| 7 Dec 2025 |
 WeetHet | I wonder if iTerm would be buildable from source after the swift work is done | 13:55:49 |
| WeetHet | And if people would even want it to be since that would mess the privileges on each update | 13:56:12 |
| WeetHet | * And if people would even want it to be since that would mess the privileges up on each update | 13:56:17 |
 Randy Eckenrode | We stopped building it because it relies on entitlements for security. | 13:56:56 |
| WeetHet | * And if people would even want it to be since that would mess the permissions up on each update | 14:01:14 |
| WeetHet | (That's the reason I'm running a binary build of Zed even though nixpkgs has a from-source one) | 14:01:50 |
| WeetHet | JIT and Apple Events don't need anything except a self signature AFAIK and the personal-information ones aren't exactly necessary for iTerm to function | 14:24:17 |
| WeetHet | At least providing a from source build with reduced entitlement set would be nice | 14:24:42 |
| Randy Eckenrode | Some things (like the AI and the browser plugin) enable the app sandbox via entitlement. | 14:29:41 |
| Randy Eckenrode | ninja: error: '/nix/store/fc9wkimiw706hdhxhk13mfcicaacjzlq-swift-llbuild-6.2.1/Library/Frameworks/llbuild.framework/llbuild.framework/llbuild', needed by 'lib/libSwiftDriverExecution.dylib', missing and no known rule to make it
| 21:44:40 |
| Randy Eckenrode | CMake is finding the framework, but it is very confused about where the framework is. Anyone know enough CMake to know why it would do that? | 21:44:56 |
| 8 Dec 2025 |
| Randy Eckenrode | Oh.
if(NOT TARGET llbuildSwift)
add_library(llbuildSwift UNKNOWN IMPORTED)
set_target_properties(llbuildSwift PROPERTIES
FRAMEWORK TRUE
INTERFACE_COMPILE_OPTIONS -F${llbuild_FRAMEWORKS}
IMPORTED_LOCATION ${llbuild_FRAMEWORKS}/llbuild.framework/llbuild)
endif()
| 00:42:21 |
| Randy Eckenrode | Probably that don’t the wrong thing. | 00:42:31 |
| Randy Eckenrode | Why does CMake set the deployment target to macOS 26 for Swift? ☹️ | 01:59:51 |
| Randy Eckenrode | It goes off the triple, which defaults to 26.0 on my system. That’s a lovely impurity. | 02:16:44 |
| Randy Eckenrode | Also, CMake doesn’t handle mixed language modules at all. You have to manually generate a bunch of yaml. | 02:17:01 |
| Randy Eckenrode | https://www.swift.org/documentation/articles/wrapping-c-cpp-library-in-swift.html#cmake | 02:17:14 |
| Randy Eckenrode | I wonder if Nix could (ab)use VFS overlays with Clang instead of doing the hacks we do today to make Nix stuff play nicely with unwrapped compilers. | 02:18:38 |
| Randy Eckenrode | Like just magically have the libc++ headers of your choice appear as if they were in the sysroot. | 02:19:06 |
| Randy Eckenrode | https://forums.swift.org/t/relationship-if-any-between-import-underlying-module-and-emit-objc-header/61287/4 | 02:21:47 |
 tiferrei | Hi folks, I seem to have corrupted my nix db somehow, I keep getting warning: error: SQLite database '/nix/var/nix/db/db.sqlite' is busy in most operations. What is the best way to sort this out? (Lix, nix-darwin) Thanks! | 10:23:49 |
| tiferrei | Fixed: Nuked the nix-darwin installation, then the whole nix installation. reinstalled nix and reapplied nix-darwin. | 12:31:58 |
 Ihar Hrachyshka | Is there a policy / rule / expectation that would not allow to merge breaking changes for darwin to bump a version of a package? context: https://github.com/NixOS/nixpkgs/pull/463817#issuecomment-3627032203
I'd think we shouldn't KNOWINGLY merge breaking changes, except in situations like security issues; and instead invest time in fixing problems before merge. But apparently there are alternative views. What's acceptable? | 15:16:04 |
| Randy Eckenrode | RFC-46 is about it. Darwin is not a tier 1 platform, so while not ideal, Darwin probably shouldn’t be blocking updates for Linux. | 15:18:34 |
| Ihar Hrachyshka | I don't understand why the rush. it's been 3 weeks total since the issue was even detected. | 15:22:29 |
 toonn | What is agreed upon is that breaking Darwin is OK after notifying darwin-maintainers and giving them a reasonable period to act. | 16:44:43 |
| Ihar Hrachyshka | reasonable does some heavy lifting here :) | 17:01:18 |
| toonn | Well in this case there wasn't even a ping to darwin-maintainers, no? | 17:03:49 |
| Ihar Hrachyshka | indeed. it was probably assumed I'll just take care of it since I "know" about the issue. | 17:11:08 |
| Ihar Hrachyshka | btw wonder if core team has opinions on way forward there. Looks like linker on darwin doesn't handle patch versions above a (high) number: https://github.com/ggml-org/llama.cpp/issues/17258
there are some hacks in the issue - splitting the "high" version number into pieces, overriding... - I wonder what would be acceptable in nixpkgs. | 17:13:56 |
