I have this local agent using Qwen 3.6 (the version that is ~17 GB). Last time I tried using this a couple months ago it worked ok. Now the agent frequently does not know how to edit a file. It takes its proposed changes and completely overwrites the entire source file with the change instead of just modifying the portion of the source code. Sometimes the agent thinks it doesn't have permission to make a change and wants me to copy and paste code. Is there an Xcode skill provided somewhere that I can use with the Agent so it knows how to work in this environment properly? Currently I have it set up via "Chat" with "Tools Enabled" and "Automatically apply changes" is on.

I have a localized description of an app. I'm trying to add it in App Store Connect. Every time I save a red exclamation point is displayed with no explanation. Then all my edits disappear on reload. So I try to add a different screenshot in the new language before saving. But that fails with a red exclamation point as well but this time with an explanation: "Please save new locales before uploading screenshots or previews." Well I can't because of the first failure. In the web inspector I can see the following error: "status" : "400", "code" : "PARAMETER_ERROR.REQUIRED", "detail" : "The parameter 'filter[appStoreVersion]' is required but was not provided",

I'm testing iAP in a sandbox account (as configured in App Store Connect under 'Sandbox Testers'). So the in app purchase works. Cool. But I wanted to retry it. So I cleared the purchase history (both in App Store Connect and on my iPad in the 'Developer' section in Settings). But when I relaunch my app the purchase still validates and my app displays the item as 'unlocked'. Figure the receipt must still be cached so I nuke the app and completely reinstall it but it appears StoreKit is still getting the receipt and it isn't being cleared because my app is displaying it as 'purchased.' Also tried rebooting the iPad. But the sandbox purchase doesn't clear. I just did a sandbox test since it is closer to real life than StoreKit Configuration so I just wanted to do it a few times to make sure all is good but making a burner test account for every purchase is kind of tiresome. Anyone know of a workaround? I might just declare victory and go back to StoreKit Configuration.

I'll try to ask a question that makes sense this time :) . I'm using the following method on NSFileManager: (BOOL) getRelationship:(NSURLRelationship *) outRelationship ofDirectoryAtURL:(NSURL *) directoryURL toItemAtURL:(NSURL *) otherURL error:(NSError * *) error; Sets 'outRelationship' to NSURLRelationshipContains if the directory at 'directoryURL' directly or indirectly contains the item at 'otherURL', meaning 'directoryURL' is found while enumerating parent URLs starting from 'otherURL'. Sets 'outRelationship' to NSURLRelationshipSame if 'directoryURL' and 'otherURL' locate the same item, meaning they have the same NSURLFileResourceIdentifierKey value. If 'directoryURL' is not a directory, or does not contain 'otherURL' and they do not locate the same file, then sets 'outRelationship' to NSURLRelationshipOther. If an error occurs, returns NO and sets 'error'. So this method falsely returns NSURLRelationshipSame for different directories. One is empty, one is not. Really weird behavior. Two file path urls pointing to two different file paths have the same NSURLFileResourceIdentifierKey? Could it be related to https://developer.apple.com/forums/thread/813641 ? One url in the check lived at the same file path as the other url at one time (but no longer does). No symlinks or anything going on. Just plain directory urls. And YES calling -removeCachedResourceValueForKey: with NSURLFileResourceIdentifierKey causes proper result of NSURLRelationshipOther to be returned. And I'm doing the check on a background queue.

I'm trying to use NSTokenField for the first time. So my custom 'representedObject' for a token has additional model data tied to it (not just the editing/display string). I noticed when I edit an existing token, type text, and hit the enter key I get the following delegate callback: - (nullable id)tokenField:(NSTokenField *)tokenField representedObjectForEditingString:(NSString *)editingString; This same delegate method is called when I type a brand new token. Is there a way to distinguish if I'm editing a token vs. creating a new one? My expectation is to be able to do something like this (made up enhancement): - (nullable id)tokenField:(NSTokenField *)tokenField representedObjectForEditingString:(NSString *)editingString atIndex:(NSUInteger)existingTokenIndex { if (existingTokenIndex == NSNotFound) { // Token is new, create a new instance MyTokenObject *newToken = //create and configure. return newToken; } else { // This would update the editing string but wouldn't discard existing data held by the token. MyTokenObject *tokenObj = [self existingTokenAtIndex:existingTokenIndex]; tokenObj.editingString = editingString; return tokenObj; } }

I have an NSTextView subclass and implements drag and drop for custom draggable data, so I override -writablePasteboardTypes and add my own type as described in the header file: // Returns an array of pasteboard types that can be provided from the current selection. Overriders should copy the result from super and add their own new types. @property (readonly, copy) NSArray<NSPasteboardType> *writablePasteboardTypes; Now my textview also accepts the drop. Drag and drop can be used to move the custom data to a different location in the text view's character range. Like grabbing a block of text and moving it. So I accept the drop in -readSelectionFromPasteboard:type: I have a variable I cache at the start of dragging like: _myDraggingItem = // Set at the start of dragging. Then when I accept the drop I just use _myDraggingItem to move it to the drop location in the text view. I don't need to actually serialize the entire object and write it on the pasteboard I can just use _myDraggingItem to move from the source location to destination location. This is local only drag and drop. it works, except when the mouse leaves the text view briefly during the dragging session. This is because I override NSTextView's - (void)cleanUpAfterDragOperation // If you set up persistent state that should go away when the drag operation finishes, you can clean it up here. Such state is usually set up in -dragOperationForDraggingInfo:type:. You should probably never need to call this except to message super in an override. - (void)cleanUpAfterDragOperation; So documentation indicates -cleanUpAfterDragOperation is for clean up after drag operation finishes so I nil out _myDraggingItem here. But -cleanUpAfterDragOperation is getting called from [NSDragDestination _draggingExited]. -draggingExited: means the drag location moved outside the view it does not mean that the dragging session is over. The drag can move back inside the view after briefly exiting, so this isn't a usable place to clean up state tied to the dragging session as the header file indicates. I must override -draggingEnded: instead. If that sounds like a bug let me know.

I typically avoid NSMetadataQuery because I always found the API to be a bit peculiar but for this feature I'm working on I'm not sure it is worth the effort to implement this functionality in my own way. Plus it seems pretty fast. What I find strange is I set the operationQueue to get notifications off the main thread. But also when I set the queue the system yells at me anytime I make a change to NSMetadataQuery that alters the query. So I found this recommendation (requirement) in the documentation : NSMetadataQuery *query = // Initialize and set up a query [query.operationQueue addOperationWithBlock:^{ [query startQuery]; }]; I find this API design to be odd, but maybe I'm just weird. So there are a bunch of properties that can be changed while the query is already running (predicate etc.) and they implicitly stop/start the query and it seems all these calls need to be routed through the query.operationQueue like above? For example if I change the predicate while the query is already started it seems I have to: [query.operationQueue addOperationWithBlock:^{ query.predicate = predicate; }]; The query already knows its operationQueue. Why does the caller have to plumb these calls through the operation queue manually? -stopQuery does not result in an error/warning when called off the operationQueue but is doing so safe? Or do I have to do: [query.operationQueue addOperationWithBlock:^{ [query stopQuery]; }]; Really what I was expecting was to provide a queue for the notification callbacks. I wasn't expecting to manually have to confine the query to its own queue.

I probably should know the answer to this but I don't, so I'll ask. When I enable state restoration and have like three windows open, and I quit the app (via Quit menu, not 'Stop In Xcode') the following happens: -NSApplication calls _closeForTermination on all the windows and this causes state restoration to fail to restore these open windows on next app launch. This behavior is not aligned with the behavior of apps like Mail. If I have two "Viewer Windows" open in Mail and I quit the app, when I relaunch the two viewer windows are restored. I can of course track this and write data to restore for these auto closed window myself but shouldn't there be an easy way to opt in to this behavior?

So after binding to NSBrowser selectionIndexPaths: https://developer.apple.com/library/archive/documentation/Cocoa/Reference/CocoaBindingsRef/BindingsText/NSBrowser.html this causes NSBrowser to do some weird stuff that seems completely unrelated to this particular binding. It starts calling NSBrowserCell methods on my cells but my cell is not a NSBrowserCell. My cell is actually a subclass of NSTextFieldCell. But NSBrowser starts sending setIsLeaf: (which I don't implement). In any case if I implement the -setLeaf: that solves that unrecognized selector, but now my cells don't draw titles. Not sure why binding to selectionIndexPaths causes this behavior? The cell stuff seems unrelated to this particular binding. I am of course using the newer but still pretty old item based APIs... do these bindings only support using NSMatrix? I do set the binding after calling -setCellClass: but makes no difference. I also just tried overriding -setSelectionIndexPaths: but NSBrowser does not use the setter.

I'm trying to get the count of how many windows an iPadOS app has 'open' (open from the user's perspective in the app switcher). This is for the sake of determining whether I should show or hide a button that takes action on every window (if there is only 1 window, the button will be hidden). According to the documentation the proper API for this is this property on UIApplication: // All of the representations that currently have connected UIScene instances or had their sessions persisted by the system (ex: visible in iOS' switcher) @property(nonatomic, readonly) NSSet<UISceneSession *> *openSessions So I print the count (only sessions with role UIWindowSceneSessionRoleApplication) when scenes are added/removed etc via appropriate lifecycle notifications like -sceneDidDisconnect: -sceneDidBecomeActive: and so forth. What I noticed is when I add a new window scene, the count increases by one so cool, that works. But when I kill a window in the App switcher the count does not decrease. I can end up in a situation where the app has only 1 window in the app switcher but the count prints 8, so this is wrong. So am I using the wrong API? How can I just get scene count in the app switcher? The documentation makes it seem like using 'connectedScenes' for this would be wrong because that property is not supposed to include 'archived' scenes in the app switcher (or is it?)? I do know I can't take action on an archived scene 'yet' but I would still show the button because whether or not the scene is archived in the app switcher is a fact that remains hidden from the user. My code will take care of that later after state restoration. Is iPadOS 26 potentially keeping scene sessions open for too long? Is there a good way to reliably detect how many scenes I have in the app switcher? Is a scene session explicitly killed by the user supposed to remain the .openSessions set? I am testing on the Simulator FWIW. iPad 26.5.

So in NSCollectionView I left click the background and start dragging to make the 'selection box' (is that what we call it?) I have a mouse with separate left and right click buttons (i.e., not an Apple Magic Mouse). Now if I accidentally hit the right click while I'm still dragging the selection box..... the right click context menu shows up in the middle of mouse dragging and the selection box freezes on screen. Besides the visual disruption this sometimes results in weird behavior. Sometimes mouse clicks stop responding for awhile in the window and I gotta click around a bunch of times until I guess the run loop "unclogs." I'm able to reproduce in Finder Icon view so it isn't just me. Steps to Reproduce: A mouse with separate buttons for left and right click. Open Finder window in icon view. Left click the background and start dragging to make the gray selection box. While dragging the mouse as you are selecting click right-click in the middle of the drag. The selection box freezes on screen. Expected Behavior: -Right click event and left mouse drag event normally shouldn't overlap. If left mouse is being dragged the right click event should be blocked. Let me know if this is Feedback worthy.

I just noticed this but am curious: Why does Xcode use list.bullet SF Symbol to toggle the "navigators" (sidebar) instead of sidebar.left? For the inspectors Xcode uses "sidebar.right" but I don't understand why the left sidebar uses a different icon from the right sidebar? Shouldn't the left sidebar icon match the right sidebar (just flipped)? Am using the wrong SF Symbol in my app to toggle the sidebar? Is the list.bullet the recommended symbol for this now? Every other app seems to use sidebar.left.

When I add a simple menu to the dock via the NSApplicationDelegate method -applicationDockMenu: and run the app from Xcode it doesn't work. -(NSMenu*)applicationDockMenu:(NSApplication*)sender { NSMenu *dockMenu = [self buildDockMenu]; if (dockMenu != nil) { NSLog(@"Returning dock menu."); return dockMenu; } else { NSLog(@"Not ready to build dock menu"); return nil; } } When I run the app, my main app window shows up but nothing logs out in -applicationDockMenu: until I click outside my app's window (so if I click the desktop background, or a Finder window, or whatever). Then after I click outside my app's main window this logs out: Returning dock menu. The "Not ready to build dock menu" message does not log out. But...when I right click on the dock icon, the menu doesn't show up. But if I stop the app from Xcode and just run it not attached to the debugger, the dock menu does show up. But this makes the debugging/testing situation not ideal.

So I'm working on adding another component to my app that uses NSOutlineView, as we do in AppKit. There will probably always be less than 25 rows here. One row is much larger than the others. Not sure if any of this matters. What I know is I noticed scrolling it is very jank. It's going way too fast. So I took a peek and see lineScroll is getting is 304 in Interface Builder. Not sure how that happened. I changed it to like 24. Then Interface Builder automatically changes it back to 304. So in -viewDidLoad I just set it: NSScrollView *scrollView = self.outlineView.enclosingScrollView; scrollView.verticalLineScroll = 24.0; scrollView.lineScroll = 24.0; But scrolling still is busted. So I subclass NSScrollView and override the setters. For some reason, NSTableView's -tile method is deciding to change the lineScroll to 304, all on its own. So every time tile is called. line scrolls get reset to 304.

I have (had) a view controller that does a bit of manual layout in a -viewDidLayout override. This was pretty easy to manage - however since introducing NSGlassEffectView into the view hierarchy I sometimes am getting hit with "Unable to simultaneously satisfy constraints" and Appkit would break a constraint to 'recover.' It appears translatesAutoresizingMaskIntoConstraints is creating some really weird fixed width and height constraints. Here I wasn't doing any autolayout - just add the glass view and set its frame in -viewDidLayout. At runtime since I do manual layout in -viewDidLayout the frames are fixed and there is no real "error" in my app in practice though I wanted to get rid of the constraint breaking warning being logged because I know Autolayout can be aggressive about 'correctness' who knows if they decide to throw and not catch in the future. In my perfect world I would probably just prefer a view.doesManualLayout = YES here - the subviews are big containers no labels so localization is not an issue for me. Rather than playing with autoresizing masks to get better translated constraints I decided to set translatesAutoresizingMaskIntoConstraints to NO and make the constraints myself. Now I get hit with the following exception: "The window has been marked as needing another Layout Window pass, but it has already had more Layout Window passes than there are views in the window" So this happens because the view which now has constraints -- I adjusted the frame of it one point in -viewDidLayout. My question is - is not legal to make changes in -viewDidLayout - which seems like the AppKit version of -viewDidLayoutSubviews. In UIKit I always thought it was fine to make changes in -viewDidLayoutSubviews to frames - even if constraints were used - this is a place where you could override things in complex layouts that cannot be easily described in constraints. But in AppKit if you touch certain frames in -viewDidLayout it can now cause this exception (also related: https://developer.apple.com/forums/thread/806471) I will change the constant of one of the constraints to account for the 1 point adjustment but my question still stands - is it not legal to touch frames in -viewDidLayout when autolayout constraints are used on that subview? It is (or at least was if I remember correctly) permitted to change the layout in -viewDidLayoutSubviews in UIKit but AppKit seems to be more aggressive in its checking for layout correctness). What about calling -sizeToFit on a control in viewDidLayout or some method that has side effect of invalidating layout in a non obvious way, is doing things like this now 'dangerous?' Shouldn't AppKit just block the layout from being invalidated from within -viewDidLayout - and leave whatever the layout is as is when viewDidLayout returns (thus making -viewDidLayout a useful place to override layout in the rare cases where you need a sledgehammer?)