| 9 Nov 2025 |
 samasaur | swift is broken on staging-next :( | 02:56:45 |
| samasaur | which i think is because of it | 02:56:51 |
 Randy Eckenrode | Yes. | 02:57:10 |
| samasaur | because when building the standard library the target is set to 10.9 for some reason | 02:57:19 |
| Randy Eckenrode | IIRC I didn’t run into that with my Swift work because I forced the deployment target to be the default for nixpkgs. | 02:58:18 |
| samasaur | what's odd is that the cmake flags are printed before that subproject is built and the last flag is -DSWIFT_DARWIN_DEPLOYMENT_VERSION_OSX=14.0 | 02:58:51 |
| Randy Eckenrode | Is it failing in Swift or some other component? | 03:04:43 |
| samasaur | appears to be Swift itself | 03:05:37 |
| Randy Eckenrode | I also set (lib.cmakeFeature "SWIFT_HOST_TRIPLE" swiftTriple) where swiftTriple = lib.replaceStrings [ "darwin" ] [ "macosx${stdenv.hostPlatform.darwinMinVersion}" ] stdenv.hostPlatform.config; | 03:06:47 |
| Randy Eckenrode | * I also set (lib.cmakeFeature "SWIFT_HOST_TRIPLE" swiftTriple) where swiftTriple = lib.replaceStrings [ "darwin" ] [ "macosx${stdenv.hostPlatform.darwinMinVersion}" ] stdenv.hostPlatform.config; | 03:07:04 |
| Randy Eckenrode | You might be able to work around it by setting the flag to suppress the error if that doesn’t work. | 03:07:50 |
| samasaur | yeah I'm pretty sure that suppressing the error would work, but ideally we could get it to target the correct version | 03:08:41 |
| samasaur | though I suppose it doesn't matter all that much | 03:08:47 |
| Randy Eckenrode | Swift is built twice IIRC in the old derivation. | 03:10:40 |
| Randy Eckenrode | Is it setting SWIFT_DARWIN_DEPLOYMENT_VERSION_OSX both times? | 03:10:49 |
| Randy Eckenrode | In particular for the concurrency backdeploy, which is useless with a 14.0 minimum version. | 03:11:03 |
| Randy Eckenrode | It looks like it. I’d try forcing the host triple. If that doesn’t work, then suppress the error for now. | 03:12:16 |
| samasaur | yes, i believe it is setting it for both, but i also think it's failing on the first one | 03:12:35 |
| Randy Eckenrode | I was building 5.10.1 with this warning as an error, so it should be possible. It’s just been so long I don’t remember whether I did something else to suppress the error. | 03:13:10 |
| Randy Eckenrode | * | 03:13:18 |
| samasaur | I think this should fix the error honestly | 03:13:29 |
| samasaur | am just mid-build at the moment so I'm letting this finish first | 03:13:40 |
| samasaur | okay this does not work | 03:43:57 |
| samasaur | trying -Wno-error=unguarded-availability in NIX_CXXFLAGS_COMPILE | 04:10:25 |
| samasaur | that didn't work (though i could have done it worng) | 04:23:31 |
| samasaur | * that didn't work (though i could have done it wrong) | 04:23:35 |
| samasaur | regardless I think I found the real root of the issue which is the COMPATIBILITY_MINIMUM_DEPLOYMENT_VERSION_OSX cmake variable, which defaults to 10.9 | 04:24:25 |
| samasaur | running a build where I override that to 14.0 | 04:24:39 |
| samasaur | which does make the "back" part of the backdeployment less effective | 04:24:59 |
| Randy Eckenrode | We don’t support older though. (I wonder why I didn’t have this issue on my branch?) | 04:56:43 |
