| 22 Oct 2024 |
 Randy Eckenrode | Unless there’s something else broken like rcodesign, I’m probably going to spend the next few days updating the stdenv documentation for Darwin. | 01:00:03 |
 samasaur | https://developer.apple.com/documentation/servicemanagement/updating-your-app-package-installer-to-use-the-new-service-management-api This app works without having to graphically run the .pkg installer, just running the app executable directly. However, it was built with real Xcode. I'm investigating now to see if it'll work with ad-hoc signatures instead of my real developer ID | 01:31:05 |
 Ihar Hrachyshka | https://github.com/NixOS/nixpkgs/pull/350384 | 01:43:03 |
| Ihar Hrachyshka | what's the approach to pick the sdk version for a package? the minimal possible that is still supported by upstream? or should I pick the latest instead? My interpretation of https://github.com/NixOS/nixpkgs/pull/338695 is the former, but perhaps I misunderstood? | 02:28:18 |
| Randy Eckenrode | That’s a good question. I don’t cover it in the documentation update, but I should. | 02:30:16 |
| Randy Eckenrode | As a rule of thumb, use the latest indicated by upstream and set the minimum version with the hook. | 02:30:30 |
| Ihar Hrachyshka | latest MINIMAL indicated by upstream you mean? | 02:30:49 |
| Randy Eckenrode | This assumes upstream is well-behaved using availability checks. | 02:30:51 |
| Randy Eckenrode | For example, Qt 6 builds with the 15.0 SDK and propagates a minimum version of 12.0. | 02:31:14 |
| Randy Eckenrode | Because that is what upstream specifies. | 02:31:20 |
| Ihar Hrachyshka | firefox has a check for 14.4+, so I wonder if I should/can bump to 15 if it still compiles. My reading is probably no. | 02:31:24 |
| Randy Eckenrode | I would probably not go beyond what upstream indicates. | 02:31:37 |
| Randy Eckenrode | I should also add that to the documentation. | 02:31:46 |
| Ihar Hrachyshka | I will wait for documentation :) | 02:32:00 |
| Ihar Hrachyshka | chat is a bit vague and I don't want it to become a word picking contest :p | 02:32:15 |
| Ihar Hrachyshka | I will review when it's up! | 02:32:24 |
| Randy Eckenrode | The reason is Apple uses the SDK version in the binary to vary behavior at runtime. Sometimes the change in behavior can be breaking, and the application has to be updated to take it into account. | 02:32:30 |
| samasaur | SUCCESS | 02:35:57 |
| samasaur | 
Download image.png | 02:36:15 |
| samasaur | this is not yet Nix-built but it isn't built with Xcode | 02:36:23 |
| samasaur | It does need Swift, though, FWIW | 02:38:51 |
| samasaur | probably could be done with Objective-C instead | 02:39:08 |
| Randy Eckenrode | https://github.com/NixOS/nixpkgs/pull/350395 | 03:09:50 |
| Randy Eckenrode | That’s the documentation piece of the Darwin refactor. I tried to touch on the various ways the SDKs and darwinMinVersionHook can be used (as well as cross-compilation and darwinMinVersion/darwinSdkVersion). | 03:10:34 |
| Randy Eckenrode | Please review for content, clarity, and wording. 🙂 | 03:10:48 |
|  jmartindf joined the room. | 04:28:07 |
| samasaur | hmm, it looks like this may only work for LaunchAgents, not LaunchDaemons — at least, LaunchDaemons must be manually approved in System Settings | 04:47:58 |
| Randy Eckenrode | Having launch agents with nice names is still a big improvement. | 04:55:35 |
| samasaur | okay, question: I am pretty sure I need to sign these app bundles, if only with an ad-hoc signature. Should I use codesign from sigtool or rcodesign? I've only ever used normal codesign before | 05:19:27 |
| samasaur | (the answer is that codesign really does not seem to work that well and rcodesign worked flawlessly) | 06:15:29 |
