| 7 Nov 2025 |
 hexa | Randy Eckenrode, emily | 18:15:51 |
 samasaur | I think it should be fine to do that | 18:44:54 |
 niklaskorz | built samasaur's PR rebased on release-25.05, seems to work fine | 18:45:08 |
| samasaur | yeah apple-sdk_26 is entirely additive, and the whole SDK version/deployment target dance only depends on macOS version in the first place | 18:47:00 |
 Ihar Hrachyshka | I haven't looked closely into ff issue yet; I was wondering if they actually use some API that is not present in xnu sources or we can repeat a trick that we did before and pull from there. | 18:47:12 |
| Ihar Hrachyshka | but if we can just pull 26 into 25.05 it's probably a cleaner path of course. | 18:47:39 |
| samasaur | yeah i remember we had a previous firefox workaround that grabbed something from the 15.0 sources before apple-sdk_15 was added to nixpkgs | 18:48:13 |
| samasaur | +
# fileport.h was exposed in SDK 15.4 but we have only 15.2 in nixpkgs so far.
lib.optionalString
(
stdenv.hostPlatform.isDarwin
&& lib.versionAtLeast version "143"
&& lib.versionOlder apple-sdk_15.version "15.4"
)
''
mkdir -p xnu/sys
cp ${apple-sdk_15.sourceRelease "xnu"}/bsd/sys/fileport.h xnu/sys
export CXXFLAGS="-isystem $(pwd)/xnu"
'';
| 18:49:00 |
| samasaur | looks like sys/fileport.h is in the 15.5 and 26.0 SDKs | 18:49:48 |
| niklaskorz | hm is there a reason we didn't backport 15.5? | 18:50:03 |
 Randy Eckenrode | 15.2 to 15.5 was a big jump. Apple bumped libc++ major versions. | 18:51:13 |
| Randy Eckenrode | IIRC. | 18:51:16 |
| Ihar Hrachyshka | haven't I seen the same fileport.h error in the log posted for the latest breakage? do we just miss a backport for the above, somehow? (same missing header( | 18:51:16 |
| Ihar Hrachyshka | * haven't I seen the same fileport.h error in the log posted for the latest breakage? do we just miss a backport for the above, somehow? (same missing header) | 18:51:19 |
| samasaur | yeah i am coming to that same conclusion | 18:51:31 |
| Randy Eckenrode | I also tend to be fairly conservative about backports. | 18:51:42 |
| Ihar Hrachyshka | I won't be able to look at it till evening so if someone can backport the missing piece (which looks like what happened?) would be great. | 18:52:11 |
| Randy Eckenrode | Does the backport tag work? That’s where I would start. | 18:52:43 |
| Ihar Hrachyshka | it was backported though https://github.com/NixOS/nixpkgs/pull/437964 | 18:53:29 |
| samasaur | oh uh | 18:53:36 |
| samasaur | nvm | 18:53:37 |
| samasaur | In file included from Unified_mm_widget_cocoa1.mm:2:
/nix/var/nix/builds/nix-10098-3737624337/firefox-145.0/widget/cocoa/nsCocoaWindow.mm:7651:15: error: use of undeclared identifier 'NSGlassEffectView'
7651 | [[NSGlassEffectView alloc] initWithFrame:self.contentView.frame];
| ^
| 18:53:47 |
| samasaur | https://github.com/NixOS/nixpkgs/pull/459417#issuecomment-3503937655 | 18:53:51 |
| samasaur | looks like it actually does need 26 | 18:53:56 |
| Randy Eckenrode | For the 26.0 SDK. | 18:53:57 |
| samasaur | assumign NSGlassEffectView is new | 18:54:03 |
| samasaur | which it appears to be | 18:54:27 |
| niklaskorz | auto backport won't work I think, doesn't apply cleanly because of the missing 15.5 backport | 18:54:28 |
| Ihar Hrachyshka | ah some new development there... :( | 18:54:33 |
| niklaskorz | but I can make a manual backport | 18:54:38 |
