with git.lix.systems:

$ niv add git -n lix-nixos-module --repo 'https://git.lix.systems/lix-project/nixos-module'
$ niv add git -n lix-lix --repo 'https://git.lix.systems/lix-project/lix'

Then, import the Lix NixOS module:

  imports = [
    (import "${sources.lix-nixos-module}/module.nix"
      (let lix = sources.lix-lix.outPath;
      in {
        inherit lix;
        versionSuffix =
          "pre${builtins.substring 0 8 lix.lastModifiedDate}-${lix.shortRev}";
      }))
  ];

On other Linux or on macOS

Currently we are still working on the installer (see tracking project). It is possible to convert an existing Nix install to Lix.

flakey-profile

This is experimental. Some people have successfully used it on macOS. We have tested it on an Arch Linux system installed a long time ago with the shell-based installer, and it works fine. This method works by replacing your system profile with one that is built by simple Nix code with flakey-profile.

You can rollback if it blows up by /nix/var/nix/profiles/default-{SECOND-HIGHEST-NUMBER}/bin/nix-env --rollback --profile /nix/var/nix/profiles/default.

Clone https://git.lix.systems/lix-project/nixos-module.git, then, inside it, run sudo nix run --extra-experimental-features 'nix-command flakes' .#system-profile.switch.

Finally, run sudo systemctl daemon-reload && sudo systemctl restart nix-daemon, or, for macOS:

sudo launchctl stop system/org.nixos.nix-daemon
sudo launchctl enable system/org.nixos.nix-daemon
sudo launchctl kickstart -k system/org.nixos.nix-daemon
Restoring a broken install after a macOS update

After updating macOS, you may get error messages like these:

~/.nix-profile: no such file or directory
/nix/var/nix/profile/default: no such file or directory
error: cannot connect to socket at '/nix/var/nix/daemon-socket/socket': Connection refused

You can fix this by opening "Disk Utility" and manually mounting the Nix Volume again. Then, run these commands to re-install the lix daemon:

sudo launchctl load /nix/var/nix/profiles/default/Library/LaunchDaemons/org.nixos.nix-daemon.plist
sudo launchctl kickstart -k system/org.nixos.nix-daemon

Manually, with nix profile

We::Qyriad have used these steps on macOS has it has seemed to work, but we would recommend flakey-profile over it.

$ sudo -H --preserve-env=SSH_AUTH_SOCK nix --experimental-features 'nix-command flakes' profile install --profile /nix/var/nix/profiles/default git+ssh://git@git.lix.systems/lix-project/lix --priority 3

If you then run sudo nix --experimental-features 'nix-command flakes' profile list --profile /nix/var/nix/profiles/default, you should get output similar to this:

Index:              0
Store paths:        /nix/store/8ma7xas2nb0i3lq8mm7fpgalv94s8pzh-nss-cacert-3.92

Index:              1
Store paths:        /nix/store/53r8ay20mygy2sifn7j2p8wjqlx2kxik-nix-2.19.2

Index:              2
Flake attribute:    packages.aarch64-darwin.default
Original flake URL: git+ssh://git@git.lix.systems/lix-project/lix
Locked flake URL:   git+ssh://git@git.lix.systems/lix-project/lix?ref=refs/heads/main&rev=98b497a1a43a4ff39263ed5259f12c5d00b4d8c0
Store paths:        /nix/store/8040hxr4rr8bpb5yp4b48709x3qs4bwb-nix-2.90.0

You may then use sudo nix --experimental-features 'nix-command flakes' profile remove --profile /nix/var/nix/profiles/default 1 to remove your original installation of Nix. This is (probably) optional.

Verification

You should now get something like the following:

~ » nix --version
nix (Nix) 2.90.0-lixpre20240324-f86b965

Information organisation

Lix has a lot of information as a project, and we want to make it accessible in a way that it can be found later if necessary.

There are various tools for keeping information in the project, and they have different purposes

Chat

Chat is good for:

The chat is expected to move way too fast to follow. As such:

Wiki (https://wiki.lix.systems)

The wiki is good for:

The wiki is not good for:

Markdown files in the Lix repo

Markdown files in the Lix repo are good for:

Markdown files in the Lix repo are bad for:

Forgejo issues (https://git.lix.systems)

Our primary issue tracker is Forgejo issues.

We are currently attempting to use the Forgejo project boards feature to communicate what people are working on; it may be replaced with better Kanban software in the future. When making project boards on Forgejo, make them on the lix-project organisation unless they are strictly contained within one project.

The issue tracker is good for:

The issue tracker is not good for:

Where to put an issue

Gerrit (https://gerrit.lix.systems)

Gerrit is good for:

Gerrit is not good for:

Pad (https://pad.lix.systems [private])

We anticipate that the pad service will be semi-permanently private by default, since it doesn't support ACLs.

The pad is good for:

The pad is not good for:

It is very important that pads remain temporary in nature, and are quickly moved to their final destination, e.g. the Wiki.

N.B. For users who aren't in the Lix core team, the service returns 500 when you attempt to login. This is a known issue that can't be fixed.

Why Lix?

(page under construction destruction. editor's note: parts of https://pad.lix.systems/lix-manifesto (PRIVATE) are ported, parts need review before posting here.)

(editor's note (ii): this page wants to be a contributor facing page, as opposed to the website page that maybe will have more general info?)

FAQ

What is Lix anyway?

Lix is a fork of CppNix 2.18, focused on stability and the user experience of both users and contributors. We want to create a safe platform to move Nix technology forward, as a piece of critical infrastructure.

To this end, we have instituted a freeze on the core, where we apply high standards to changes to the core of the system and pursue testing and stability as our first priority on the core. Our long term vision is to shrink and decouple parts of the core, and move features like Flakes to the periphery of the system.

Style Guide

Not just about code, a style guide is a list of decisions we've made, that we want to be consistent about going forwards. It does not need to be comprehensive of all possible issues, nor does it need to confine itself to trivial topics such as formatting. It's a tool for ourselves so we don't forget where we've been, and can avoid solving the same problems again.

Don't be shy about adding to it. Things written here do impose some burden, but the hope is that they lessen other burdens in the long run. Use your judgement about what's worth it.

Please fix style issues in existing code as you encounter them. Style is aspirational, a journey not a destination. :)

Style Guide

Language and terminology

Language

Most existing Lix documentation is written in British English. We intend to continue with that.

Terminology

(FIXME: unsure if this should be in the style guide but ... it kinda should be -jade)

Style Guide

Code

Code changes

Tests

If at all practicable, all new code should be tested to some extent. If writing a test is hard, we need to prioritize making it easier, and potentially block features if that is the case.

Documentation

Reference documentation should be added, in addition to release notes (doc/manual/rl-next-dev), for user visible changes.

For notable dev facing changes, consider adding release notes in doc/manual/rl-next-dev. This is not critical for all changes; in some cases it may make more sense to write it up in dev documentation instead, and indeed it may be ok to defer writing that dev documentation (it's helpful to create an issue to not forget).

Benchmarking

Changes that touch the core of the evaluator or other performance critical code in Lix should be benchmarked.

See bench/README.md for instructions.

Changelist size

If a CL is too long to review, it should be split up into smaller pieces with tests. The exact length varies but passing the 1000 line mark should give significant thought to splitting.

Commit messages

Include at least a sentence or two as to why you are making a commit. For example, it can be nice to have the reproduction of a bug in the commit message. The commit message is the message for your review.

There's no particular format or specific style for commit messages; just make sure they're descriptive and informative.

C++

While we hope to migrate the Lix interpreter from C++ to Rust eventually, C++ is a language that is likely to exist for a long time, and we may end up having to use it in other contexts.

Lix is a C++20 codebase. Features of C++20 that compile on all supported platforms can be used.

NULL vs nullptr

nullptr where at all possible.

Static vs anonymous namespace

Prefer anonymous namespace, both currently exist in the codebase (jade: any other opinions?).

Type Aliases with typedef vs using

Prefer using declarations, as they can be used in more places, can be templatized, and have clearer syntax. Both currently exist in the codebase. (Qyriad)

TODO/FIXME/XXX Comments

jade: this is not consistent with the conventions I use, needs further discussion imo (TODO: block in pre-commit hook, used in local tree but should never pass code review, FIXME(name||feature): its busted, someone should go fix it later, XXX: this is bad, we are writing down that it is ugly but leaving it as-is as we didn't figure out a better way)

Something along the lines of:

Header files

Filenames

Headers should end with .hh. This reduces the likelihood anyone will try to include them from C files, which would require following the rules of both languages and is easy to get wrong.

The implementation of the functions declared in a .hh file should be in a .cc file of the same name, absent reasons to do otherwise.

Order-independence

Headers should not care what order they're loaded in.

The exception, for now, is config.h in the lix repo. This must always come before all other headers. This observation should not be taken to imply it must always be that way, but at the moment it's helpful to be aware of.

Idempotence

Use #pragma once, it helps. You can see this in most existing header files.

///@file and header documentation

///@file should be at the top of all nix headers - Doxygen and other tools use it to decide whether a header should have documentation generated for definitions in it. See the relevant Doxygen documentation for more details.

Strongly consider adding a description of the purpose of a header file at the top of it in with @brief A sentence saying what it is for.

Examples:

/**
 * @file
 * @brief This header is for meow meow cat noises.
 */

/// @file
/// @brief meow meow meow

Source files

Filenames

Source files should end with .cc.

Python

Python is a widely used tooling language in the Lix codebase. It's used in the test suite in functional2, a pytest based test suite intending to replace the Bash functional test suite. It's what the release engineering in releng/ is written in.

In general, follow PEP-8 while writing Python code: identifier style should be the typical Python one per PEP-8, indents are 4 spaces, etc. In the near future, we will autoformat Python code with ruff.

Type annotations

Type annotations should be used where they are helpful, and should be used for parameter and return types of functions. If it is too much of a bother to convince the type checker of something, it is acceptable to put an Any there.

Prefer inserting Any if the alternative is lying to the type checker via # type: ignore comments.

dataclasses is a useful library that produces nicer, more typed, code, and it is widely used. It's the correct choice for most classes which are just blobs of data.

Doc strings

Please write them. Single-line ones are better than none at all.

We use the Sphinx docstring format for parameter documentation, though we currently do not generate Sphinx docs for our Python codebases. See this Sphinx tutorial for an example and the Sphinx domains reference for reference.

Nix language

Unsurprisingly Nix contains Nix code. Some amount is tests and a lot is packaging.

We use the nixfmt formatter on files outside the test suite. It's run through treefmt with pre-commit hooks. Nix code outside the test suite is expected to be formatted.

Test suite files need not be formatted with the formatter at this time, but please consider doing so with new tests that don't rely on formatting.

with

Prefer not to use with to bring things into scope as it obscures the source of variables and degrades language server diagnostics.

Use let inherit (attrset) attrs instead.

Meson

Generally based on the style in Meson's docs made consistent and with a couple tweaks; notably multiline function calls are done in "block style" (think like rustfmt does it), rather than aligned, e.g.:

executable('sdlprog', 'sdlprog.c',
  win_subsystem : 'windows',
  dependencies : sdl2_dep,
)

rather than:

executable('sdlprog', 'sdlprog.c',
           win_subsystem : 'windows',
           dependencies : sdl2_dep)

Meson's docs go back and forth on this, but we also put a space before and after the colon for keyword arguments (so win_subsystem : 'windows', rather than win_subsystem: 'windows').

Style Guide

Operations

Operational Conventions

Code Review

Self Stamping and Merging

On our Gerrit, core members have permissions to +2 any arbitrary CL, because sometimes we should be able to get something in quickly, and talk about it afterwards. In almost every case, the author of a change should not +2 their own CL, however Lix members may use their best judgement so long as they talk about it with the team when they can. Some cases where skipping synchronous review is a good idea:

Just make sure to talk about what you do :)

Gerrit

Everything about Gerrit: how to get started, advanced tips and tricks, and more!

Gerrit

Getting Started with Gerrit

Thanks for showing interest in contributing to Lix! Gerrit can seem daunting at first, but it is our hope that you'll learn to navigate it and use it confidently after finishing this tutorial.

Perhaps the first question you have is: "Why Gerrit"? Well, glad you asked! In fact, we have an entire page describing that. But in short: it's just very nice for working with code. Instead of "PR from a branch" model that Github uses, Gerrit assigns a Change-Id to a commit, which is preserved even when the commit itself changes. This is perhaps familiar to Jujutsu VCS users.

This model allows us to largely ignore branches, which have long been known to introduce many operational complexities. Instead, we can focus on getting the work done, with our tooling supporting us in this goal!

Setting Up

Here are a few things you will need. Let's go over them one by one.

  1. Go to our Gerrit instance and sign in. It will prompt you to sign in with Github SSO. If you don't want to use your Github account, that's fine; just hit us up on Matrix and we'll create you an account!
  2. Get your SSH key ready! If you don't have one, run ssh-keygen -t ed25519 to generate the SSH keypair with ed25519 cypher. It is the most secure SSH cypher currently available, and quite ergonomic to use! Passphrase isn't necessary for Gerrit, but if you want to use it, we recommend to also set up ssh-agent.
  3. Go to your user settings in Gerrit, scroll down to "SSH keys", and add your public SSH key (from ~/.ssh/id_ed25519.pub).

Configuring Git Repo

Now, we need to make some setup for the Git repo. If you haven't yet cloned Lix repo, use this fancy one-liner to do it (just remember to change USERNAME to your actual username!):

git clone "ssh://USERNAME@gerrit.lix.systems:2022/lix" && (cd "lix" && mkdir -p `git rev-parse --git-dir`/hooks/ && curl -Lo `git rev-parse --git-dir`/hooks/commit-msg https://gerrit.lix.systems/tools/hooks/commit-msg && chmod +x `git rev-parse --git-dir`/hooks/commit-msg)

If you want to get a deeper understanding, or you already cloned the repo from Forgejo - read on!

This is a little awkward, but yes - we store a Git repo in both Forgejo and Gerrit. This is a technical limitation of our infra. The "actual" repo lives in Gerrit - and it is automatically mirrored to Forgejo. This means that we want to push changes directly to Gerrit.

  1. If you cloned the repository from Forgejo, rewrite the origin to point to Gerrit instead: git remote set-url origin ssh://USERNAME@gerrit.lix.systems:2022/lix
  2. Gerrit wants a Change-Id footer in your commit message to work (and track changes to your commit). Adding it by hand is very inconvenient; thankfully, there's a Git hook that adds this footer if it doesn't already exist, even on amends of existing commits. It is added automatically if you are in nix develop shell. Otherwise, you can add it manually:
 mkdir -p .git/hooks
 curl -Lo .git/hooks/commit-msg https://gerrit.lix.systems/tools/hooks/commit-msg
 chmod +x .git/hooks/commit-msg
  1. Now you can just git commit your change. No need to create a separate branch - your commit is the unit of review here!
  2. Gerrit also expects you to push in an unusual way, too: git push origin HEAD:refs/for/main. Just run this command to automate it: git config remote.origin.push HEAD:refs/for/main. Now you'll be able to git push like normal!
  3. Now, you can just git push, and you'll see a link to your CL (Change List)! Now pat yourself on the back and wait for review :)

What Next?

Now, you should be getting a review in a few days - especially if your CL is quite small. Generally, Lix team goes through the open CLs quite often, and helps the new ones get to completion. Sometimes, people might be busy or just miss your CL - so don't take it to heart, and ask in Matrix for a review!

Once you got a review, you can address some issues - and fix others. When fixing issues, don't create new commits - use git commit -a instead to amend the first commit that you pushed to Gerrit initially. Remember, each commit is a separate review piece - so unless you want to create a new CL, just amend the existing commit!

When the review process is done and everybody is satisfied, you'll get approvals from maintainers, and your change will become "Ready to submit". Now is your time to shine! Unlike Github, Gitlab or Forgejo, maintainers aren't the ones who can press the "Submit" button. You can get a last look at the change - and when you decide you're ready, push the "Submit" button, and your change will be part of the codebase. Congratulations!

Now, perhaps this experience was too easy, and you want to learn more Gerrit. Fair enough! Here are some pointers for you:

Good luck on your journey, and thanks for your contribution!

Gerrit

Review Flow

The review flow is quite similar to how Github does it, but there are a few differences here too. The UI also hides a few pretty powerful features!

Reviewer Flow

You can scroll down to the changes and start a review:

You can click on a line number to add a comment. Like on Github, the comments will be saved in a "draft" state, and will be submitted all at once when you finish the review.

When you expand the change to a file, the file will be marked as "reviewed" automatically - for your convenience. If you press "Expand All" button, you'll have to mark files as "reviewed" manually.

Finally, you can also change the "revisions" of comparison: you can choose to review the diff between Patchset [version] 3 and Patchset [version] 4! FIXME: this might be done automatically on re-review?

Finally, when the review is done, you scroll up and press the "Reply" button to submit a review, where you can add a final comment:

There, it's possible to mark some comments as resolved (maybe they're a non-important nitpick!), and also add a Code-Review vote. This is similar to Github's "approve/comment/request changes" on a PR, but a little more flexible. For a change to be ready for merge, it needs +2 on Code Review. Lix maintainers can +2 your change by themselves, so in normal circumstances, your change will require maintainer approval or 2 reviews from non-maintainers.

Reviewee Flow

There are two ways to view the review. One is to look at it directly in code:

Screenshot from 2025-03-29 13-13-12.png

This will show you the whole comment thread.

Another way is to scroll down to the change log and expand a review:

This will show you the entire review at once, with threads collapsed.

Choose whichever you like best!

Merge Flow

Once approved, the change can be merged. Maintainers will sometimes do that, but we generally prefer to leave the decision to the person who submitted the change! Take a final look at your change, verify that everything looks good, and press the "Submit" button!

Gerrit

Why Gerrit?

Gerrit produces better code:

Gerrit makes reviewers' lives easier and reduces review round trips:

Gerrit makes your life easier as a contributor:

That being said, there are some downsides:

Gerrit

Gerrit with Jujutsu

Jujutsu has a very similar model to Gerrit! How nice it would be to integrate them!

Unfortunately, Jujutsu currently doesn't support Gerrit natively. You'll have to rely on a couple of workarounds to get it working.

  1. You need to colocate with Git, because Jujutsu doesn't quite understand Gerrit's push semantics.
  2. You need to add Change-Id footer that is required by Gerrit. Jujutsu doesn't support git hooks; you may use regular Git to amend the commit message to have this footer. Cleaner solution is to teach Jujutsu to do this: see this (untested) example on how to teach Jujutsu to append Change-Id footer.
  3. You need to git push using regular git command.

See this issue for tracking progress.

Gerrit

Tips and Tricks

SSH Tuning

Add these lines to your ~/.ssh/config:

Host gerrit.lix.systems
    User YOUR_GERRIT_USERNAME
    Port 2022
    # Keep sessions open for a bit in the background to make connections faster:
    ControlMaster auto
    ControlPath /tmp/ssh-%r@%h:%p
    ControlPersist 120

Now you can use ssh://gerrit.lix.systems/lix instead of ssh://USERNAME@gerrit.lix.systems:2022/lix URL, and have faster iterations on a change!

Splitting Commits

Sometimes a commit that was supposed to be a single feature gets out of hand. You have options. git rebase -i is a "default" suggestion, but you might want to look at git-revise or even at Jujutsu, which has a pretty cool squash workflow to do this!

Gerrit

Advanced Features

If you feel confident in your Gerrit-fu, this page is for you. Perhaps you've already noticed how Gerrit brings a few good improvements to the workflow just based on the commit-centric design. Now we are getting to the really cool stuff: Gerrit-specific features that further enhance your experience!

WIP Changes

Want to mark a change as WIP? You can do it in the Web UI. Or you can do it immediately on push:

git push origin HEAD:refs/for/main%wip

Submitting Chains of Changes

In traditional Git forges, it is hard to submit multiple changes to one repo at once. Perhaps you want to avoid unnecessary rebuilds - or broken intermediate states. You might have to coordinate with people, plan downtime, or create meta-PRs to do that.

Thankfully, Gerrit has topics for that! Assign a topic to necessary CLs, or set it on push:

git push origin HEAD:refs/for/master%topic=<username>-<ISO8601 date>-<topic name>

For example:

git push origin HEAD:refs/for/master%topic=kfearsoff-2025-03-29-rebranding

Why the convoluted naming scheme? Unfortunately, Gerrit doesn't namespace topics, so we do a funny naming scheme to prevent accidental collisions. We are truly sorry for this.

Once all changes in a topic are reviewed, they can be gracefully sent all at once!

Syncing Multiple Repos

In traditional Git forges, it is always a huge pain when you need to submit a batch of changes at once to multiple repos.

Topics to the rescue, again! Just follow the instructions for chains of changes, but in multiple repos. Isn't that awesome?

Gerrit

Troubleshooting

"Remote Unpack Failed" on push

Run a git fetch and try again.

Re-run a CI

Do an empty commit amend: git commit -a. This will change the commit ID (because the date has changed!), so it will now be a new CL revision: you can git push it and pray for good CI run!

Working in the Lix codebase

See also: See also: https://git.lix.systems/lix-project/lix/src/branch/main/doc/manual/src/contributing/hacking.md

Working in the Lix codebase

Codebase overview

The Lix system is constituted of two broad parts, the evaluator and the store daemon. The two pieces may run on the same machine or on different machines.

For example, in a remote build setup like https://hydra.nixos.org, one node is running several evaluators in parallel, and builds are dispatched to several builder nodes.

(fyi to anyone editing this, double click the image in the preview to edit it)

Evaluator

The evaluator is an AST tree-walking evaluator with lazy semantics.

Notable files:

Known design flaws

Lix team plans

Store protocol

The store protocol is a hand rolled binary protocol with monotonically increasing versioning. It runs over a few different transports such as ssh (src/libstore/ssh-store.cc) or Unix sockets (src/libstore/uds-remote-store.cc).

Known design flaws

Lix team plans

Store daemon

The store daemon takes derivations (≈ execve args and dependencies) and realises (builds or substitutes) them. It also implements store path garbage collection.

Lix's local store implementation currently uses a SQLite database as the source of truth for mapping derivation outputs to derivations as well as maintaining derivation metadata.

Notable files:

Known design flaws

Lix team plans

Working in the Lix codebase

Freezes and recommended contributions

Suggested contributions

Consider taking an issue marked E-help wanted: assign it to yourself and have a go. Feel free to ask for help in the development channel. The Lix team wants these issues fixed, but they are not high on our agenda to fix ourselves.

When in doubt, please ask the Lix team before beginning work, to make sure it is in line with our current priorities.

New builtins

One thing that is a little bit tricky and can get contributions canned late in the process is new builtins. We should write up a more full document on this, but the gist is that the API of any builtins needs to stand the test of time as they have pretty strong compatibility and usefulness expectations.

As such, it would be greatly appreciated if work on builtins starts with an issue on lix-project/lix tagged RFD discussing the use case for the builtin, then, once there is rough agreement on the use case, write a simple document giving examples and design of the proposed API. Then, we can gather feedback before too much time is put into implementations that might not see the light of day.

I don't wanna write C++. How can I help?

This is totally reasonable, C++ is not super fun, but this isn't to say there isn't help if you do choose to. Nonetheless, there are still many issues on our tracker that do not require writing any C++ that would help out immensely.

Here's some possible places to look:

Freezes

This document describes the state of freeze that Lix is in.

We do expect to have main always be in a state to run on machines you care about, unconditionally. Nightly builds should not be a problem to run in production in any freeze state.

The purpose of this policy is to set expectations of what we are looking for in contributions, rather than to set hard rules.

Lix is currently in status "milkshake".

ice cube

No major features or code changes are accepted that touch the core (e.g. the hot paths of the evaluator, the store), absent a good justification, good test coverage and a strong belief that they will not cause regressions. In this state, we don't recommend external contributors do substantive work outside the roadmap without speaking with the Lix team first. However, this is a guideline, and such work can be done if discussed and planned carefully beforehand.

Simple surface features with low impact are likely to be accepted with tests, assuming that they do not impact reliability.

New tests are gleefully accepted.

Bug fixes (with tests) are gleefully accepted.

For example, the following would require discussion with the Lix team before work begins, as it is likely to not fit our goals:

For example, the following would likely be accepted assuming it has tests, without needing prior discussion:

hard ice cream

Changes that improve maintainability of the core are accepted, with careful review depending on their significance. Changes that add more complexity to the core need to pass scrutiny.

Features at the edge are accepted if they have low impact, assuming that they have tests.

soft serve

FIXME

milkshake

All kinds of changes are acceptable, but we still strive to keep main always as stable as possible and a safe decision to daily-drive for all your nixy needs. Please don't jam the ice cream machine!

Working in the Lix codebase

Bug tracker organisation

We have a repo of directly imported nix bugs at https://git.lix.systems/nixos/nix. Please don't file bugs in there, we want the IDs to match. When we import a bug, we might put notes on there as we triage it, and potentially close it.

Bug labels on NixOS/nix

Dispositions:

Closed, marginal

Working in the Lix codebase

Gerrit

What is Gerrit and why do people like it?

Gerrit is a code review system from Google in a similar style to Google's internal Critique tool, but based on Git, and publicly available as open source. It hosts a Git repo with the ability to submit changes for review and offers mirroring to other repos (like https://git.lix.systems/lix-project/lix). It has an entirely different review model to GitHub (and Forgejo, GitLab, etc that copy GitHub's review model), where, instead of pull requests, you have changelists (CLs): reviews on individual commits, with each revision of a commit being a different "patchset", rather than reviewing an entire branch at a time. CLs may be merged one by one or in a batch.

Although this has some learning curve, we expect that you will find it pleasant to work with after figuring it out. It has some rough edges and strong opinions that take some getting used to, but it has served us well and saved us an inordinate amount of time both as reviewers and change authors. The rest of this document gives some pointers on the workflows we use with Gerrit.

People like Gerrit because it makes the following things trivial or easy, all of which are somewhere between annoying and impossible on GitHub modeled systems:

Gerrit produces better code:

Gerrit makes reviewers' lives easier and reduces review round trips:

Gerrit makes your life easier as a contributor:

That being said, there are some downsides:

Learning materials

Our installation

Gerrit is at https://gerrit.lix.systems

The Gerrit SSH server is running on port 2022. The repo URLs are:

Hit the d key on any change to download it, which will give you the right URLs.

SSH config

You might like to add the following configuration to your ~/.ssh/config:

Host gerrit.lix.systems
    User YOUR_GERRIT_USERNAME
    Port 2022
    # Keep sessions open for a bit in the background to make connections faster:
    ControlMaster auto
    ControlPath /tmp/ssh-%r@%h:%p
    ControlPersist 120

Basic workflow for a change

The unit of code review is a "change", which yields a single commit when "submitted" (merged). The commit message is taken from the change description in Gerrit; in our experience this tends to lead to more comprehensive commit messages.

For a change to be merged, it must have the following four "votes", in Gerrit's terminology:

If you're newly part of the core team you will need to add yourself to the Gerrit lix group, otherwise you can't set the Has-Release-Notes or Has-Tests labels. If you're not, this doesn't affect you.

When all of those labels are set, a change becomes Ready to submit, in Gerrit's termology, and Gerrit will give you a Submit button in the top right: Screenshot_20240507_165352.png

By convention, the change author has the final say on clicking the Submit button (note: this is the opposite of the Github convention), and there is no special permission to merge a change once it has been fully reviewed (the permissions are in the reviewer +2'ing it). This gives you a last chance to have a look at your change before merging it.

Workflow tips

Local branches and commits

Gerrit is very mean to you if you don't have your local commit history in a linear presentable state, which takes getting used to but it is very low overhead once you get used to it. In short, amended commits become "patchsets", new commits become changes, and multiple commits help link your changes together as a "relation chain".

Note: if you're coming from Chromium, this is different to how they use Gerrit, where multiple commits become patchsets, and only the first commit on a local branch creates a new change.

Gerrit’s commit-msg hook generates a new Change-Id for each commit you make, which in turn creates a new change that gets reviewed separately. To update an existing change after review feedback, amend or squash your changes into your old commit, keeping its Change-Id unchanged, then push.

Consider not pushing for review before it is clean, or split commits up with git-revise (good) or jj (better) after the fact, amending as you work. If you want a backup of your changes, you can fork it on Forgejo and push to that fork.

Basic Pushing

If you cloned the repo from Forgejo, be sure to change your remote URL to point to Gerrit before continuing. Assuming your remote is called origin (which is the default):

git remote set-url origin ssh://{username}@gerrit.lix.systems:2022/lix

Then you can push to Gerrit with:

git push origin HEAD:refs/for/main

If you get tired of doing this every time, you can automate it by setting the .git/config as follows:

git config remote.origin.push HEAD:refs/for/main

You will have to do that in each fresh check-out. Once it's done, git push will work without additional options.

If you get a “remote unpack failed” error while pushing, run git fetch then try again.

If you wish to push a change and immediately mark it as WIP, you can push with -o wip, or make that the default behavior by checking Set new changes to "work in progress" by default in Gerrit's user settings, under "Preferences".

Topics & Push Arguments

A Gerrit topic may be set on push with:

git push origin HEAD:refs/for/main%topic=foo

Which will create all pushed changes with the topic "foo". Topics are helpful for grouping long series of related changes.

A change may also be marked as "work in progress" on push:

git push origin HEAD:refs/for/main%wip

Gerrit has documentation on other push arguments you can use here, but it also takes a help argument whose output is more canonical and might be easier to understand, which you can view with:

git push origin HEAD:refs/for/main%help

At the time of this writing (2024/06/26), that output looks like this:

$ git push origin @:refs/for/main%help
Total 0 (delta 0), reused 0 (delta 0), pack-reused 0 (from 0)
remote: Processing changes: refs: 1, done
remote:
remote: Help for refs/for/branch:
remote:
remote:  --base BASE                            : merge base of changes
remote:  --cc CC                                : add CC to changes
remote:  --create-cod-token                     : create a token for consistency-on-dema
remote:                                           nd (default: false)
remote:  --deadline NAME                        : deadline after which the push should
remote:                                           be aborted
remote:  --edit (-e)                            : upload as change edit (default: false)
remote:  --hashtag (-t) HASHTAG                 : add hashtag to changes
remote:  --help (-h)                            : display this help text (default: true)
remote:  --ignore-automatic-attention-set-rules : do not change the attention set on
remote:  (-ias, -ignore-attention-set)            this push (default: false)
remote:  --label (-l) LABEL+VALUE               : label(s) to assign (defaults to +1 if
remote:                                           no value provided)
remote:  --merged                               : create single change for a merged
remote:                                           commit (default: false)
remote:  --message (-m) MESSAGE                 : Comment message to apply to the review
remote:  --no-publish-comments (--np)           : do not publish draft comments
remote:                                           (default: false)
remote:  --notify [NONE | OWNER |               : Notify handling that defines to whom
remote:  OWNER_REVIEWERS | ALL]                   email notifications should be sent.
remote:                                           Allowed values are NONE, OWNER,
remote:                                           OWNER_REVIEWERS, ALL. If not set, the
remote:                                           default is ALL.
remote:  --notify-bcc USER                      : user that should be BCC'd one time by
remote:                                           email
remote:  --notify-cc USER                       : user that should be CC'd one time by
remote:                                           email
remote:  --notify-to USER                       : user that should be notified one time
remote:                                           by email
remote:  --private                              : mark new/updated change as private
remote:                                           (default: false)
remote:  --publish-comments                     : publish all draft comments on updated
remote:                                           changes (default: false)
remote:  --ready                                : mark change as ready (default: false)
remote:  --remove-private                       : remove privacy flag from updated
remote:                                           change (default: false)
remote:  --reviewer (-r) REVIEWER               : add reviewer to changes
remote:  --skip-validation                      : skips commit validation (default:
remote:                                           false)
remote:  --submit                               : immediately submit the change
remote:                                           (default: false)
remote:  --topic NAME                           : attach topic to changes
remote:  --trace NAME                           : enable tracing
remote:  --wip (-work-in-progress)              : mark change as work in progress
remote:                                           (default: false)
remote:
To ssh://gerrit.lix.systems:2022/lix
 ! [remote rejected]     HEAD -> refs/for/main%help (see help)
error: failed to push some refs to 'ssh://gerrit.lix.systems:2022/lix'

Pulling

Pulling from Gerrit will work normally. It's worth keeping in mind that sometimes a CL you're working on has been edited in the web UI or by another contributor, so the commit in your repo isn't the latest. Rebasing will usually make the duplicate go away; this is part of the normal rebase semantics, not Gerrit magic. You might consider making rebase-on-pull your default.

Sandbox branches

This feature has some notable ways to shoot yourself in the foot. We still support it, since it allows for running CI builds on things before they become proper CLs. If you don't need that and don't want to worry about the footguns, consider using a branch on a Forgejo fork for sharing WIP code.

In particular, if a commit is in any branch already including a sb/ branch, it will be rejected with the error "no new changes" if it is later pushed to refs/for/main. This can be worked around by amending all the commits so they are distinct, or by git push origin HEAD:refs/for/main%base=$(git rev-parse origin/main), which forces the merge-base

Use refs/heads/sb/USERNAME/*.

CI rerun

Push the CL again with a no-changes commit amendment if you want to force CI to rerun.

Finding CLs to review

Consider bookmarking: https://gerrit.lix.systems/q/status:open+-is:wip+-author:me+label:Code-Review%3C2

Working in the Lix codebase

Improving build times

Setup

Use a clang stdenv:

nix develop .#native-clangStdenvPackages

Then delete build/ if you were using gcc before. Enable build-time profiling with:

just setup; meson configure build -Dprofile-build=enabled

Then run the build: just compile.

Enabling build-time profiling itself costs about 10% of compile time but has no other disadvantage.

Build time reports

Use maintainers/buildtime_report.sh build/ to generate a build time report. This will tell you where all our build time went by looking at the trace files and producing a badness summary.

Sample report

Build-time report sample 
lix/lix2 » ClangBuildAnalyzer --analyze buildtimeold.bin
Analyzing build trace from 'buildtimeold.bin'...
**** Time summary:
Compilation (551 times):
  Parsing (frontend):         1465.3 s
  Codegen & opts (backend):   1110.9 s

**** Files that took longest to parse (compiler frontend):
 10478 ms: build/src/libstore/liblixstore.so.p/build_local-derivation-goal.cc.o
 10319 ms: build/src/libexpr/liblixexpr.so.p/primops.cc.o
  9947 ms: build/src/nix/nix.p/flake.cc.o
  9850 ms: build/src/libexpr/liblixexpr.so.p/eval.cc.o
  9751 ms: build/src/nix/nix.p/profile.cc.o
  9643 ms: build/src/nix/nix.p/develop.cc.o
  9296 ms: build/src/libcmd/liblixcmd.so.p/installable-attr-path.cc.o
  9286 ms: build/src/libstore/liblixstore.so.p/build_derivation-goal.cc.o
  9208 ms: build/src/libcmd/liblixcmd.so.p/installables.cc.o
  9007 ms: build/src/nix/nix.p/.._nix-env_nix-env.cc.o

**** Files that took longest to codegen (compiler backend):
 24226 ms: build/src/libexpr/liblixexpr.so.p/primops_fromTOML.cc.o
 24019 ms: build/src/libexpr/liblixexpr.so.p/primops.cc.o
 21102 ms: build/src/libstore/liblixstore.so.p/build_local-derivation-goal.cc.o
 16246 ms: build/src/libstore/liblixstore.so.p/store-api.cc.o
 14586 ms: build/src/nix/nix.p/.._nix-build_nix-build.cc.o
 13746 ms: build/src/libexpr/liblixexpr.so.p/eval.cc.o
 13287 ms: build/src/libstore/liblixstore.so.p/binary-cache-store.cc.o
 13263 ms: build/src/nix/nix.p/profile.cc.o
 12970 ms: build/src/nix/nix.p/develop.cc.o
 12621 ms: build/src/libfetchers/liblixfetchers.so.p/github.cc.o

**** Templates that took longest to instantiate:
 42922 ms: nlohmann::basic_json<>::parse<const char *> (69 times, avg 622 ms)
 32180 ms: nlohmann::detail::parser<nlohmann::basic_json<>, nlohmann::detail::i... (69 times, avg 466 ms)
 27337 ms: nix::HintFmt::HintFmt<nix::Uncolored<std::basic_string<char>>> (246 times, avg 111 ms)
 25338 ms: nlohmann::basic_json<>::basic_json (293 times, avg 86 ms)
 23641 ms: nlohmann::detail::parser<nlohmann::basic_json<>, nlohmann::detail::i... (69 times, avg 342 ms)
 20203 ms: boost::basic_format<char>::basic_format (492 times, avg 41 ms)
 17174 ms: nlohmann::basic_json<>::json_value::json_value (368 times, avg 46 ms)
 15603 ms: boost::basic_format<char>::parse (246 times, avg 63 ms)
 13268 ms: std::basic_regex<char>::_M_compile (28 times, avg 473 ms)
 12757 ms: std::__detail::_Compiler<std::regex_traits<char>>::_Compiler (28 times, avg 455 ms)
 10813 ms: std::__detail::_Compiler<std::regex_traits<char>>::_M_disjunction (28 times, avg 386 ms)
 10719 ms: std::__detail::_Compiler<std::regex_traits<char>>::_M_alternative (28 times, avg 382 ms)
 10508 ms: std::__detail::_Compiler<std::regex_traits<char>>::_M_term (28 times, avg 375 ms)
  9516 ms: nlohmann::detail::json_sax_dom_callback_parser<nlohmann::basic_json<... (69 times, avg 137 ms)
  9112 ms: std::__detail::_Compiler<std::regex_traits<char>>::_M_atom (28 times, avg 325 ms)
  8683 ms: std::basic_regex<char>::basic_regex (18 times, avg 482 ms)
  8241 ms: std::operator<=> (438 times, avg 18 ms)
  7561 ms: std::vector<boost::io::detail::format_item<char, std::char_traits<ch... (246 times, avg 30 ms)
  7475 ms: std::vector<boost::io::detail::format_item<char, std::char_traits<ch... (246 times, avg 30 ms)
  7309 ms: std::reverse_iterator<std::_Bit_iterator> (268 times, avg 27 ms)
  7131 ms: boost::stacktrace::basic_stacktrace<>::basic_stacktrace (246 times, avg 28 ms)
  6868 ms: boost::stacktrace::basic_stacktrace<>::init (246 times, avg 27 ms)
  6518 ms: std::reverse_iterator<std::_Bit_const_iterator> (268 times, avg 24 ms)
  5716 ms: std::__detail::_Synth3way::operator()<std::variant<nix::OutputsSpec:... (182 times, avg 31 ms)
  5303 ms: nix::make_ref<nix::SingleDerivedPath, nix::DerivedPathOpaque> (178 times, avg 29 ms)
  5244 ms: std::__uninitialized_move_a<boost::io::detail::format_item<char, std... (246 times, avg 21 ms)
  4857 ms: std::make_shared<nix::SingleDerivedPath, nix::DerivedPathOpaque> (178 times, avg 27 ms)
  4813 ms: std::__detail::_Synth3way::operator()<std::variant<nix::TextIngestio... (158 times, avg 30 ms)
  4648 ms: nlohmann::detail::json_sax_dom_callback_parser<nlohmann::basic_json<... (69 times, avg 67 ms)
  4597 ms: std::basic_regex<char>::basic_regex<std::char_traits<char>, std::all... (10 times, avg 459 ms)

**** Template sets that took longest to instantiate:
 55715 ms: std::__do_visit<$> (3603 times, avg 15 ms)
 47739 ms: std::__detail::__variant::__gen_vtable_impl<$>::__visit_invoke (11132 times, avg 4 ms)
 43338 ms: nlohmann::basic_json<$>::parse<$> (85 times, avg 509 ms)
 43097 ms: std::__detail::__variant::__raw_idx_visit<$> (2435 times, avg 17 ms)
 32390 ms: nlohmann::detail::parser<$>::parse (83 times, avg 390 ms)
 30986 ms: nix::HintFmt::HintFmt<$> (1261 times, avg 24 ms)
 30255 ms: std::__and_<$> (25661 times, avg 1 ms)
 29762 ms: std::unique_ptr<$> (2116 times, avg 14 ms)
 28609 ms: std::__tuple_compare<$>::__eq (2978 times, avg 9 ms)
 27560 ms: nlohmann::detail::parser<$>::sax_parse_internal<$> (167 times, avg 165 ms)
 27239 ms: std::variant<$> (1959 times, avg 13 ms)
 26837 ms: std::__invoke_result<$> (10782 times, avg 2 ms)
 25972 ms: std::tuple<$> (5714 times, avg 4 ms)
 24247 ms: std::__uniq_ptr_data<$> (2116 times, avg 11 ms)
 24061 ms: std::__result_of_impl<$> (9029 times, avg 2 ms)
 23949 ms: std::__uniq_ptr_impl<$> (2116 times, avg 11 ms)
 21185 ms: std::optional<$> (2502 times, avg 8 ms)
 21044 ms: std::pair<$> (4989 times, avg 4 ms)
 20852 ms: std::__or_<$> (24005 times, avg 0 ms)
 20203 ms: boost::basic_format<$>::basic_format (492 times, avg 41 ms)
 20184 ms: std::tie<$> (2895 times, avg 6 ms)
 19938 ms: nlohmann::basic_json<$>::create<$> (668 times, avg 29 ms)
 19798 ms: std::allocator_traits<$>::construct<$> (5720 times, avg 3 ms)
 19182 ms: std::__detail::__variant::_Variant_base<$> (1959 times, avg 9 ms)
 19151 ms: std::_Rb_tree<$>::_M_erase (2320 times, avg 8 ms)
 19094 ms: std::_Rb_tree<$>::~_Rb_tree (2022 times, avg 9 ms)
 18735 ms: nlohmann::basic_json<$>::basic_json (243 times, avg 77 ms)
 18546 ms: std::__detail::_Synth3way::_S_noexcept<$> (2542 times, avg 7 ms)
 17174 ms: nlohmann::basic_json<$>::json_value::json_value (368 times, avg 46 ms)
 17111 ms: nlohmann::detail::conjunction<$> (907 times, avg 18 ms)

**** Functions that took longest to compile:
  2091 ms: _GLOBAL__sub_I_primops.cc (../src/libexpr/primops.cc)
  1799 ms: nix::fetchers::GitInputScheme::fetch(nix::ref<nix::Store>, nix::fetc... (../src/libfetchers/git.cc)
  1388 ms: nix::Settings::Settings() (../src/libstore/globals.cc)
  1244 ms: main_nix_build(int, char**) (../src/nix-build/nix-build.cc)
  1021 ms: nix::LocalDerivationGoal::startBuilder() (../src/libstore/build/local-derivation-goal.cc)
   918 ms: nix::LocalStore::LocalStore(std::map<std::__cxx11::basic_string<char... (../src/libstore/local-store.cc)
   835 ms: opQuery(Globals&, std::__cxx11::list<std::__cxx11::basic_string<char... (../src/nix-env/nix-env.cc)
   733 ms: nix::daemon::performOp(nix::daemon::TunnelLogger*, nix::ref<nix::Sto... (../src/libstore/daemon.cc)
   589 ms: _GLOBAL__sub_I_tests.cc (../tests/unit/libutil/tests.cc)
   578 ms: main_build_remote(int, char**) (../src/build-remote/build-remote.cc)
   522 ms: nix::fetchers::MercurialInputScheme::fetch(nix::ref<nix::Store>, nix... (../src/libfetchers/mercurial.cc)
   521 ms: nix::LocalDerivationGoal::registerOutputs[abi:cxx11]() (../src/libstore/build/local-derivation-goal.cc)
   461 ms: nix::getNameFromURL_getNameFromURL_Test::TestBody() (../tests/unit/libutil/url-name.cc)
   440 ms: nix::Installable::build2(nix::ref<nix::Store>, nix::ref<nix::Store>,... (../src/libcmd/installables.cc)
   392 ms: nix::prim_fetchClosure(nix::EvalState&, nix::PosIdx, nix::Value**, n... (../src/libexpr/primops/fetchClosure.cc)
   390 ms: nix::NixArgs::NixArgs() (../src/nix/main.cc)
   388 ms: update(std::set<std::__cxx11::basic_string<char, std::char_traits<ch... (../src/nix-channel/nix-channel.cc)
   340 ms: _GLOBAL__sub_I_primops.cc (../tests/unit/libexpr/primops.cc)
   332 ms: nix::flake::lockFlake(nix::EvalState&, nix::FlakeRef const&, nix::fl... (../src/libexpr/flake/flake.cc)
   305 ms: _GLOBAL__sub_I_lockfile.cc (../src/libexpr/flake/lockfile.cc)
   300 ms: nix_store::opQuery(std::__cxx11::list<std::__cxx11::basic_string<cha... (../src/nix-store/nix-store.cc)
   296 ms: nix::parseFlakeRefWithFragment(std::__cxx11::basic_string<char, std:... (../src/libexpr/flake/flakeref.cc)
   289 ms: _GLOBAL__sub_I_error_traces.cc (../tests/unit/libexpr/error_traces.cc)
   278 ms: nix::ErrorTraceTest_genericClosure_Test::TestBody() (../tests/unit/libexpr/error_traces.cc)
   274 ms: CmdDevelop::run(nix::ref<nix::Store>, nix::ref<nix::Installable>) (../src/nix/develop.cc)
   269 ms: nix::flake::lockFlake(nix::EvalState&, nix::FlakeRef const&, nix::fl... (../src/libexpr/flake/flake.cc)
   257 ms: nix::NixRepl::processLine(std::__cxx11::basic_string<char, std::char... (../src/libcmd/repl.cc)
   251 ms: nix::derivationStrictInternal(nix::EvalState&, std::__cxx11::basic_s... (../src/libexpr/primops.cc)
   249 ms: toml::result<toml::basic_value<toml::discard_comments, std::unordere... (../src/libexpr/primops/fromTOML.cc)
   238 ms: nix::LocalDerivationGoal::runChild() (../src/libstore/build/local-derivation-goal.cc)

**** Function sets that took longest to compile / optimize:
 10243 ms: std::vector<$>::_M_fill_insert(__gnu_cxx::__normal_iterator<$>, unsi... (190 times, avg 53 ms)
  9752 ms: bool boost::io::detail::parse_printf_directive<$>(__gnu_cxx::__norma... (190 times, avg 51 ms)
  8377 ms: void boost::io::detail::put<$>(boost::io::detail::put_holder<$> cons... (191 times, avg 43 ms)
  5863 ms: boost::basic_format<$>::parse(std::__cxx11::basic_string<$> const&) (190 times, avg 30 ms)
  5660 ms: std::vector<$>::_M_fill_insert(std::_Bit_iterator, unsigned long, bo... (190 times, avg 29 ms)
  4264 ms: non-virtual thunk to boost::wrapexcept<$>::~wrapexcept() (549 times, avg 7 ms)
  4023 ms: std::_Rb_tree<$>::_M_erase(std::_Rb_tree_node<$>*) (1238 times, avg 3 ms)
  3715 ms: boost::stacktrace::detail::to_string_impl_base<boost::stacktrace::de... (166 times, avg 22 ms)
  3705 ms: std::vector<$>::_M_fill_assign(unsigned long, boost::io::detail::for... (190 times, avg 19 ms)
  3326 ms: boost::basic_format<$>::str[abi:cxx11]() const (144 times, avg 23 ms)
  3070 ms: void boost::io::detail::mk_str<$>(std::__cxx11::basic_string<$>&, ch... (191 times, avg 16 ms)
  2839 ms: boost::basic_format<$>::make_or_reuse_data(unsigned long) (190 times, avg 14 ms)
  2321 ms: std::__cxx11::basic_string<$>::_M_replace(unsigned long, unsigned lo... (239 times, avg 9 ms)
  2213 ms: std::_Rb_tree<$>::_M_get_insert_hint_unique_pos(std::_Rb_tree_const_... (203 times, avg 10 ms)
  2200 ms: boost::wrapexcept<$>::~wrapexcept() (549 times, avg 4 ms)
  2093 ms: std::vector<$>::~vector() (574 times, avg 3 ms)
  1894 ms: bool std::__detail::_Compiler<$>::_M_expression_term<$>(std::__detai... (112 times, avg 16 ms)
  1871 ms: int boost::io::detail::upper_bound_from_fstring<$>(std::__cxx11::bas... (190 times, avg 9 ms)
  1867 ms: boost::wrapexcept<$>::clone() const (549 times, avg 3 ms)
  1824 ms: std::_Rb_tree_iterator<$> std::_Rb_tree<$>::_M_emplace_hint_unique<$... (244 times, avg 7 ms)
  1821 ms: toml::result<$> toml::detail::sequence<$>::invoke<$>(toml::detail::l... (93 times, avg 19 ms)
  1814 ms: nlohmann::json_abi_v3_11_2::detail::serializer<$>::dump(nlohmann::js... (39 times, avg 46 ms)
  1799 ms: nix::fetchers::GitInputScheme::fetch(nix::ref<$>, nix::fetchers::Inp... (1 times, avg 1799 ms)
  1771 ms: boost::io::detail::format_item<char, std::char_traits<char>, std::al... (190 times, avg 9 ms)
  1762 ms: std::__detail::_BracketMatcher<$>::_BracketMatcher(std::__detail::_B... (112 times, avg 15 ms)
  1760 ms: std::_Function_handler<$>::_M_manager(std::_Any_data&, std::_Any_dat... (981 times, avg 1 ms)
  1733 ms: std::__detail::_Compiler<$>::_M_quantifier() (28 times, avg 61 ms)
  1694 ms: std::__cxx11::basic_string<$>::_M_mutate(unsigned long, unsigned lon... (251 times, avg 6 ms)
  1650 ms: std::vector<$>::vector(std::vector<$> const&) (210 times, avg 7 ms)
  1650 ms: boost::io::basic_altstringbuf<$>::overflow(int) (190 times, avg 8 ms)

**** Expensive headers:
178153 ms: ../src/libcmd/installable-value.hh (included 52 times, avg 3426 ms), included via:
  40x: command.hh 
  5x: command-installable-value.hh 
  3x: installable-flake.hh 
  2x: <direct include>
  2x: installable-attr-path.hh 

176217 ms: ../src/libutil/error.hh (included 246 times, avg 716 ms), included via:
  36x: command.hh installable-value.hh installables.hh derived-path.hh config.hh experimental-features.hh 
  12x: globals.hh config.hh experimental-features.hh 
  11x: file-system.hh file-descriptor.hh 
  6x: serialise.hh strings.hh 
  6x: <direct include>
  6x: archive.hh serialise.hh strings.hh 
  ...

173243 ms: ../src/libstore/store-api.hh (included 152 times, avg 1139 ms), included via:
  55x: <direct include>
  39x: command.hh installable-value.hh installables.hh 
  7x: libexpr.hh 
  4x: local-store.hh 
  4x: command-installable-value.hh installable-value.hh installables.hh 
  3x: binary-cache-store.hh 
  ...

170482 ms: ../src/libutil/serialise.hh (included 201 times, avg 848 ms), included via:
  37x: command.hh installable-value.hh installables.hh built-path.hh realisation.hh hash.hh 
  14x: store-api.hh nar-info.hh hash.hh 
  11x: <direct include>
  7x: primops.hh eval.hh attr-set.hh nixexpr.hh value.hh source-path.hh archive.hh 
  7x: libexpr.hh value.hh source-path.hh archive.hh 
  6x: fetchers.hh hash.hh 
  ...

169397 ms: ../src/libcmd/installables.hh (included 53 times, avg 3196 ms), included via:
  40x: command.hh installable-value.hh 
  5x: command-installable-value.hh installable-value.hh 
  3x: installable-flake.hh installable-value.hh 
  2x: <direct include>
  1x: installable-derived-path.hh 
  1x: installable-value.hh 
  ...

159740 ms: ../src/libutil/strings.hh (included 221 times, avg 722 ms), included via:
  37x: command.hh installable-value.hh installables.hh built-path.hh realisation.hh hash.hh serialise.hh 
  19x: <direct include>
  14x: store-api.hh nar-info.hh hash.hh serialise.hh 
  11x: serialise.hh 
  7x: primops.hh eval.hh attr-set.hh nixexpr.hh value.hh source-path.hh archive.hh serialise.hh 
  7x: libexpr.hh value.hh source-path.hh archive.hh serialise.hh 
  ...

156796 ms: ../src/libcmd/command.hh (included 51 times, avg 3074 ms), included via:
  42x: <direct include>
  7x: command-installable-value.hh 
  2x: installable-attr-path.hh 

150392 ms: ../src/libutil/types.hh (included 251 times, avg 599 ms), included via:
  36x: command.hh installable-value.hh installables.hh path.hh 
  11x: file-system.hh 
  10x: globals.hh 
  6x: fetchers.hh 
  6x: serialise.hh strings.hh error.hh 
  5x: archive.hh 
  ...

133101 ms: /nix/store/644b90j1vms44nr18yw3520pzkrg4dd1-boost-1.81.0-dev/include/boost/lexical_cast.hpp (included 226 times, avg 588 ms), included via
:
  37x: command.hh installable-value.hh installables.hh built-path.hh realisation.hh hash.hh serialise.hh strings.hh 
  19x: file-system.hh 
  11x: store-api.hh nar-info.hh hash.hh serialise.hh strings.hh 
  7x: primops.hh eval.hh attr-set.hh nixexpr.hh value.hh source-path.hh archive.hh serialise.hh strings.hh 
  7x: libexpr.hh value.hh source-path.hh archive.hh serialise.hh strings.hh 
  6x: eval.hh attr-set.hh nixexpr.hh value.hh source-path.hh archive.hh serialise.hh strings.hh 
  ...

132887 ms: /nix/store/h2abv2l8irqj942i5rq9wbrj42kbsh5y-gcc-12.3.0/include/c++/12.3.0/memory (included 262 times, avg 507 ms), included via:
  36x: command.hh installable-value.hh installables.hh path.hh types.hh ref.hh 
  16x: gtest.h 
  11x: file-system.hh types.hh ref.hh 
  10x: globals.hh types.hh ref.hh 
  10x: json.hpp 
  6x: serialise.hh 
  ...

  done in 0.6s.

Manually looking at traces

Note that the summary in the report can miss details like why one particular header is bad; to find that out, use a trace viewer to inspect the JSON trace file; we suggest rg -t json -uu error\.hh build/ | less to find some .cc trace that the bad header (in this example, error.hh) appears in.

You can look at individual file traces by opening some file like build/src/libcmd/liblixcmd.so.p/command.cc.json in https://ui.perfetto.dev or another Chrome-trace-json compatible trace viewer like Speedscope.

This will produce a flamegraph of the trace (screenshot shows Perfetto):

The most general spans of compile time are at the top, and the constituent spans are shown as you go down.

Successful build time reduction CLs

See the build-time-optimisation Gerrit topic for more related things.

Working in the Lix codebase

Commit porting (cherry-picking) guide

This guide is primarily about CppNix->Lix commit ports and was written when we were doing a lot of those before 2.90. It also applies to Lix->Lix ports, though, those are much easier, often just taking clicking "cherry pick" in the gerrit UI.

Don't forget, using Gerrit is a bit different than other systems.

Lix -> Lix ports

We generally only backport fixes from main if there are severe bugs that users will hit in production, or if we had to patch the derivation in nixpkgs. The only real thing to remember about this is to use git cherry-pick -x. Gerrit will match up the Change-Ids and link the CLs on release branches to the other CLs on release branches or main. Then, push as if any other change: git push origin HEAD:refs/for/release-2.92 (e.g.).

You can also use the cherry-pick button in the Gerrit UI.

CppNix -> Lix ports

Commit style:

GitHub encourages messy branch history, so the cherry-picking procedure can be rather annoying in a codebase like Lix which mandates that every commit passes CI. However, this document describes how we have generally done it.

Also note that we have renamed src/ to lix/, which may or may not confuse git sometimes.

Procedure (focused mostly on CppNix -> Lix)

single commits

try git cherry-pick -x first. if this works, excellent. if not, apply the usual cherry-picking procedures:

full prs

single-commit prs were mostly picked using cherry-pick -x -m1 to keep the association with the upstream pr number for clarity. this implicitly squashes the pr into a single commit so it's only useful for single-commit prs. (some prs that have broken intermediate commits also benefit from this, but see above for that)

when pushing these to gerrit please set a topic like backport-<pr-number> using push options (-o topic=backport-<pr-number> in git push) to delineate one picked pr from a pr that depends on it

Working in the Lix codebase

Misc tips

pyright is not dealing with import paths correctly for functional2

I don't know why it's doing this but it seems to be assuming that the functional2 test suite is relative to the workspace root.

You can jam the following into :CocLocalConfig for vim, and probably the workspace config for VSCode and it should fix it.

{
    "python.analysis.include": ["tests"],
    "python.analysis.extraPaths": ["tests"],
}

buildbot user style to make the pulsing pills bearable

@keyframes pulse_animation {
  0% { transform:scale(.9) }
  50% { transform:scale(1) }
  to { transform:scale(.9) }
}

.pulse {
    animation-duration: 10s !important;
}

FIXME: someone should PR this, now that we have the ability to patch buildbot

run all lix vm tests locally

tests=$(
    nix eval --json --impure \
        --apply '
            let f = n: t:
                if __isAttrs t
                then (if t.type or "" == "derivation"
                    then (if t.system == __currentSystem
                            then [ n ]
                            else [])
                    else __concatMap (m: f "${n}.${m}" t.${m}) (__attrNames t))
                else [];
            in f ".#hydraJobs.tests"
        ' \
        .#hydraJobs.tests \
        | jq -r '.[]'
)

nix build --no-link -L ${tests[@]}

check out current patchset of a cl by git alias

put this in a gitconfig that can configure aliases:

[alias]
	cocl = !\
	 ps=$(\
	   ssh $(git config remote.origin.gerriturl) \
	     gerrit query --format=json --current-patch-set $1 \
        | jq -sr .[0].currentPatchSet.ref \
	 ) && git fetch origin $ps && git checkout FETCH_HEAD && true

then run as git cocl <cl-number>. needs a git config remote.origin.gerriturl gerrit.lix.systems, or some other url that ssh can connect to. (could've extracted it from the remote url but we didn't want to do that much shell)

git stuff

git-revise

git-revise is a cool tool for splitting and shuffling commits in-memory without breaking your working tree. it's great.

It also has some broken stuff with respect to gerrit commit-msg hooks. However, this can be fixed (this is opt-in because some commit-msg hooks make unsound assumptions but the gerrit one should be fine):

# allow gerrit git hooks to run on git-revise
[revise "run-hooks"]
        commit-msg = true

Making git clean clean the stuff that isn't removed by make clean

If you don't want to use git clean -x to remove all git-ignored stuff, but want to remove things that are generated in Lix's build process but aren't removed by make clean, apply this patch: no-ignore-not-cleaned.patch

Working in the Lix codebase

Building Locally

See hacking.md in the Lix repo for the main documentation. Extra tips can go here.

Working in the Lix codebase

RISC-V support

Goal: install lix on a riscv64-linux system

The target is a DevTerm R-01, so it's an AllWinner D1 RISC-V processor @ 1GHz, with 1GB of memory and 32GB of microSD.

We can't run the Lix installer without building it, because there's no canned build for it. So let's try building it natively:

$ rustup
$ git clone https://git.lix.systems/lix-project/lix-installer
$ cd lix-installer
$ RUSTFLAGS="--cfg tokio_unstable" cargo install --path .

This doesn't work because there's some conditional complication^Wcompilation that doesn't cover riscv64. So we need to open self_test.rs and add an entry:

    #[cfg(all(target_os = "linux", target_arch = "riscv64"))]
    const SYSTEM: &str = "riscv64-linux";

At this point, it will, in principle, build. In practice, however, 1GB is just not enough RAM. If you add some swap it'll make it to the last step, but then it wants 1.5GB+ for that. I wouldn't try it on a system with less than 2GB, and ideally more.

Ok, native build is a bust unless I want to let it thrash all night. So let's cross-compile it on ancilla, which is, conveniently, already running nixos.

The nix-installer flake doesn't come with riscv64 cross support, and rather than try to figure it out I just winged it with nix-shell. I am skipping over a lot of false starts and blind alleys here as I ran into things like dependency crates needing a cross-compiling gcc, or rust not having a stdlib on riscv64-musl.

$ git clone https://git.lix.systems/lix-project/lix-installer
$ cd lix-installer
$ $EDITOR shell.nix
with import <nixpkgs> {
  crossSystem.config = "riscv64-unknown-linux-gnu";
};
mkShell {
  nativeBuildInputs = with import <unstable> {}; [ cargo rustup ];
}

$ nix-shell
[long wait for gcc to compile]

$ export RUSTUP_HOME=$PWD/.rustup-home
$ export CARGO_HOME=$PWD/.cargo-home
$ rustup default stable
$ rustup target add riscv64gc-unknown-linux-gnu
$ edit src/self_test.rs
[apply that same patch to SYSTEM]

The build invocation is a bit more complicated here, because we need to tell it where to find the linker:

$ RUSTFLAGS="--cfg tokio_unstable" cargo build \
    --target riscv64gc-unknown-linux-gnu \
    --config target.riscv64gc-unknown-linux-gnu.linker='"riscv64-unknown-linux-gnu-gcc"'
[another long wait]

$ file target/riscv64gc-unknown-linux-gnu/debug/lix-installer
target/riscv64gc-unknown-linux-gnu/debug/lix-installer:
  ELF 64-bit LSB pie executable, UCB RISC-V, RVC, double-float ABI,
  version 1 (SYSV), dynamically linked,
  interpreter /nix/store/g4xam7gr35sziib1zc033xvn1vy9gg8m-glibc-riscv64-unknown-linux-gnu-2.38-44/lib/ld-linux-riscv64-lp64d.so.1,
  for GNU/Linux 4.15.0, with debug_info, not stripped

Since we couldn't do a static musl build it needs the nix ld.so, but we can get around that!

$ scp target/riscv64gc-unknown-linux-gnu/debug/lix-installer root@riscv:.
$ ssh root@riscv
# ./lix-installer
-bash: ./lix-installer: no such file or directory

# ldd ./lix-installer
/nix/store/.../ld-linux-riscv64-lp64d.so.1 => /lib/ld-linux-riscv64-lp64d.so.1
[other output elided]

# /lib/ld-linux-riscv64-lp64d.so.1 ./lix-installer
The Determinate Nix installer (lix variant)
[...]

Sadly we can't actually use it to install, because nix_package_url needs a default value, and on RISC-V, it doesn't have one! It's self_test.rs all over again except it doesn't manifest until runtime.

So, off to src/settings.rs we go. It doesn't need to be a valid URL, just something URL-shaped.

/// Default [`nix_package_url`](CommonSettings::nix_package_url) for unknown platforms
pub const NIX_UNKNOWN_PLATFORM_URL: &str =
    "https://releases.lix.systems/unknown-platform";

    #[cfg_attr(
        all(target_os = "linux", target_arch = "riscv64", feature = "cli"),
        clap(
            default_value = NIX_UNKNOWN_PLATFORM_URL,
        )
    )]

Rebuild, re-push, re-run:

# /lib/ld-linux-riscv64-lp64d.so.1 /opt/lix-installer install linux
Error: 
   0: Planner error
   1: `nix-installer` does not support the `riscv64gc-unknown-linux-gnu` architecture right now

Ok, missed a few places in settings.rs, let's put a quick and dirty hack in there:

            #[cfg(target_os = "linux")]
            (_, OperatingSystem::Linux) => {
                url = NIX_UNKNOWN_PLATFORM_URL;
                nix_build_user_prefix = "nixbld";
                nix_build_user_id_base = 30000;
                nix_build_user_count = 32;
            },

            #[cfg(target_os = "linux")]
            (_, OperatingSystem::Linux) => {
                (InitSystem::Systemd, linux_detect_systemd_started().await)
            },

It also needs a tarball to install; jade_ kindly updated the flake for it to support riscv64, so we just check it out (or, well, check out review branch 1444) and then nix build -L .#nix-riscv64-linux.binaryTarball and away we go.

This, it turns out, also doesn't work, because the installer is hardcoded to expect the directory the tarball contains to start with nix-*. You can either unpack and repack the tarball to meet that requirement, or find all the places in lix-installer that assume that and edit them -- they're in src/action/base/move_unpacked_nix.rs and src/action/base/setup_default_profile.rs.

Finally, this particular kernel lacks seccomp support -- in order to get it working, I had to edit the lix (not lix-installer) package.nix and add (lib.mesonEnable "seccomp-sandboxing" false) to the meson flags.

And with that done, it works!

root@devterm-R01:~# uname -a && nix --version
Linux devterm-R01 5.4.61 #12 PREEMPT Wed Mar 30 14:44:22 CST 2022 riscv64 riscv64 riscv64 GNU/Linux
nix (Lix, like Nix) 2.90.0pre20240613_dirty
Working in the Lix codebase

Branches

The Lix repository contains multiple releases in parallel. The branches work as follows:

Version types

Git tags

Git tags are created with the format 2.90.0.

Docker tags

Working in the Lix codebase

Working Groups

What are they?

Lix doesn't have a very formal development structure - ice cream isn't a stranger to being fluid, after all! But all of the contributors have a different shape - so they are naturally drawn to different parts of the project. Documenting that is useful - for making communication more streamlined, and for a high-level overview on what activities are there!

People working on some part of the project are called "working group". But that's not all! There might be other people that are looking to help that area - they would be interested to join! Some cool projects depend on others, and so people who monitor the prerequisite project would be tracking it. And of course, some projects can be understaffed, in other words: help wanted.

C++ Code

IPC Rewrite

horrors is doing it, @raitobezarius and @jade_ on review

Evaluator Changes

@piegames

Rust Code

Lix Installer

@KFears

Help wanted!

NixLang parser rewrite

@piegames wants to do it, waiting for Rustification of the codebase

Lix Logging

@jade_ wants to do it, blocked on Rustification

@KFears is interested to join

Documentation

See this: https://wiki.lix.systems/books/lix-contributors/page/docs-rewrite-plans

There also exist issues in Lix bug tracker: https://git.lix.systems/lix-project/lix/issues?labels=151

Integration Work

Lix Rustification

@jade_ drives it

@KFears and @piegames are tracking the progress

Infra

@raitobezarius

@KFears is interested to join (a bit :P)

Release Engineering

@jade_ knows a lot, but the project hasn't seen much activity lately due to burnout

@KFears is interested to join

Testing

Lix integration tests are a bit of a mess: some are in pytest, some are in Bash scripts

https://git.lix.systems/lix-project/lix/issues?labels=121

tests/functional2

@Commentator2.0 and @helle mainly, with @piegames, @pennae and others handling reviews and helping out

External Issues

There are a bunch of different ones! Some are integration work, some are Go, some are Java, some are Python, some are NixLang. Take a look!

https://git.lix.systems/lix-project/external-bugs

Working in the Lix codebase

Dealing with CI

Logging in to CI for restarts

Go via https://buildkite.com/sso/lix-project and use your Lix account.

Design planning

This is the place where we plan major changes to various components of the code base, to tackle the big issues. Treat everything as an unimplemented idea unless stated otherwise. Make sure to link to implementation efforts in order to keep track of them.

Ideas and plans are usually written by individuals or groups of individuals, this is not an official roadmap. Anyone in the Lix community may add to it.

Design planning

regexp engine investigation

nix uses libstdc++'s std::regex. it uses whatever version of libstdc++ the host system has.

which it invokes in both std::regex_replace std::regex_match modes.

nix occasionally uses the flags std::regex::extended and std::regex::icase which determine the available features - it's always either no flags, or both of these together. there's also a couple things that use the flag std::regex::ECMAScript. when the constructor is called without a flags parameter, the flags default to std::regex::ECMAScript (see method signature in C++23 32.7.2), so really we have only two cases.

std::cregex_iterator and std::sregex_iterator are used.

there's a header regex-combinators.hh which defines regex::group and regex::list.... and a couple others that are unused. but those are just trivial textual things, not extensions, so we can ignore the file.

getting the C++ standard

someday when C++23 is official you will be able to pirate the PDF. otherwise, you can clone https://github.com/cplusplus/draft and check out the tag n4950 which is the current formally adopted working draft as of 2024-03-14 and is intended to have the same technical content as the final standard. you can then invoke make in the source subdirectory which will produce std.pdf. you will need LaTeX installed. if you're ever not sure which working draft is the one that became a particular version of the standard, Wikipedia will probably tell you...

(personally I install texlive.combined.scheme-full from nixpkgs on all my machines that have room for it, but this is surely more than necessary, it just makes me feel warm and fuzzy -- Irenes)

chapter 32 is the one that documents regular expressions.

open questions that require reading the standard

required functionality

the extended flag, per the C++ standard, "Specifies that the grammar recognized by the regular expression engine shall be that used by extended regular expressions in POSIX.". it references POSIX, Base Definitions and Headers, Section 9.4.

the ECMAScript flag "Specifies that the grammar recognized by the regular expression engine shall be that used by ECMAScript in ECMA-262, as modified in [section 32.12 of the C++ standard]." it references ECMA-262 15.10. the changes in 32.12 are important and probably do create real compatibility issues for us, though fortunately it's only a single page.

if we complete this chart we can use it to assess which existing engines would meet our needs, or how much of a pain in the ass it would be to make a new one

the columns are the two ways it gets invoked

extended + icase ECMAScript
Syntactic constructs -- --
(TODO: fill in every construct here)
Semantics -- --
Case-insensitivity yes ?
(TODO: fill in other behaviors here)
Design planning

Dreams

This page documents the dreams of the Lix team. These are features which we have generally not roadmapped yet, and which may not have complete and thoroughly thought-through plans, and which we would like to think about more completely before implementing. We are writing them down publicly so that others can dream with us.

fixing ux

slaying the hydra

these are problems that make hydra sad

Design planning

Language versioning

This document is extremely a draft. It needs some editing and discussion before it can be made into a useful thing. It's been simply copy pasted out of the pad in its current form.

See also

musings

puck: honestly, having language version as part of a scopedImport-style primop would be funny horrors: we're shitposting about setting language version from the source accessor

horrific writeup

basic mechanism

add a new syntactic element that is only valid at the head of a file and used only to declare language requirements. nix versions that cannot satisfy all requirements must reject this element to situations in which two nix versions parse the same file differently, or even evaluating the same file to different derivation hashes. any kind of comment as used by eg GHC is not viable for nix for this reason.

proposed syntax for the first implementation: use $( $feature: ident )+;

anything ahead of this directive could be either unversioned nix code or versioned nix code (see below for details), but since the directive is only valid at the head of a file or expression this "code" can only be comments. this kind of locks us into supporting the current comment syntax forever, but the comment syntax is rather fine so this won't be a problem.

each feature may declare a syntactical requirement for the file, a semantic requirement, or possible both (cf rust editions, or perl use v<something>).

features may be global, namespaced to their implementations, or live in a reserved experimental namespace an implementation can add to and remove from as it wishes with absolutely no guarantee of future evaluatility.

syntactic features

syntax is entirely local to the file itself and has few to no intercompatibility constraints with other code. a very useful syntax requirement would something like no-url-literals, which might strip the syntactic ability to parse url-like sequences of characters into strings and, rather than nix currently does the experimental feature of the same name simply throwing a parse error, parse them as eg a lambda with a sequence of divisions in its body.

(realistically no-url-literals would not appear in practice, instead it should be implied by use itself since url literals are such an obvious misfeature)

semantic features

semantic features produce evaluation changes that could be achieved any other way. examples of this are:

semantic changes may escape the expression that requires them and usually some of amount of cross-compatibility with other semantic versions must be given. using the same examples as above, considerations can include:

this is in fact a full classification of cross-compatibility issues: side-effects changing, evaluation outputs changing, and evaluation inputs changing. side-effects need not be considered very much since nixlang is supposed to be pure and all side-effects that are not part of the store interface must already be considered incidental. evaluation outputs changing can be handled by optional lint or runtime warnings when a versioned evaluation structure passes a semantic version boundary without being annotated as an intentional behavioral leak. evaluation inputs changing is a non-issue because nix plugins and the ExternalValue infrastructure already make it impossible to rely on the type system being fully specified at the time an expression is written

inter-file inter-actions

by default language features must not be propagated across an unadorned import boundary to retain compatibility with existing nix code (eg nixpkgs, which will not be able to switch for quite some time). in some circumanstances it is however required to propagate language features across imports to provide a consistent and meaningful interface, eg in the case of a hypothetical requiredLanguageFeatures attribute for a flake. to allow for both of these requirements to peacefully coexist we add a new primop:

scopedImportUsing
:: { features ? <current language features> :: AttrSetOf bool
     ## ^ language features as would be specified by `use ...;`.
     ## selecting a default-off feature is achieved by setting its key to `true`,
     ## deselecting a default-on feature is achieve by setting its key to `false`.
     ## nesting is not needed because features are identifiers. future changes to
     ## the use interface may extend the type of this set.
   , newGlobals ? env: env :: AttrSetOf Any -> AttrSetOf Any
     ## ^ function to produce the new global environment. it receives the default globals
     ## set for the target expression language features (as calculated form `features` and
     ## the target `use` clause) and produces a new set.
     ## `scopedImport` behavior is recovered by setting this to `const newEnv`.
   }
-> PathLike
## ^ imported path as in `scopedImport`
-> Any
## ^ import result. may be cached, most immediately using the intransparent internal
## object id of the provided features and the globals set. this mimics the beavior
## or `import` in cppnix

if the imported expression selects a different set of language features the features specified by scopedImportUsing are ignored.

scopedImportUsing is available in the builtins set and crucially, can be replaced. this allows a hypothetical flake implementation to replace both scopedImportUsing and import with its own versions that provide propagation behaviors that might be expected from such a library:

additionally the current language features might be made available through a builtin value languageFeatures by such a replacement of scopedImportUsing.

builtins versioning, global versions

a language feature may add or remove elements of builtins or the global environment. as mentioned earlier this does not pose a large hazard since evaluation is sufficiently unespecified that this must already be expected to happen.

interactions with eg nixpkgs lib

nixpkgs lib (and other libraries) will have to cater to the smallest common denominator when exposing library functions/constants as they do now. if we change a function to have a different prototype and a library reexports it from builtins to its own namespace the language features used by the code importing the library do not matter. to make this problem less unbearable we may want to introduce a concept of library objects and a "use library" directive like eg python from ... import ... that can pass language features down to the library being imported in some way.

as a first approximation is would be sufficient to encourage libraries to version their namespaces in such a way that accessing a namespace that relies on language features not present in the current evaluator will fail to evaluate (eg by providing the library itself as a plain set and each version as an attribute that (lazily) imports the specific version of the library needed to fulfill the requested version).

bad ideas for features to remove/change in the first langver

Feature detection

jade: I think we might want to be able to feature detect certain features, e.g. new builtin args, which can be done without, but we would like to know if they are there.

builtins.nixVersion has been defanged, which means that an alternate cross impl compatible mechanism needs to be created.

Minimally thought-through proposal

builtins.features is an attribute set, where individual attribute names are exposed with the value true if they are implemented by a given implementation.

Attribute names are of the format:

"domainname.feature", for example, "systems.lix.somefeature".

Design planning

Docs rewrite plans

Here, for now (public edit link): https://pad.lix.systems/lix-docs-planning

Design planning

Nix lang v2

The Nix language unfortunately is full of little and big design accidents. Only so much can be fixed without breaking backwards compatibility.

Our goal is to design an improved Nix language revision, working title "Nix 2". To keep the scope manageable, the first iteration of language improvements will be restricted to be mostly backwards compatible and only require minimal migration effort. This allows us to test the process on a smaller scale, as well as allows us to get the quick and easy improvements out as soon as possible for others to use.

Join the discussion on Matrix: #nix-lang2:lix.systems

The rough action plan is:

  1. Fork the grammar and gate its usage behind a feature flag.
  2. Use the new grammar as a playground to experiment and implement fixes and improvements to the language, free of any constraints of backwards compatibility.
  3. Figure out language versioning and prepare interoperability.
  4. Provide a migration path, stabilize the new language, and make it available to users.

Initial language changes

Fixing floats

Grammar: All floats must have a digit before the .. This is a hard requirement for making some of the other proposed syntax changes parse unambiguously in the first place.

Moreover, floating point semantics are currently broken in several ways. They need to strictly follow IEE754 double semantics instead.

Given that such a switch is not easy to make in a safe way, as an intermediate solution all floating point operations should be forbidden, effectively making floating point values opaque to the language.

Set slicing

Partially adapted from https://github.com/NixOS/rfcs/pull/110.

Sets can be sliced using set.[key1, key2] and set.{key1, key2}. The first returns a projection of the listed keys into a list, the second one a subset. All keys must be identifiers (or string identifiers), scoped to the attribute set.

[TBD: it is unclear as to whether interpolation is useful and how easy it is to implement] Identifiers may be interpolated: set.[key1, ${key2}] is equivalent to [ set.key1, set.${key2} ], set.{key1, ${key2}} is equivalent to { key1 = set.key1; ${key2} = set.${key2}; }.

Slicing into lists is a replacement for using with:

dependencies = python.pkgs.[
    arabic-reshaper
    babel
    beautifulsoup4
    bleach
    celery
    chardet
    cryptography
];

List and Set unpacking

In a list, elements which are lists themselves can be unpacked with the * operator. They will be concatenated in-place. ["hello", *list, "world"] is equivalent to ["hello"] ++ list ++ ["world"]

This can be easily combined with set slicing. The operator precedence facilitates patterns like the following:

  configureFlags = [
    "--without-ensurepip"
    "--with-system-expat"
    *(optionals (!(stdenv.isDarwin && pythonAtLeast "3.12")) [
      #  ./Modules/_decimal/_decimal.c:4673:6: error: "No valid combination of CONFIG_64, CONFIG_32 and _PyHASH_BITS",
      # https://hydra.nixos.org/build/248410479/nixlog/2/tail
      "--with-system-libmpdec",
    ])
    *(optionals (openssl != null) [
      "--with-openssl=${openssl.dev}",
    ])
  ];

In a set, one can unpack elements like this:

let baz = { bar = "foo"; }; in { foo = "bar"; *baz.{bar}; }

This combines well with optionalAttrs:

{
  meta = with lib; {
    maintainers = with maintainers; [ matthewbauer qyliss ];
    platforms = platforms.unix;
    license = licenses.bsd2;
  };

  HOST_SH = stdenv'.shell;

  *lib.optionalAttrs stdenv'.hasCC {
    # TODO should CC wrapper set this?
    CPP = "${stdenv'.cc.targetPrefix}cpp";
  };
  
  *attrs;

  *lib.optionalAttrs (attrs.headersOnly or false) {
    installPhase = "includesPhase";
    dontBuild = true;
  };

  # Files that use NetBSD-specific macros need to have nbtool_config.h
  # included ahead of them on non-NetBSD platforms.
  postPatch = lib.optionalString (!stdenv'.hostPlatform.isNetBSD) ''
    set +e
    grep -Zlr "^__RCSID
    ^__BEGIN_DECLS" $COMPONENT_PATH | xargs -0r grep -FLZ nbtool_config.h |
        xargs -0tr sed -i '0,/^#/s//#include <nbtool_config.h>\n\0/'
    set -e
  '' + attrs.postPatch or "";

}

It also allows to have "local" let bindings for just some of the keys, without having to move them out of the entire attrset:

{
  key1 = "value1";
  *let
    stuff = "foo";
  in
  {
    inherit stuff;
    key2 = stuff;
  };
}

As with convential set declaration, duplicate keys are not allowed.

Note that the pattern of inherit (foo) bar baz; is equivalent to *foo.{bar, baz};.

Pipe operator function application: |>

This is being worked on in RFC 148

In nixpkgs, there is the lib.pipe function which will allows to write g f a as pipe a [f g]. Especially with deep nested and complicated data transformations, it makes the code flow from left to right and thus easier to read. Sadly, it is under-used because many people are not aware of it.

The fundamental problem it tries to solve though is that function calls are prefix, i.e. that a data processing chain with multiple entries is read from right to left. (Or, when adding parentheses, from the inside to the out side.)

Therefore, we introduce the |> operator. a |> f |> g is equivalent to g(f(a)).

List indexing

Introduce list.INDEX on lists as syntax sugar for builtins.elemAt list index. list.${index} interpolation for dynamic variables also works like it does for attribute sets. To avoid type ambiguities at runtime, ident.${expr} is reserved for dynamic attribute access only, dynamic list indexing still requires using builtins.elemAt

Optional: We could even introduce .last .tail and .length as attributes. Need to think about that. Is a bad idea because of dynamic typing.

Function list destructuring

The same way as function arguments can be destructured into an attrset with {…}, it should also work with lists. Some restrictions:

This, together with list indexing syntax, will make tuple-style code constructs a first-class citizen of the language. Replacing nameValuePair alone is expected to give significant performance gains (short lists are heavily optimized in the evaluator).

Disallow inner-attribute merging

Nix has syntax sugar for merging attrsets within attrset declarations: { a = {}; a.b = "hello"; } will be fused into { a = { b = "hello"; }; } at parse time.

This feature, only rarely used, does not compose well with other features like rec attrsets, leading to unintuitive semantics and potential foot guns: https://git.lix.systems/lix-project/lix/issues/350, https://github.com/NixOS/nix/issues/6251, https://github.com/NixOS/nix/issues/9020, https://github.com/NixOS/nix/issues/11268, https://md.darmstadt.ccc.de/xtNP7JuIQ5iNW1FjuhUccw#inherit-from-scopes-differently-than-inherit

Since these problems would be deeply aggravated by the new set unpacking syntax (defined below), it is best to completely remove this feature altogether. Since it only is convenience syntax sugar, no replacement syntax is necessary.

Expand inherit syntax

The inherit syntax is adapted to be both more powerful and more consisten with the slicing syntax. The inherit (from) is made redundant and deprecated for removal in a future language revision. Inherit can also be used outside of attrsets and let bindings now, and will behave as if it was in a let binding.

inherit lib.{mkIf, types};
inherit {
  lib.mkif,
  types.{attrsOf, listOf, string}
};
# Mixing old with new style syntax: Do we want to allow this?
inherit
  lib.mkif
  types.{attrsOf, listOf, string}
;
# This only makes sense within attrsets really
inherit foo;

Proper keywords for null, true and false

I don't know why these are builtins instead of keywords but at this point I assume it's because it was faster to implement.

Proper syntax nodes for all arithmetic expressions

No more __sub and __lessThan. These had no reason but laziness to exist in the first place.

? and or operator

All line endings must be \n

The current handling of \r is so jank that we'd better do without.

CRLF line endings are allowed within the file for Windows compat, but in strings the line endings get consistently normalized to LF only.

All files must be valid UTF-8 text

The world runs on UTF-8, and most tools these days expect UTF-8 encoded input by default. There's no reason to allow other encodings or invalid byte sequences.

Sane escape sequences for strings

Indented strings

Don't strip indentation of first line

The current behavior is just weird, both for single-line strings (commonly used for unquoted ") and multi-line strings. The new behavior is also what Haskell does (in its new multiline strings proposal).

Indented strings work with tabs

Programming languages may be opinionated, but making some features work only with space indentation is crossing a line.

Tabs and spaces can be mixed as part of the string's content, but not for the string's indentation. Indentation is calculated based on the longest common prefix.

Old cruft to remove

https://wiki.lix.systems/link/21#bkmrk-bad-ideas-for-featur

Remove unquoted URLs

DSL or not, you'll survive typing those two additional extra characters.

Remove let {} syntax

And also the special body attribute.

__override special attribute

No more magic attributes please. __functor is already bad enough.

Fix tokenization rules

https://md.darmstadt.ccc.de/xtNP7JuIQ5iNW1FjuhUccw?view=#token-boundaries-aren%E2%80%99t-real-and-will-hurt-you-cf-nix-iceberg

Status: Partially implemented in https://gerrit.lix.systems/c/lix/+/1984

Autocaller must die

Status: Not implemented

wtf?

❯  nix-build --expr '[[[ ({a}: [a]) ]]]' --arg a 'with import <nixpkgs> {}; hello' --no-link
fetching path input 'path:/nix/store/nyysli8lhjf03jgyvrf7mlxrlgnqn9qp-source'
/nix/store/kwmqk7ygvhypxadsdaai27gl6qfxv7za-hello-2.12.1

Future language changes

Some changes to the syntax would make large chunks of existing code invalid. These need to be postponed until proper versioning and migration tooling have been figured out.

Comma separated lists (confidence: high)

Currently list items are space separated. This has two major drawbacks:

  1. This is inconsistent with most other language syntax features, which use , or ; as item separator.
  2. Not having them requires using parentheses around function calls in lists. Those are currently easy to forget, causing confusing type issues for beginners. (This would be less of an issue if we had a type system that could catch the mistake early on …)

Function declaration (confidence: low)

NUL bytes must be supported within strings

Status: Blocked on rewriting the garbage collector to be compatible

0-terminated strings were a mistake, and we should not make any concessions in the language to implementations who use them. Especially when they're buggy.

Paths

Comments

While we are touching the syntax, let's leave some space here to discuss code comments.

TODO

Design planning

Flake stabilisation proposal

Preface

FIXME: this page hasn't been reviewed by Lix Core team members, so it's effectively a draft/suggestion/pre-RFC/dream, whatever. It's not an official design document, but thought has been put into making it good, anyway.

Problem Statement

Flakes are a mess. They are extremely popular (so it's very painful to discard them), but they are also deeply flawed in so many ways, and their compat story is non-existent. Let's go through a few things that are traditionally associated with flakes, but they don't need to be.

On the last 2 points, see this: https://samuel.dionne-riel.com/blog/2024/05/07/its-not-flakes-vs-channels.html

Overall, flakes did too much at once. We can sort those out one by one. Deprecating NIX_PATH and channels would be a bit tricky, but we can try to re-use flake registries for the same functionality.

Also, flakes have a very bad backwards-compatibility story. Worse than that, we are a CppNix fork, so we want to provide a migration path for a reasonable amount of time. CppNix also completely doesn't have forward compatibility. This means that doing any changes to the flake.nix or to flake.lock will break flakes for CppNix users. This is really bad, it essentially means we're removing flakes outright, so this isn't something that we want to do.

With those preparations out of the way, we can now get to the flakes.

Flake Components

Flakes themselves have many moving parts.

The most cursed part is how tightly connected all of that is. flake.lock records the inputs to builtins.fetchTree. These inputs are parsed from flake.nix. The real abstraction here is inputs URL parser. Everything else is implementation details that leak out into public interfaces.

So the situation is tricky. Code changes leak out, there are no useful versioning mechanisms, we need to make changes in such a way as to not break upstream, and the adoption is large enough that we don't want to break things. But thankfully, there is a way to deal with it, largely inspired but Opentofu's approach.

The Plan

Stage 0: Fork the Interfaces

First, we must fork the interfaces. Instead of having ossified flake.nix and flake.lock interfaces that we have no control over - we fork them into different files. Naming is TBD, but let's use flake.lix and flake.lick in this discussion. More specifically, the procedure looks like this:

This completely changes the compatibility story, because we no longer have to think about upstream usage: we only read, never modify the files the upstream uses. Together with adding sane versioning, we can isolate the versioning to just our project, and make changes (including backwards-incompatible ones) in a sane manner.

Stage 1: Eating Spaghetti

Next, we need to decouple implementation, flake.lix and flake.lick from each other. For the latter two, we already have separate version on "manifest" file and the lockfile; it's a good start. Let's discuss what needs to be done to unveil this spaghetti:

Stage 2: Improving the Interfaces

There's a lot that can be done here. Cross-compilation, version resolution, newer fields, and more - all of that belongs to this stage.

Stage 3: Maintenance

This path is backwards-compatible throughout, so we can maintain an upgrade path without much issue. We can have a directory with subdirectories for each major version. Those subdirectories will also handle upgrading the lockfile; then, we'll always have a path to upgrade from CppNix flakes to Lix flakes: you just execute all of the existing upgrades in order.

Truly backwards-incompatible changes would be adding absolutely necessary metadata, without which the previous version is useless. npm has this: their oldest lockfile (you can call it "v0") didn't have a version field, and it also didn't record checksums. It simply doesn't contain any metadata that better lockfiles do, so the only way to move forward is to extract whatever you can from it, and generate a newer-version lockfile from scratch with that data.

As long as we only need the version and checksum (which seems to be the case), the only source of breaking changes I see is security vulnerabilities. If e.g. NAR hashing is proven to be vulnerable - it's probably for the best to not rely on the already existing hashes at all.

Notes

This plan doesn't have to be executed as sequentially as it's described. Really, we can have something like a from-scratch rewrite for flakes and include it in the first flake.lix+flake.lick versions. Or we could only add the versioning code. Or we could add versioning and version resolution, or versioning and cross-compilation, or literally anything else, as long as versioning is definitely present.

Appendix A: Flakes are a broken abstraction

Some parts of this were already mentioned, but flakes are pretty broken on fundamental levels. The lockfile essentially containing arguments for a C++ function are an example of that. This isn't an abstraction pretty much by definition - it does not abstract away the details. A good example of a lockfile is version = 3 for Cargo:

[[package]]
name = "anstyle-wincon"
version = "3.0.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "61a38449feb7068f52bb06c12759005cf459ee52bb4adc1d5a7c4322d716fb19"
dependencies = [
 "anstyle",
 "windows-sys 0.52.0",
]

[[package]]
name = "anstyle"
version = "1.0.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "038dfcf04a5feb68e9c60b21c9625a54c2c0616e79b72b0fd87075a056ae1d1b"


[[package]]
name = "windows-sys"
version = "0.52.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "282be5f36a8ce781fad8c8ae18fa3f9beff57ec1b52cb3de0789201425d9a33d"
dependencies = [
 "windows-targets 0.52.5",
]

...

We more or less have the source (as bad as it is implemented in flakes), and we have checksum (which is NAR hash) and name - but we are missing the version abstraction. There's some complexity to unpack here (for example, it isn't trivial to say what "version" means for a tarball or a filesystem path), but flakes don't even try - they just completely ignore the need for this abstraction, and use C++ implementation details instead.

Another issue is the general confusion about what flakes are supposed to be, and how Nix plays into that. Nix is a lot of things, but the way it ended up working out is that Nix is a builder abstraction: you use Nix to build packages, and the packages may have dependencies, and yadda yadda. But because Nix is so general, it can be used to build a "meta-package" of all "installed packages", and you can also use it to build OS configs, so you can essentially build a system meta-package. The whole NixOS system is just a big meta-package that consists of other packages.

This is a blessing and a curse: expressing the entire system as one package is cool and has its advantages, but this is also a very hacky way to use the build system that is Nix. It's like using the Makefile to configure your system. CppNix developed a lot of stuff to keep this approach going: channels, NIX_PATH, nixos-rebuild scripts, nix-env and other things are all used to make the experience more tolerable. So it's a lot of hacks on top of a rather quirky way to use the build system. The biggest example on how it manifests is NixOS configuration: we use it to create different build manifests for the resulting system, and we don't have other ways to interact with the system, like a package manager. This is a tough place to be in: the NixOS approach has a lot of really good properties, but it's also inherently limited because the build system is used as a configuration engine and a package manager.

Flakes are confused and stupid because they try to be a package manager for Nix, but they are a shitty package manager, and they also don't even try to resolve many of the hard questions that arise from using Nix itself as a package manager. They don't have a concept of "libraries", so everyone still uses Nixpkgs lib. They don't have version resolution, or a concept of versions. They don't really integrate with Nix profiles, they don't integrate with NixOS, they don't draw good boundaries between what different units of NixLang code do: provide library functions, create packages, create configuration, or whatever else.

There are only three package manager things that flakes actually tried to do: it's installables/runnables abstraction (just barely counts), manifest+lockfile usage (the idea itself is good but impl is awful), and defining a schema. Everything else doesn't address the issue at hand in the slightest: some of the ideas are good and should be decoupled from flakes, and some of the ideas are awful.

Regarding installables/runnables: it's a step in the right direction for drawing boundaries between packages, libraries and configs. But the way it's implemented is also bad. The definition for installables is a huge nothing-burger: basically, an "installable" is a store path or a thing that resolves into a store path (this is more or less what's said in the glossary). This definition gives you exactly nothing, and reminds of a horribly ill-defined "derivation" stuff ([1], [2], [3], [4], [5], [6]). The actually useful thing here is "runnables", which are things you can nix run. It's also barely defined (mostly just using the store path and appending /bin/<name> to it lol) and absolutely isolated from all of the Nixpkgs and NixOS work, so it ends up being completely useless in practice.

This document mainly goes over how to unbreak the flakes and make them work at all, but creating coherent abstractions on top of the unbroken flakes is a whole other dimension of pain and integration work. In practice, integrating flakes into Nix properly will end up requiring "owning the stack" or close to it: being very free to refactor and unbreak many hacks in Nixpkgs and NixOS.

Appendix B: Some thoughts on post-stabilisation world

Something that would make a lot of sense is drawing further boundaries between units of NixLang code. Flakes would be a NixLang package manager, and like there is a distinction between "binaries" and "libraries" in proper programming languages like Rust, there would be "flake types" for NixLang. Some easy examples include: "NixLang library", "Nix plugin", "configuration", "package manifest", "binary/runnable", "generic package". Just using those "flake types" for manifests doesn't do much good: there needs to be tight integration with Nixpkgs. In fact, Nixpkgs might start composing flakes instead of just NixLang code directories: this might be a great change for the better.

To give some examples on how integrating flakes would look like, we can take inspiration from dreams page. Let's discuss flake-related items:

So basically, flakes subsystem needs to be an actual package manager with actual units (flakes). Then, flakes will actually make sense and be good, and we'll finally be able to have nice things, like not having Nixpkgs be a gigantic fs tree with dubious abstractions. I mean, pointing to Rust again (because it's good): Cargo doesn't just operate on fs trees and let you handle the rest like an old-school thing like Nix forces you to do, Cargo has many abstractions to decouple fs tree from things you care about: workspaces, crates, modules, etc. When flakes become Cargo and give us proper composition - we'll know we've done a good job.

Design planning

Observability and Protocol Design

jade: I think that we should start protocol design by thinking about who needs what information, which is most cleanly hit by looking at how observability architecture looks. Let's get cracking on what observability we need/want in Lix.

Context

Old profiling pad for the Nix language: https://pad.lix.systems/lix-profiling. This might want/need to be a different system than the overall observability architecture since it affects the evaluator primarily and has specialized needs (e.g. high performance).

Old protocol investigation pad: https://pad.lix.systems/lix-protocol-investigation

Discussion

Let's have this discussion in a pad here so we can have good live editing: https://pad.lix.systems/lix-observability

Design planning

Replacement CLI design & Profiles

Draft pads:

Design planning

Nix bootstrapping

Pad:

https://pad.lix.systems/VjA-WMSQS42dh-ghL98Uow

Design planning

Improving IFD

pennae's idea:

magic_ifd_primop = \f i -> import f >>= i
# nix:
importAndThen (res: <stuff>) some-ifd.outPath

Observation: a magic IFD primop that immediately does something to the result of an import can be "thunked" (thus processed in the background) and offer an "interesting" fix to IFDs. We cannot just make import be lazy because of the evaluator internals.

Subsequent observations:

Why is this interesting?

The capability of forking / parallelizing the work.

Questions:

Release names

Release names are the names of frozen desserts. There's a list on Wikipedia of frozen desserts, but of course, others can be added. The purpose of release names is that they are cute, and they are not necessarily picked for any reason.

Used release names:

Here are some ideas of release names:

Easy Drive-by Contributions

This page lists easy contributions to do without additional context. Perfect to do on a quiet weekend by yourself!

The page is free to edit, so if you've solved one of the issues, or concluded that it's not really easy to drive-by at all - feel free to remove or comment out the issue!

Docs

Python

Rust

NixLang

Go

Java

C++

Discussion notes

If you have a discussion or a meeting and are writing things down and want the meeting notes to persist, here is the correct place. Try to always move all relevant information off any scratch pads into appropriate more permanent places, like the wiki.

The template is optional and meant to help not forget any of the important aspects

Discussion notes

Discussion notes template

Meeting name

Discussion notes

2025-06-27 Wiki work

Wiki work

Summary: We have a mess of information in the pad system which is not well organized by quality, and it's hard to deal with it.

Action items

Meeting notes