| 8 Jun 2021 |
 toonn | I think so. | 22:07:22 |
| toonn | I'd even go so far as to say none, thought I'm not 100% on that. | 22:07:40 |
 LnL | guess you could ask whether it's worth to deal with the source releases, which often have missing stuff, given that tbd files are in place now | 22:11:13 |
| LnL | but for now a good distinction is probably our own CF from swift-corelibs and whatever that thus depends on | 22:12:22 |
| LnL | we also have our own generated stubs, currently those are used for both header only source releases and frameworks | 22:16:17 |
| LnL | the later can come from the sdk in the future but I'm not sure about the former | 22:16:37 |
| LnL | emily: I'd be interested to know if building something that requires >=11.0 features works on an older release (eg. building on 10.15, copying the result to 11.0 and running it there) | 22:21:22 |
| LnL | if that works it means we're decoupled enough from the host system and then I don't see any reason not to expose "future" sdk versions for packages that need it | 22:24:06 |
 emily | hard to see how it wouldn't at least, it's just extracting the same SDK files into the nix store when all's said and done right? not like we use any system compilers or anything that could choke on the source | 22:25:19 |
| emily | I don't have access to a macOS <11 machine to test this with though :( | 22:25:27 |
| emily | I used the macosPackages_11_0_1 thing while trying to get Swift to compile on Darwin | 22:25:39 |
| LnL | well, we're still linking against a 10.12 Libsystem etc. which I'm not sure would work but in theory yeah we should be isolated enough | 22:26:44 |
| LnL | I have a machine that's still on 10.15 if you have an expression that links against 11.x features I could build it | 22:27:47 |
| toonn | emily: Any headway into that? It's something I've gotten stuck on before. | 22:30:56 |
| emily | some, but it's still a pile of hacks that doesn't fully build | 22:31:42 |
| emily | i had to pull in a lot of newer stuff from SDKs / source releases | 22:31:55 |
| emily | and ultimately swift doesn't officially support the SDK version NixOS is on so it'd probably be a lot easier with a bump :p | 22:32:36 |
| emily | I think 10.15 is the lowest they go down to | 22:32:45 |
| toonn | Working on it (bump to 10.13 for now) but all these stdenv builds are harshing the buzz. | 22:34:37 |
| 9 Jun 2021 |
| LnL | toonn: FYI https://hydra.nixos.org/build/145123408#tabs-summary | 11:31:20 |
| toonn | LnL: Great : ) | 16:55:04 |
| toonn | Now I wait until this bootstrap-tools is uploaded so I can add a commit and push the LLVM 11 stdenv PR? | 16:55:35 |
| LnL | oh, it's actually done | 16:56:53 |
| LnL | * oh, it's actually already done | 16:56:59 |
| LnL | you can make a commit like this for now, I'll upload it to the mirror once everything looks good https://github.com/NixOS/nixpkgs/pull/85151/commits/9989b7dd8cfbda0d097442c78430a76a0195fa71 | 16:58:31 |
| LnL | hmm wait, where are the build products... | 16:59:28 |
| LnL | :( | 17:00:39 |
| LnL | looks fine to me at first glance https://gist.github.com/LnL7/17c67c2da4a01167734e48fccdb294d8 | 17:04:23 |
| toonn | THat's the same tarball hash I got. | 17:04:58 |
| LnL | yeah but that hydra-build-products file should result in a list of links on the hydra build like this | 17:05:32 |
