“show when run”

Area: CoreML / Machine Learning Describe the issue: On iOS 26.4, calling MLModel(contentsOf:configuration:) to load an .mlpackage model hangs indefinitely and eventually kills the app via watchdog. The same model loads and runs inference successfully in under 1 second on iOS 26.3.1. The hang occurs inside eort_eo_compiler_compile_from_ir_program (espresso) during on-device AOT recompilation triggered by MLE5ProgramLibraryOnDeviceAOTCompilationImpl createProgramLibraryHandleWithRespecialization:error:. A C++ exception (__cxa_throw) is thrown inside libBNNS.dylib during the exception unwind, which then hangs inside __cxxabiv1::dyn_cast_slow and __class_type_info::search_below_dst. Swift's try/catch does not catch this — the exception originates in C++ and the process hangs rather than terminating cleanly. Setting config.computeUnits = .cpuOnly does not resolve the issue. MLE5ProgramLibrary initialises as shared infrastructure regardless of compute units. Steps to reproduce: Create an app with an .mlpac

Hi all, I've been working on a pure-Swift port of Google's Gemma 4 text decoder that plugs into mlx-swift-lm as a sidecar model registration. Sharing it here in case anyone else hit the same wall I did, and to get feedback from the MLX team and the community before I propose anything upstream. Repo: https://github.com/yejingyang8963-byte/Swift-gemma4-core Why As of mlx-swift-lm 2.31.x, Gemma 4 isn't supported out of the box. The obvious workaround — reusing the Gemma 3 text implementation with a patched config — fails at weight load because Gemma 4 differs from Gemma 3 in several structural places. The chat-template path through swift-jinja 1.x also silently corrupts the prompt, so the model loads but generates incoherent text. What's in the package A from-scratch Swift implementation of the Gemma 4 decoder (Configuration, Layers, Attention, MLP, RoPE, DecoderLayer) Per-Layer Embedding (PLE) support — the shared embedding table that feeds every decoder layer through a gated MLP as a third residual KV sharing

Summary I'm implementing NEURLFilter with the com.apple.developer.networking.networkextension.url-filter-provider entitlement for a system-wide URL filtering feature. The feature works perfectly in development-signed builds (connecting successfully to my PIR server over extended testing) but every production-signed build fails before any network call is made. NEURLFilterManager reports .serverSetupIncomplete (code 9). After installing the NetworkExtension debug profile, the unredacted com.apple.CipherML logs reveal the cause: no privacy proxy is provisioned for this bundle identifier, and the connection is configured proxy fail closed. Environment iOS 26 Entitlement: com.apple.developer.networking.networkextension.url-filter-provider Extension point: com.apple.networkextension.url-filter-control PIR server configured via NEURLFilterManager.setConfiguration(...) Privacy Pass issuer configured Dev-signed builds: working correctly, connecting to the PIR server Production-signed builds (both TestFlight and distri

We've filed FB22448572 for this, but posting here in case others are hitting the same issue. After upgrading macOS from 26.3.2 to 26.4, SecItemCopyMatching returns errSecAuthFailed (-25293) when reading kSecClassGenericPassword items from the default login keychain. The keychain reports as unlocked, but all authenticated operations fail. The error doesn't self-resolve — we've observed it persisting for 7+ minutes across repeated calls and process restarts. The only workaround we've found is SecKeychainLock(nil) followed by SecKeychainUnlock(nil, 0, nil, false), which prompts the user for their password and clears the stale state. Apple's own security CLI tool also fails while the keychain is in this state: $ security show-keychain-info ~/Library/Keychains/login.keychain-db security: SecKeychainCopySettings .../login.keychain-db: The user name or passphrase you entered is not correct. The trigger seems to be process lifecycle — a new process accessing the keychain early in startup (e.g., from the app

I've been trying over the past few days to use a PIV-programmed Yubikey to perform mTLS (i.e. mutual client cert auth) in my custom app. My understanding is that I need to feed NSURLSession a SecIdentity to do so. Yubico's instructions state that I need their Yubico Authenticator app for this, but this directly contradicts Apple's own documentation here. I dont need NFC/lightening support, and I only need support for my specific app. When I plug in my key to my iPhone and have TKTokenWatcher active, I DO see com.apple.pivtoken appear in the logs. And using Yubico's SDK, I CAN get data from the key (so I'm pretty sure my entitlements and such are correct). But using the below query to get the corresponding (fake? temporary?) keychain item, it returns NULL no matter what I do: let query: [String: Any] = [ kSecClass as String: kSecClassIdentity, kSecReturnRef as String: true, kSecAttrTokenID as String: apple.com.pivtoken, // Essential for shared iPads kSecMatchLimit as String: kSecMatchLimitOne ] var item: CFTyp

Hello, I’ve noticed an issue with custom keyboard extensions in iOS 26 that seems specific to native apps. When a custom keyboard is opened in apps like Messages, Notes, or Safari, there’s an extra strip of native grey space around the left, right, and top edges of the keyboard. This extra margin cannot be rendered over by the keyboard’s own views. Interestingly, this behaviour does not occur in third-party apps like Instagram. It also wasn’t present in earlier iOS versions. The result is that keyboards with custom or non-grey backgrounds look visually inconsistent (they appear framed by unwanted grey). Has anyone else run into this? Is this a known change in iOS 26, or could it be a bug? Any guidance or official clarification would be appreciated. Thanks!

Hi there! I'm trying to get my first app released. It's been 40 rejections for over 2 months. And there is a TV show which I am hoping to go on, but I can't without the app being released. Some of the issue has been my newbieness to Apple admittedly but things have become bizaare this week particularly. Each review is a drip feed of one or two problems rather than a full review of all problems in one go - surely this has a cost associated which Apple is bearing unnecesarily? In the last week, all rejection reasons could have been stated months ago. The problem is that it appears that a bot is testing the app not a human. For instance a reviewer flagged that a Continue button is not pressable. When in reality I added an intentional delay of 120ms for app stability, the reviewer didn't or couldn't try to tap the button twice. Separate example: twice reviewers have brought up my sign in button has not worked (04/01 & 03/16) - and twice I have explained that non subscribers are not allowed to sign in

I need to bring attention to a very suspicious copycat incident that suggests a potential metadata scraping issue. I submitted my original sports insights app, KickPick, for review on April 2nd, 2024. To my shock, I discovered a competitor app already live on the store with same app name and nearly identical app information (descriptions, metadata) and a very similar UI structure. Domain Registration: The infringing developer registered their website domain on the exact same day I submitted my app for review (April 2nd). This is not a coincidence; it points to automated monitoring or scraping of new submissions/metadata. Plagiarized Legal Assets: While the UI has some variations, they have copied my Privacy Policy and Terms of Service verbatim. These documents were custom-written for my specific project logic, yet they appear on their site with only the company name changed. App Information: The app descriptions and store metadata are almost identical to what I provided in my pending submission. It is highly

Context: We are developing an SDK to support global age verification regulations (e.g., Texas HB 18, Brazil’s LGPD). We plan to use the DeclaredAgeRange and PermissionKit frameworks. We want to verify if our proposed Block-by-Default sequence for non-compliant states is legally and policy-compliant according to Apple’s standards. Detailed Workflow Description: Initial Authentication: After the user logs in, the SDK calls requestAgeRange(ageGates: 13, 16, 18). Handling Sharing Status: If Declined: If the user declines age sharing (.declinedSharing), the SDK blocks app access and displays a popup guiding them to enable sharing in System Settings. Age Verification Results: Adult (VERIFIED, 18+): Immediate access to the game. Non-Regulated Region (UNKNOWN): Access to the game is allowed. Minor (SUPERVISED, 13-17): Step A (Age Gate): We check if they are 13+. If they are under 13, we block access and show an Underage notice. Step B (Family Sharing): If they are 13+, we check if Family Sharing is linked. I

Two Previously Approved Apps Pending Review — Business Event Tomorrow April 9th Hello everyone, I'm reaching out here hoping for guidance or assistance from the App Review team, as we are running out of time. We have two apps that were previously approved, had their publication canceled to implement minor updates, and are now stuck in the review process with our business event happening tomorrow, April 9th. Here is the full timeline: — App ID: 6761207761 • April 1 (1:44 PM): Approved by App Review. • Publication canceled to implement aesthetic/UI changes only. • April 2 (10:46 PM): Resubmitted with aesthetic updates. • April 3 (2:30 AM): Rejected — a device authentication feature was added without providing a test code to reviewers. • Feature was removed entirely. • April 4 (12:29 AM): Resubmitted. • April 8 (3:54 AM): Reviewed — Guideline 2.1 (Information Needed), reviewer asked two questions. • April 8 (4:21 AM): We responded to both questions via App Store Connect and tapped Update Review. • As of