LTO is broken on darwin, which can lead to problems for packages that enable it by default. There’s a PR that aims to fix it, but it’s almost 3 years old. It would be really great if someone who’s familiar with clang in Nix could shepherd this to completion.
So if you have a PR which is stuck, or have seen a PR which is stuck,
why not try and post it here. And if you think you can help with a PR
that has been posted then you’re awesome!
First, thanks to everybody posting in this thread, and especially to
Aaron for raising the issue.
I’d like to make a suggestion, though:
When you post a PR in this thread, please include a summary of what
sort of help is needed. Without this, it’s very difficult to look
through the list and see things I could help with – I’d need to open
every PR individually.
Also please put the PR links on their own lines so Discord will expand them into summaries, which will help when skimming the thread to find PRs that I have some experience with.
Hi all! I have been working[1] on deprecating the type loaOf
for quite some time and at this point the PR only needs reviewing. Any help would be appreciated, in particular testing the software not covered by nixos tests.
[1]: Deprecate types.loaOf by rnhmjoj · Pull Request #63103 · NixOS/nixpkgs · GitHub
https://github.com/NixOS/nixpkgs/pull/73267
The bot reports build errors on the darwin
platform. I have no access to darwin
or time/ability to figure this out even if I did. If someone was motivated to take a look at this it would be much appreciated as there are 2 CVEs listed against mariadb
which are fixed by this release.
https://github.com/NixOS/nixpkgs/pull/73042
Would appreciate input on two things.
(1) The pull request adds “MPRIS” functionality which seems to require X11, which makes spotifyd fail when X11 is not available. Unsure how to handle this but I would like to see spotifyd working by default without X11.
(2) The version bump broke the spotifyd service module for me and I had to manually clear out its cache. I do not know if we should handle this and if so, how.
https://github.com/NixOS/nixpkgs/pull/72987
This PR went unnoticed but it’s a relatively important fix: it should also be backported. I have been running it on my server so I think it’s already good to be merged.
Solved
https://github.com/NixOS/nixpkgs/pull/73905
Would be thankful for a quick function test on Darwin/MacOSX so I can tick the last box in my first PR.
If anyone has a Darwin/MacOSX machine and 5 minutes for this, please do as follows:
- Checkout the PR
- Run at the top of the worktree
nix-shell -E 'with import ./default.nix {}; mkShell { buildInputs = [ ammonite fish ]; }'
- If there’s no error, run
fish
in the resulting shell - Run
amm
in the resulting fish shell - Check that the Ammonite REPL pops up
- Type
exit
or hit control-D, whichever you prefer - Report what happened in the PR thread
https://github.com/NixOS/nixpkgs/pull/65835
Unfortunately, this package is unmaintained, so I don’t know whom to ping to look at this PR.
I have one question there because of the impact on package closure size of +78.35 MB (46.52%): Should gstreamerSupport
be false by default (thus the effect of this PR be a non-default option)?
https://github.com/NixOS/nixpkgs/pull/83321
This is relatively simple PR but it touches the core of the python build system and is therefore a mass rebuild. I need some assistance in testing and reviewing.
I have a new PR in distress:
https://github.com/NixOS/nixpkgs/pull/85319
The issue is I don’t have a darwin
machine so I can’t fix the darwin
build without putting in unreasonable amounts of effort. Maybe some darwin
wizard will help me out.
https://github.com/NixOS/nixpkgs/pull/50891
Is stuck, waiting for final review with two other reviewers already resigned.
Here are two similar patches waiting for boost17x that improve the closure size usage of packages which depend on boost17x (such as nheko
). If either one will be merged it’d be great:
My PR doesn’t build on aarch64. I got “error: linker aarch64-linux-gnu-gcc
not found” and I have no idea if it’s a problem with one of the Rust dependencies or if it’s a problem with the aarch64 builder.
This PR exists to upgrade fontconfig (the current version is now nearly 3 years out of date), and introduces breaking changes to fontconfig’s configuration. Because of this, it’s causing issues for lots of people who use nix on non-NixOS, because both system’s fontconfig and nix’s fontconfig will try to read configuration from /etc. As far as I can tell the PR is fine, and the only blocker seems to be the huge number of reverse-dependencies (i.e. hydra rebuilds). There’s also something about backporting old versions of fontconfig to generate font caches for older versions, but I’m not really sure what’s going on there.
SSH - Add crypot options.
Example ssh config after this PR:
programs.ssh = {
...
macs = [ "hmac-sha2-512-etm@openssh.com" "hmac-sha2-512" ];
ciphers = [ "chacha20-poly1305@openssh.com" ];
kexAlgorithms = [ "curve25519-sha256@libssh.org" ];
...
};
services.openssh = {
...
macs = [ "hmac-sha2-512-etm@openssh.com" ];
ciphers = [ "chacha20-poly1305@openssh.com" ];
kexAlgorithms = [ "curve25519-sha256@libssh.org" ];
...
};
I have a not too big nodePackages
fix for an issue which passes all tests possible, but every once in a while it reaches a merge conflict status… And what’s annoying even more is that running node’s generate.sh
takes about half a day (for me) and afterwards I have to cherry-pick only the relevant changes to the commit and discard the rest …
Improvements to gitlab-runner
module:
https://github.com/NixOS/nixpkgs/pull/86946
Fix for armv7l/aarch64 image builder:
https://github.com/NixOS/nixpkgs/pull/82718