| 1 Dec 2025 |
 Randy Eckenrode | I think I’m going to have to package Swift 5.10.1. The C++ bootstrap compiler is useless. | 03:14:09 |
| Randy Eckenrode | This sucks because eventually the 14.4 SDK will be dropped, so what happens then? | 03:15:36 |
 emily | have you reported the bootstrapping issues upstream? | 03:44:01 |
| emily | I assume that the C++ compiler will become more useful over time | 03:44:14 |
| Randy Eckenrode | The problem is the 26.0 SDK exposes macros unconditionally, which the C++ bootstrap compiler can’t support because swift-syntax is written in Swift. | 04:12:08 |
| Randy Eckenrode | It appears to be working with the 14.4 SDK, but that makes me nervous about the long term maintainability. | 04:12:43 |
| Randy Eckenrode | emily: FYI, KosmicKrisp works with mpv and Wine. It doesn’t implement all the features yet that DXVK requires, but they will eventually. | 04:49:23 |
| Randy Eckenrode | SwiftPM 6.2 requires Swift Build. That’s lovely. | 14:17:54 |
 toonn | Welcome back, emily : ) | 14:20:25 |
| toonn | RandyEckenrode: Can you explain again why an extra intermediate Swift compiler doesn't work? I get wanting to avoid it but relying on the older SDK doesn't seem like a great workaround. | 14:22:22 |
| Randy Eckenrode | The problem (using the 26.0 SDK, which I stopped doing for now) is that going from Swift with no macro support to Swift with macro support requires a compiler with macro support, but all you have is one with no macro support. | 14:31:22 |
| toonn | Oh, what. And 5.10.1 is the missing link non-macro compiler supporting compiling a macro compiler? | 14:32:07 |
| Randy Eckenrode | Yes, but it requires the 14.4 SDK. | 14:41:38 |
| Randy Eckenrode | Fortunately, the 14.4 SDK does appear to work with Swift 6.2, so I’m doing that instead of building Swift 5.10.1. | 14:42:24 |
| toonn | Is the bootstrap compiler basically frozen or an active project? | 14:43:25 |
| Randy Eckenrode | The issue is more that our update plans involve removing the 14.4 SDK next year. | 14:44:14 |
| Randy Eckenrode | We may end up having to start from a binary bootstrap compiler. | 14:45:53 |
|  matthewcroughan changed their profile picture. | 14:59:06 |
 dotlambda | We might want to add some functionality to desktopToDarwinBundle:
https://github.com/NixOS/nixpkgs/pull/466480 | 16:58:00 |
| Randy Eckenrode | Yeah. There is more it can do. We can also drop the pre-11.0 workarounds too. | 17:05:30 |
| emily | isn't this something the bootstrap compiler could handle by ignoring the macros or such? | 17:10:50 |
| emily | the fact that the bootstrap compiler is being actively worked on implies that it's meant to be useful for something, no? | 17:11:03 |
| emily | all we need it to do is be able to compile Swift again right? | 17:11:21 |
| emily | nice! I wonder how performance is with mpv compared to MoltenVK. | 17:11:31 |
| Randy Eckenrode | Bootstrapping new platforms. | 17:12:14 |
| Randy Eckenrode | https://matrix.to/#/!lheuhImcToQZYTQTuI:nixos.org/$22gcuynbC7YnGvlCdcr8ATU2-lGn89JU7fcyrcOzoUk?via=nixos.org&via=matrix.org&via=nixos.dev | 17:13:37 |
| Randy Eckenrode | Not mpv, but mpv is probably not pushing the GPU very hard just for plain playback. The Matrix seemed fine playing back in 4k HDR. | 17:14:21 |
| emily | is that with e.g. profile=high-quality, maybe interpolation=on and a fancy tscale, possibly some Vulkan-using video filters? | 17:15:00 |
| emily | the out of the box settings don't tax the GPU, but adjusting it for high playback quality in various situations can be expensive enough that my old Intel machine would struggle | 17:15:29 |
| emily | some of the scaling algorithms are pretty expensive, I think profile=high-quality uses ewa_lanczossharp these days | 17:15:49 |
