!djTaTBQyWEPRQxrPTb:nixos.org

Nixpkgs Architecture Team

231 Members
https://github.com/nixpkgs-architecture, weekly public meetings on Wednesday 15:00-16:00 UTC at https://meet.jit.si/nixpkgs-architecture53 Servers

Load older messages


SenderMessageTime
6 Feb 2023
@figsoda:matrix.orgfigsoda
In reply to @gytis-ivaskevicius:matrix.org
Hey guys, not sure what you guys are up to these days but I'd appreciate if you guys could follow up on this blueprints idea, maybe making it a standard in nix ecosystem
https://github.com/NixOS/nixpkgs/pull/208020#issuecomment-1419168718

The blueprints still looks very nixpkgs-esque to me

I believe that we should not encourage people to depend on nixpkgs and this should be resolved in nix itself.

I can see a callPackage like pattern being lifted to the nix level (maybe in a separate repository core, kind of like nixpkgs.lib), but things like stdenv are strictly nixpkgs, it breaking change in nixpkgs will cause this flake to break, and I'm not seeing anything that refers to nixpkgs in the flake outputs.

about the issue, I like the perSystem pattern that flake-parts uses, maybe this can be supported in a more official way (something in nixpkgs like nixosSystem, or even ootb flakes support)

14:56:25
@figsoda:matrix.orgfigsoda
In reply to @gytis-ivaskevicius:matrix.org
Hey guys, not sure what you guys are up to these days but I'd appreciate if you guys could follow up on this blueprints idea, maybe making it a standard in nix ecosystem
https://github.com/NixOS/nixpkgs/pull/208020#issuecomment-1419168718
*

The blueprints still looks very nixpkgs-esque to me

I believe that we should not encourage people to depend on nixpkgs and this should be resolved in nix itself.

I can see a callPackage like pattern being lifted to the nix level (maybe in a separate repository core, kind of like nixpkgs.lib), but things like stdenv are strictly nixpkgs, it having a breaking change in nixpkgs will cause this flake to break, and I'm not seeing anything that refers to nixpkgs in the flake outputs.

about the issue, I like the perSystem pattern that flake-parts uses, maybe this can be supported in a more official way (something in nixpkgs like nixosSystem, or even ootb flakes support)

14:59:45
@figsoda:matrix.orgfigsoda *

The blueprints still looks very nixpkgs-esque to me

I believe that we should not encourage people to depend on nixpkgs and this should be resolved in nix itself.

I can see a callPackage like pattern being lifted to the nix level (maybe in a separate repository core, kind of like nixpkgs.lib), but things like stdenv are strictly nixpkgs, it having a breaking change in nixpkgs will cause this flake to break, and I'm not seeing anything that refers to nixpkgs in the flake outputs.

about the issue, I like the perSystem pattern that flake-parts uses, maybe this can be supported in a more official way (something in nixpkgs like nixosSystem, or even ootb support in nix)

15:02:43
@infinisil:matrix.orginfinisil @room: The next meeting will take place in ~10 minutes, we'll be there to discuss RFC 140 - meeting link - live stream - meeting notes 15:21:39
@gytis-ivaskevicius:matrix.orgGytis Ivaskevicius

I can see a callPackage like pattern being lifted to the nix level (maybe in a separate repository core, kind of like nixpkgs.lib), but things like stdenv are strictly nixpkgs, it having a breaking change in nixpkgs will cause this flake to break, and I'm not seeing anything that refers to nixpkgs in the flake outputs.

No, I think this should stay part of nixpkgs. Just that users have a fool proof option to make package depend on whichever nixpkgs revision they feel like

about the issue, I like the perSystem pattern that flake-parts uses, maybe this can be supported in a more official way (something in nixpkgs like nixosSystem, or even ootb support in nix)

perSystem is not part of the idea, it's being addressed as a separate issue. Quite possibly differently

15:23:02
@nbp:mozilla.orgnbp
In reply to @figsoda:matrix.org

The blueprints still looks very nixpkgs-esque to me

I believe that we should not encourage people to depend on nixpkgs and this should be resolved in nix itself.

I can see a callPackage like pattern being lifted to the nix level (maybe in a separate repository core, kind of like nixpkgs.lib), but things like stdenv are strictly nixpkgs, it having a breaking change in nixpkgs will cause this flake to break, and I'm not seeing anything that refers to nixpkgs in the flake outputs.

about the issue, I like the perSystem pattern that flake-parts uses, maybe this can be supported in a more official way (something in nixpkgs like nixosSystem, or even ootb support in nix)

Let's hope not, as my plan was to get rid of its crappy logic to make it trivially extensible.
15:25:10
@infinisil:matrix.orginfinisil
In reply to @gytis-ivaskevicius:matrix.org
Hey guys, not sure what you guys are up to these days but I'd appreciate if you guys could follow up on this blueprints idea, maybe making it a standard in nix ecosystem
https://github.com/NixOS/nixpkgs/pull/208020#issuecomment-1419168718
We're currently focusing on RFC 140 as the first major change, but after that's done and implemented we'll have to find the next thing to work on, which might be something from the future work section, from the other NAT issues or something else
15:26:09
@infinisil:matrix.orginfinisilThere was also a discussion yesterday about parallelising the work more15:26:34
@infinisil:matrix.orginfinisilI think it all hinges on making that first RFC a success though15:27:56
@growpotkin1:matrix.orggrowpotkin ( Alex Ameen ) A fool proof way to depend on a particular version is a pinned form of Nixpkgs in a flake. You can pin an arbitrary number and create an overlay to cherry pick. 15:30:15
@growpotkin1:matrix.orggrowpotkin ( Alex Ameen )Override is your friend also 15:31:01
@infinisil:matrix.orginfinisilEhh, overlays and overrides have serious composability problems, I really think we need something better to replace them15:31:22
@infinisil:matrix.orginfinisil John Ericson Robert Hensing (roberth) tomberek: Joining? 15:32:26
@blaggacao:matrix.org@blaggacao:matrix.org Gytis Ivaskevicius: from a 30s look the blueprints idea is almost 100% overlap with RFC140. Correct me if I'm wrong. 16:19:12
@blaggacao:matrix.org@blaggacao:matrix.orgpkg-evolution.png
Download pkg-evolution.png
16:21:03
@blaggacao:matrix.org@blaggacao:matrix.org It would be pkg-fun, I guess. 16:21:16
@gytis-ivaskevicius:matrix.orgGytis Ivaskevicius
In reply to @growpotkin1:matrix.org
A fool proof way to depend on a particular version is a pinned form of Nixpkgs in a flake. You can pin an arbitrary number and create an overlay to cherry pick.
My option gives you a possibility of doing that
16:24:41
@blaggacao:matrix.org@blaggacao:matrix.orgBeyond features, though, the name of the game is the process to obtain convergence and wide-spread adoption.16:27:23
@blaggacao:matrix.org@blaggacao:matrix.org * Beyond features, though, the name of the game is the process to obtain convergence and wide-spread adoption. 16:27:40
@blaggacao:matrix.org@blaggacao:matrix.org
In reply to @infinisil:matrix.org
I think it all hinges on making that first RFC a success though
Or ^^ this, in other words.
16:29:11
@tomberek:matrix.orgtomberek
In reply to @gytis-ivaskevicius:matrix.org
Hey guys, not sure what you guys are up to these days but I'd appreciate if you guys could follow up on this blueprints idea, maybe making it a standard in nix ecosystem
https://github.com/NixOS/nixpkgs/pull/208020#issuecomment-1419168718
Yes.... the blueprints concept is something I'd like to examine and shepherd.
18:00:05
@winterqt:nixos.devWinter (she/her)
In reply to @blaggacao:matrix.org
It just got harder and became an evangelization issue after playing a little with it. But that's borderline off topic, here, anyways.
Can I ask what this issue is even saying? I'd consider myself very experienced in Nix and yet I am very, very confused.
18:28:59
@winterqt:nixos.devWinter (she/her) If I'm this confused, I'd say the upstream maintainers are absolutely clueless as to what you're saying -- I don't even see a call to action or anything. 18:31:07
@infinisil:matrix.orginfinisil I think it's David Arnold (blaggacao) trying to make a point for adding a Nix build, maybe it should be linked to https://github.com/frappe/bench/pull/1429 18:31:10
@blaggacao:matrix.org@blaggacao:matrix.org

There are so many different ways to approach Nix. It's always a gamble.

This istance asumes maintainers are familiar with PATH and sys.addsitedirs() and tries yo explain what it does for them in a two-script capsule what they currently have been trying to achieve with multiple ansible playbooks.

18:45:33
@blaggacao:matrix.org@blaggacao:matrix.org* There are _so_ many different ways to approach Nix. It's always a gamble. This istance assumes maintainers are familiar with `PATH` and `sys.addsitedirs()` and tries yo explain what it does for them in a two-script capsule what they currently have been trying to achieve with multiple ansible playbooks.18:45:48
@blaggacao:matrix.org@blaggacao:matrix.org* There are _so_ many different ways to approach Nix. It's always a gamble. This istance assumes maintainers are familiar with `PATH` and `sys.addsitedirs()` and tries to explain what it does for them in a two-script capsule. As soon as they refuse to get distracted by wired path names, they will recognize the packages and libraries that they currently are procuring via ansible playbooks in a very heavy handed approach that makes up 20% of the `bench` codebase.18:47:24
@piegames:matrix.org@piegames:matrix.org

There are so many different ways to approach Nix. It's always a gamble.

If your way of approaching Nix results multiple people already deeply familiar with Nix saying that they don't understand your text (add me to that list, I read the code blocks twice and they still look like some arcane spell to me), then I guarantee you'll lose that gamble for everybody else too.

21:49:01
@piegames:matrix.org@piegames:matrix.org

evangelization issue

As soon as they refuse to get distracted by wired path names, they will recognize the packages and libraries that they currently are procuring via ansible playbooks in a very heavy handed approach that makes up 20% of the bench codebase.

I can't really put it into words, but I find those sentences deeply unsettling. It gives me the vibes of some pretentious 16th century missionary trying to convert those primitive people or something. (I apologize for the metaphor, but it currently best describes how I feel about this.)
I know there are running jokes about Nix being a cult/sect/religion and about "converting" people to Nix, yet I've never felt it being hostile until now

21:49:04
@blaggacao:matrix.org@blaggacao:matrix.orgOnce language and stance becomes a distraction itself, it's a sign for me to go fish. Thank you.22:05:32

Show newer messages


Back to Room ListRoom Version: 9