| 30 Sep 2025 |
Emma [it/its] | yes | 01:03:07 |
Emma [it/its] | probably the biggest offender in my case is ommitting @aws-sdk/client-s3 for the majority of users that dont care about s3 support lol | 02:01:09 |
Emma [it/its] | additional question: is there any reason a buildNpmPackage result has an explicit dependency on ${NODEJS}-sources? | 03:09:14 |
Winter | (i’ll try and remember to respond to all of these tomorrow, going to bed; but rapid fire: just removing the directories but keeping the same lockfile would be good, and no; i just haven’t fixed it yet) | 04:04:34 |
Emma [it/its] | ah i see, thanks for the responses Winter!
good night :) | 04:06:46 |
Cobalt | Doesn't importNpmLock do this though? Iirc in build logs it does them on at a tims | 04:12:19 |
Emma [it/its] | are they separate derivations though? or is it jut the standard npm logging when it fetches them | 04:20:53 |
Cobalt | It appears so, tree from one of the packages, https://git.cobalt.rocks/-/snippets/32. At least the sources appear to be on drv each
(Eval time is a bit higher because of this but it helps with dedup). | 13:31:06 |
Cobalt | * It appears so, tree from one of the packages, https://git.cobalt.rocks/-/snippets/32. At least the sources for npm deps appear to be on drv each
(Eval time is a bit higher because of this but it helps with dedup). | 13:31:12 |
| NixOS Moderation Bot banned @joepie91:pixie.town (divsive behavior). | 19:23:48 |
| 1 Oct 2025 |
dish [Fox/It/She] | I hate nodePackages | 22:00:23 |
dish [Fox/It/She] | been working on clearing out a few more high-impact packages from that set(ones that have a lot of packages they depend on, which can help us really remove a lot of stuff from that file) | 22:01:45 |
dish [Fox/It/She] | * been working on clearing out a few more high-impact packages from that set(ones that have a lot of packages they depend on, which can help us really remove a lot of stuff from that file quickly, which can actually help eval time and also maintainability while the set is still around) | 22:02:05 |
| 2 Oct 2025 |
Winter | me too | 03:51:08 |
Winter | i made a hammer to kill it for this reason… didn’t have the spoons to work on migrating things much though, thank you! | 03:51:27 |
dish [Fox/It/She] | its been a somewhat long-term fixation for me | 03:55:53 |
dish [Fox/It/She] | we've only recently gotten good enough tooling for things like yarn berry and pnpm(yes ik pnpm tooling has been around for a while, but the bugs have been worked out) | 03:56:20 |
dish [Fox/It/She] | frankly I'm finding too many library-only packages in here | 03:58:37 |
dish [Fox/It/She] | the issue with having several of these PRs in flight is that if theres libraries that are shared between only packages in different PRs and nothing else, i cant remove those since reasons | 03:59:30 |
dish [Fox/It/She] | and i could keep track of those but thats way too much work | 03:59:38 |
dish [Fox/It/She] | so eventually the package set will be regenerated and there should hopefully be a giant drop in line count then | 04:00:02 |
dish [Fox/It/She] | maybe we update the package set once per release until there's just a handful of packages | 04:00:44 |
dish [Fox/It/She] | (say, less than 50) | 04:00:48 |
dish [Fox/It/She] | since theres currently still over 150 packages in the set on master branch atm | 04:01:39 |
dish [Fox/It/She] | and there would be way less if projects would vendor a package-lock.json >:( | 04:02:27 |
dish [Fox/It/She] | and i do not want to shove like 50 package-lock files into nixpkgs | 04:02:35 |
dish [Fox/It/She] | the whole reason im doing this is to reduce the amount of bloat in nixpkgs and it would prob be more efficient to keep the set over having a bunch of package-locks lol | 04:03:00 |
dish [Fox/It/She] | I think this is something that nixpkgs needs to solve(some way of creating a trusted source of lockfiles that can be fetched externally so that we don't have to vendor them inside nixpkgs) | 04:17:09 |
Emma [it/its] | wouldnt you have to do that anyways to avoid IFD? | 04:18:50 |
dish [Fox/It/She] | too many projects have gotten the habit from sindresorhus of not having lockfiles in-tree >.> | 04:19:03 |