| 12 Jan 2026 |
 emily | I don't think we can do by-name/sw/swiftPackages; that doesn't follow the rules and nixpkgs-vet will complain about it. I'd like scopes in by-name, but it's not a thing we can really do right now. | 19:29:24 |
| 13 Jan 2026 |
 Randy Eckenrode | Not sure. The swbuild tool uses its own command-line parser. It’s pretty funky. A custom frontend could just use Swift Argument Parser and delegate to the library for everything else. | 01:06:56 |
| Randy Eckenrode | This is a helper for keeping the boilerplate down. I don’t see how we can allow switching build systems in a hook. One of the keys to bootstrapping is not pulling in the wrong build system. | 01:09:29 |
| Randy Eckenrode | lib.getDev and lib.getOutput "dev" work as expected. What other ones are there? | 01:10:15 |
| Randy Eckenrode | Ideally, I could get prebuilts working and drop all this stuff. | 01:10:26 |
| Randy Eckenrode | I didn’t think so. It’s a temporary thing so I can keep the old Swift stuff around to reference while I work. The final PR will probably do it like Tcl does with tcl-modules.nix. | 01:11:36 |
| Randy Eckenrode | If I can, I can revisit using custom helper hooks I guess. I hate hooks because Bash sucks, and debugging them sucks. | 01:12:51 |
| Randy Eckenrode | More pressing right now is how exactly to handle stdlib discovery. | 01:14:38 |
| emily | that would be nice for sure. what was the issue with that again? | 01:14:38 |
| Randy Eckenrode | It requires figuring out how it works and (probably) patching SwiftPM to allow it. | 01:14:56 |
 samasaur | aren't they only recently officially supported for swift-syntax only? | 01:15:14 |
| samasaur | and no support for any other package | 01:15:20 |
| Randy Eckenrode | There is some magic for Swift Syntax, but it seems to be really brittle. I was able to get it to parse a prebuilts entry in workspace-state.json, but SwiftPM didn’t do anything with it. | 01:15:31 |
| Randy Eckenrode | The more pressing need is figuring out how to discover the stdlib. | 01:15:47 |
| Randy Eckenrode | The current approach patches the compiler to find it at an arbitrary path, which isn’t very good for cross-compilation. | 01:16:19 |
