Thanks Quinn! Another thing that could help these scenarios is if pthread_chdir_np and pthread_fchdir_np were public, as you could then safely chdir and use a short relative sun_path. Filed as FB23555258. Thanks!

Thanks!

Thanks! :)

I posted two follow up questions here https://developer.apple.com/forums/thread/807374 and here https://developer.apple.com/forums/thread/80737 Thanks!

Thank you so much! 🙌🏻 And sorry for the delay in replying, have been down with a nasty cold 🤧

Thanks for your insights Kevin! I've added a standalone comment explaining Qt's use of NSApplication

Thanks Albert! See additional comment below.

I've since learned about launchctl asuser, which allows me to launch a process over SSH with an Aqua session, and with stdin/out/err attached: torarne@macos-15-5-0 % ./foo Session ID: 100044 sessionIsRoot: false sessionHasGraphicAccess: false sessionHasTTY: true sessionIsRemote: true ^C torarne@macos-15-5-0 % sudo launchctl asuser 501 ./foo Session ID: 100003 sessionIsRoot: false sessionHasGraphicAccess: true sessionHasTTY: true sessionIsRemote: false ^C

This seems to work now (in Xcode 16.4 at least)! Thanks to whoever fixed this! ❤️

You can create NSAdaptiveImageGlyphs for the emoji keyboard by turning photos into "stickers". As far as I can tell this uses the same underlying APIs for text as Genmojis will when released.

Agreed. A clear indication up front (UI or warning/error in system logs or as detailed in the returned error in the network API layer) that the executable can't even ask for network permissions would have helped diagnose this.

It's a reasonable bug, however, IMHO the "fix" here should be to reject the extension as invalid at load time. Executable need to include build UUIDs. Agreed. A clear indication up front (UI or warning/error in system logs or as detailed in the returned error in the network API layer) that the executable can't even ask for network permissions would have helped diagnose this.

Posted https://developer.apple.com/forums/thread/761001 tl;dr, golang doesn't produce LC_UUID, which is important for the network privacy subsystem

Nice, that helps to see the properties! Unfortunately we use a CAMetalLayer in an NSView, not MTKView, so we're not seeing those properties shown in the screenshot above. Is there a way to drill down in the inspector from a NSView to its layer?

The reason for the synthetic event not behaving the same as the real key event was that the synthetic event didn't have an eventRef, causing _matchesKeyEquivalent to bail out early. Calling syntheticEvent.eventRef (https://developer.apple.com/documentation/appkit/nsevent/1525143-eventref?language=objc) forces NSEvent to lazily create the eventRef.