| 20 Apr 2026 |
 viraptor | I was not lucky... stdlib.dev does indeed contain SwiftShim, but it's not found. Unfortunately I know nothing about Swift module lookup system. I'll do some reading unless you know the answer immediately.
/nix/store/ifz6r2b8q64r73iq249h49yddficziy4-new_sdk/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib/swift/Swift.swiftmodule/x86_64-apple-macos.swiftinterface:5:8: error: Unable to find module dependency: 'SwiftShims'
import SwiftShims | 11:49:40 |
 Randy Eckenrode | Are you using swiftPackages.swift or .swiftc? | 11:51:21 |
| Randy Eckenrode | That SDK path is odd. | 11:52:13 |
| viraptor | .swiftc - I think it's working better now with .swift! thank you. | 11:53:14 |
| viraptor | 🎉 it did work. | 11:53:57 |
| Randy Eckenrode | The stdlib is separated from the compiler. .swift sets up a symlink farm with the stdlib, compiler, and other needed dependencies. | 11:54:22 |
| Randy Eckenrode | Like on Linux it will include Foundation. | 11:54:34 |
| Randy Eckenrode | Nix makes building multiple packages from one source an enormous pain in the ass, but we want these separate so you don’t pull in the full Swift closure just for linking to its stdlib (and for cross eventually). | 11:55:40 |
| viraptor | That makes sense. Ok, I'm getting yet-another-tool-missing error, but I can work with that. The success is close - I can smell it... | 11:55:49 |
| Randy Eckenrode | I was considering using Swift for some things, but this experience has me wanting to stick with .NET and F#. | 11:56:53 |
| Randy Eckenrode | Or <insert favorite non-Swift language/ecosystem here>. I just happen to really like F#. | 12:08:34 |
| Randy Eckenrode | The lack of a bootstrap story, the proprietary stuff in the toolchain (no testing on Darwin without Xcode!), the narrow vision on building iOS apps, the increasing complexity, etc | 12:10:31 |
| Randy Eckenrode | * | 12:10:44 |
| viraptor | F# is a fine choice. One day I'll get motivated enough to start somehting in it rather than C#... | 12:10:47 |
| Randy Eckenrode | There’s an open issue for vendoring deps that’s almost five years old. There’s no good story for Linux distros to package Swift applications and libraries. | 12:11:33 |
| Randy Eckenrode | My branch is a second attempt at vendoring because the first broke with SwiftPM 6.x. It works, but it’s a hack. | 12:12:51 |
| viraptor | I acquired some hate towards xcode recently for my own reasons so I'm happy to replace their proprietary tools one by one, in whatever way possible. (it's easy-ish when you just go for 1:1 compatibility - if the result matches, it's correct...) | 12:12:57 |
| Randy Eckenrode | SwiftPM’s dependency model is based on having a full Git repo. 😕 | 12:16:26 |
 kdn | how do I get dmesg/error logs for the linux-builder? | 12:57:57 |
| kdn | is there any way to add full disk access to Nix binaries (bash) on MacOS automatically? | 14:00:51 |
 Sarah Clark | Has anybody else hit a failing dav1d build when building a python package?
❯ nix-build -A dav1d
error:
… while calling the 'abort' builtin
at /Users/seclark/development/nixpkgs/lib/customisation.nix:339:7:
338| else
339| abort "lib.customisation.callPackageWith: ${error}";
| ^
340|
error: evaluation aborted with the following error message: 'lib.customisation.callPackageWith: Function called without required argument "xxHash" at /Users/seclark/development/nixpkgs/pkgs/by-name/da/dav1d/package.nix:9'
| 17:22:04 |
| Sarah Clark | I hit it while check-building python3Packages.celery | 17:22:43 |
| Randy Eckenrode | Yes. Sometimes my Jujutsu invents a commit that deletes xxHash and some other package. It’s really annoying. | 17:22:55 |
| Sarah Clark | Nuts. And yes, I'm using Jujutsu | 17:23:15 |
| Randy Eckenrode | I assumed it was because of the multi-merge branch I’m on that is rebased on master and staging. | 17:23:31 |
| Randy Eckenrode | I find which commit has those changes, split them out, then drop them. | 17:23:53 |
| Sarah Clark | I do have staging-next as one of my branches | 17:24:01 |
| Sarah Clark | Is there a straightforward way of doing this? | 17:24:39 |
| Randy Eckenrode | Unfortunately, no. You my have some luck finding the Git commit doing git log <path to xxHash> then looking up the change id from that. | 17:25:55 |
| Randy Eckenrode | * Unfortunately, no. You might have some luck finding the Git commit doing git log <path to xxHash> then looking up the change id from that. | 17:26:02 |
