| 28 Jan 2024 |
 Paul Meyer (katexochen) | In reply to @qbit:tapenet.org
it does in my experience It definitely shouldn't.
https://github.com/NixOS/nixpkgs/blob/e4b4793ffe04112a121d0a23e2010c576bddefdd/pkgs/build-support/go/module.nix#L235 | 16:09:26 |
 @qbit:tapenet.org | nice | 16:09:55 |
| Paul Meyer (katexochen) | In reply to @qbit:tapenet.org
go test $(go list ./... | grep -v /vendor/) would probably be a good thing to have yes, my solution uses go list (but is a bit more sophisticated). | 16:10:10 |
| @qbit:tapenet.org | so find should be go list there probably | 16:10:10 |
| @qbit:tapenet.org | sweet | 16:10:25 |
| Paul Meyer (katexochen) | but it will cause a lot of built failures | 16:10:46 |
| Paul Meyer (katexochen) | as the current way excludes tests in most cases | 16:11:08 |
| Paul Meyer (katexochen) | you can check the following for a preview ;)
https://github.com/katexochen/nixpkgs/commits/buildGoModule/tests/ | 16:11:49 |
| @qbit:tapenet.org | nice | 16:12:07 |
| Paul Meyer (katexochen) | While I still have some work to do, I think I will just open a draft PR now | 16:12:14 |
 K900 | go: flag needs an argument: -parallel (from $GOFLAGS) | 16:28:13 |
| K900 | Huh | 16:28:15 |
| Paul Meyer (katexochen) | That should already be done here, shouldn't it?
https://github.com/NixOS/nixpkgs/blob/e4b4793ffe04112a121d0a23e2010c576bddefdd/pkgs/build-support/go/module.nix#L209 | 16:34:21 |
| Paul Meyer (katexochen) | Ah, -p is not short for -parallel. But
By default, -parallel is set to the value of GOMAXPROCS.
| 16:36:30 |
| Paul Meyer (katexochen) | GOMAXPROCS might be the thing we actually want to set. | 16:37:19 |
| Paul Meyer (katexochen) | Draft for the go list thing: https://github.com/NixOS/nixpkgs/pull/284568 | 16:39:48 |
| 31 Jan 2024 |
|  Federico Damián Schonborn changed their profile picture. | 03:36:40 |
| Federico Damián Schonborn changed their profile picture. | 06:22:16 |
| 6 Feb 2024 |
| @qbit:tapenet.org | the other bummer about our build for go stuff is that we can't take advantage of a global build cache | 15:26:06 |
| @qbit:tapenet.org | i don't know how it possibly could... just a bummer for speed | 15:27:23 |
| Paul Meyer (katexochen) | Well, I think the way Bazel handles this is by not using the higher level go "build" command but the lower level commands to build each package in separate to get a linkable archive, then invoke the linker on those archives when building the final binary. | 15:47:09 |
| @qbit:tapenet.org | interesting | 16:23:44 |
| Paul Meyer (katexochen) | https://github.com/bazelbuild/rules_go/blob/07317c7ce2ae6e6cda269c0c3ec7eac171363936/go/tools/builders/compilepkg.go#L538-L560 | 16:34:07 |
|  Moritz Sanft set a profile picture. | 17:17:54 |
 Sandro 🐧 | Bazel also thought it would be a good idea to vendor all kinds of libraries and copy their build systems... | 17:40:41 |
| Sandro 🐧 | and fetch archives against GitHub recommendations in the docs and completely breaking when they updated git-archive | 17:41:17 |
| Paul Meyer (katexochen) | Sandro 🐧: not sure how that is related to their mechanism of building Go code. | 18:12:52 |
| Paul Meyer (katexochen) | I think if you really want to break up Go builds into multiple derivations (which is nothing I propose here), the Bazel approach sounds more reasonable and correct to me than somehow injecting build cache. | 18:16:15 |
| 8 Feb 2024 |
|  symys joined the room. | 18:50:55 |
| 10 Feb 2024 |
|  anthr76 joined the room. | 14:32:02 |
