!OqhvaDMJdKYUicLDiE:nixos.org

Nixpkgs Stdenv

136 Members
49 Servers

Load older messages


SenderMessageTime
31 Jul 2024
@philiptaron:matrix.orgPhilip Taron (UTC-8)
In reply to @raitobezarius:matrix.org
Then I'd move setting the env var in the sandbox inside the package manager impl
Say more on this?
21:59:58
@raitobezarius:matrix.orgraitobezariusI wouldn't hack in the nixpkgs packaging of the package managers22:00:38
@raitobezarius:matrix.orgraitobezariusThe details are vague to me as well but I think a post patch approach seems wrong to me, it has to be minimally a patch, a proper solution to upstream with a proper PR number / issue which describes what are we trying to solve as a usecase and who uses it in practice and how22:01:43
@raitobezarius:matrix.orgraitobezariusLike maybe the hack is to have a FD per loglevel and have the stdenv always write all of them, etc.22:03:43
@raitobezarius:matrix.orgraitobezarius(and let the user choose a dynamic nix3-log level or a global setting)22:04:24
@raitobezarius:matrix.orgraitobezariusOr maybe more things should be structured traces, etc.22:04:35
@philiptaron:matrix.orgPhilip Taron (UTC-8)Yeah, I've seen on basically all Nix-logging related PRs upstream a lament that Nix doesn't provide a way to let derivations do something structured. There's a theme of "what we're doing isn't right, but we don't know what better is." https://github.com/NixOS/nix/pull/4914#issuecomment-1008806246 in particular.22:06:35
@raitobezarius:matrix.orgraitobezariusI will refrain from snarky comments about the state of things being quite obvious to me22:08:46
@raitobezarius:matrix.orgraitobezariushttps://github.com/lf-/nix-otel seems to show an interesting way to pursue22:08:58
@raitobezarius:matrix.orgraitobezariusNow, I'd find it interesting whether stdenv should have more OTEL awareness and whether we should just get into that ecosystem22:09:42
@philiptaron:matrix.orgPhilip Taron (UTC-8)That is a cool direction, for sure.22:15:15
@raitobezarius:matrix.orgraitobezariusHence the question about userbase and audience for this feature22:17:23
@raitobezarius:matrix.orgraitobezariusIf there's no such a thing, we can (a) hide the verbosity asap for the next staging and generalize the pipeline so we can easily prepare for a proper solution (b) propose a good design for the next next staging22:18:16
@raitobezarius:matrix.orgraitobezariusJade is not in this channel but I guess she could better represent what Lix may have in mind for this as I just got home from holidays22:19:08
@philiptaron:matrix.orgPhilip Taron (UTC-8)

Right now, the audience I know about that wants this feature is Qyriad and myself. 😬 As she wrote in the original PR,

even to Nixpkgs veterans it's far from obvious what the final array of hooks for a build might be, or what order they even run in (a source of confusion if one causes an error!).

For me, open telemetry doesn't do the same job as log-staring. I want the paths, the information about whether it was source'd or eval'd, and all the rest of the logging that Qyriad added. And being able to not invalidate the hashes is essential because logging that changes the thing being logged is really irksome.

We definitely can hide the verbosity ASAP. I rather expect having that be split off the PR (or the PR be transformed into just that change).

22:25:46
@philiptaron:matrix.orgPhilip Taron (UTC-8) Since NIX_DEBUG is so well documented in nixpkgs, but it's completely not clear how to go set it as a user, one goal I have is to pave that path a little more. Now that I know and where to patch Nix and Lix so as to set it, I personally am unblocked, but the broader goal of enabling wizards-in-training to get more information from their systems definitely isn't. 22:28:53
1 Aug 2024
@cleverca22:matrix.orgcleverca22 joined the room.12:53:35
5 Aug 2024
@yannis:mozilla.orgyannis changed their display name from yannis to yannis|pto.08:48:05
6 Aug 2024
@philiptaron:matrix.orgPhilip Taron (UTC-8)stdenv folks: I'm "pencils down" on https://github.com/NixOS/nixpkgs/pull/331560 and I would appreciate a review and merge before the staging train leaves the station. It resolves the "hook" messages several have complained about.03:03:12
8 Aug 2024
@qyliss:fairydust.spaceAlyssa Ross set a profile picture.18:48:47
10 Aug 2024
@rosscomputerguy:matrix.orgTristan Ross changed their profile picture.03:13:27
11 Aug 2024
@emilazy:matrix.orgemily joined the room.22:12:53
13 Aug 2024
@yannis:mozilla.orgyannis changed their display name from yannis|pto to yannis.22:55:31
22 Aug 2024
@asmundesen:matrix.orgArtur Manuel joined the room.13:01:45
@asmundesen:matrix.orgArtur Manuel changed their profile picture.14:53:11
27 Aug 2024
@oliver.falvai:oliverfalvai.xyzoliver.falvai joined the room.10:52:20
@aloisw:kde.org@aloisw:kde.org left the room.18:03:14
30 Aug 2024
@artur:glasgow.social(artur 'manuel) joined the room.08:11:28
2 Sep 2024
@emilazy:matrix.orgemily Philip Taron (UTC-8): what would it take to nerd-snipe you into picking up https://github.com/NixOS/nixpkgs/issues/21471 https://github.com/NixOS/nixpkgs/pull/87909 https://github.com/NixOS/nixpkgs/pull/44583? :) 16:40:05
@philiptaron:matrix.orgPhilip Taron (UTC-8) emily my nerd-snipe stack is a little too deep to take on these really intense PRs at the moment. Gotta hunt some easier-to-test-and-less-likely-to-bite-back game for a little bit. 18:53:55

Show newer messages


Back to Room ListRoom Version: 9