Codex will no longer be available in Xcode 26.3 starting on Friday, May 8th. Please upgrade to Xcode 26.4 or later to continue using Codex.

ENVIRONMENT Device : MacBook Pro 16,1 CPU : 6-Core Intel Core i7, 2.6 GHz RAM : 16 GB DDR4 GPU : AMD Radeon Pro 5300M + Intel UHD 630 macOS : 26.5 (Build: 25F71) Xcode : 26 (also tried Xcode 16 — same issue) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ISSUE SUMMARY Xcode crashes immediately on launch after updating to macOS 26.5. The issue started right after the macOS update — Xcode was working perfectly before. Both Xcode 26 and Xcode 16 are affected. Fresh reinstall does not help. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ERROR — xcodebuild -version output dyld: Library not loaded: @rpath/DVTSystemPrerequisites.framework/ Versions/A/DVTSystemPrerequisites Reason: code signature in DVTSystemPrerequisites.framework not valid for use in process — mapping process is a platform binary, but mapped file is not. zsh: abort xcodebuild -version ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CRASH REPORT DETAILS Incident ID : 370D1162-4CA5-4FFE-9193-45EBD350BF85 Exception : EXC_BAD_ACCESS (SIGKILL — Code Signature Invalid) Termination : CODESIGNING, Code 2, Invalid Page Crash Date : 2026-05-14 Failing Frameworks: DVTSystemPrerequisites.framework MobileDevice.framework (/Library/Apple/System/Library/ PrivateFrameworks/MobileDevice.framework) Crashed Thread 7 Stack: dyld3::MachOFile::trieWalk dyld4::Loader::hasExportedSymbol dyld4::JustInTimeLoader::applyFixups _CFBundleLoadExecutableAndReturnError -[NSBundle loadAndReturnError:] __watchForSims_block_invoke -[OSActivityStream setDeviceDelegate:] ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHAT I HAVE ALREADY TRIED sudo xattr -cr /Applications/Xcode.app sudo codesign --force --deep --sign - /Applications/Xcode.app sudo codesign --force --deep --preserve-metadata=identifier,entitlements --sign - /Applications/Xcode.app Re-signed DVTSystemPrerequisites.framework individually Deleted DerivedData, Xcode caches, preferences Fresh reinstall of Xcode (multiple times) Tried installing Xcode 16 — same crash Raised Apple Support Ticket: 102889658162 — no resolution received Nothing has worked. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ROOT CAUSE ANALYSIS macOS 26.5 introduced stricter platform binary validation in dyld. On Intel Mac (x86_64), this validation is rejecting Xcode's own frameworks at runtime — even after re-signing. This does not appear to be a user-fixable issue. This seems to be a macOS 26.5 regression specifically affecting Intel Mac users. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ REQUEST Is anyone else experiencing this on Intel Mac? Has anyone found a working workaround? Can Apple Engineering please investigate this as a macOS 26.5 regression for Intel Macs? My entire development workflow is blocked. Any help is greatly appreciated. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Hey, I am running Xcode 26.5 and I am not able to option + click anymore on a variable or a type that is not native (String, Bool etc works). The developer documentation preview is also broken.

We would like to use Xcode's new agentic coding feature (probably with Claude Agent). I had a look at the privacy statement, and under "Third-Party Agentic Coding Tools" it says the following: Capabilities include but are not limited to code searching, retrieving build logs, building source code, and arbitrary code execution. This is all expected, except for the bit about "arbitrary code execution". What does this mean, exactly? This sounds like a backdoor to my machine, surely that's not it? Does "code" refer to the code in my project, or any code the agent may provide? I know that other agentic coding tools may execute terminal commands, which you have to (or can select to) confirm first. Is that what is meant by arbitrary code execution? How arbitrary are we talking? And can I configure Xcode to always ask my permission first before any code is executed?

Hi, Overview I would like to add a Swift Package to my Xcode project. Problem In Xcode when I select Project > Package Dependencies and Press + button to add a dependency it keeps spinning and doesn't end Points to note: Xcode 26.5 (17F42) macOS 26.5 (25F71) This is a private GitHub repository I am able to clone the Swift Package URL on the command line. Xcode has the GitHub account I am able to push and pull code on that GitHub account using Xcode My failed attempt rm -rf ~/Library/Developer/Xcode/DerivedData rm -rf ~/Library/Caches/org.swift.swiftpm plutil -remove IDESwiftPackageAdditionAssistantRecentlyUsedPackages ~/Library/Preferences/com.apple.dt.Xcode.plist I have removed GitHub account in Xcode and added it again I am confused if I am missing something very obvious, any help on this is much appreciated. Questions Is there any workaround or way to fix this issue? Why does this take such a long time when I have a specific URL, either it needs to add it or return that the package is not found. Why is it searching?

I was having some trouble installing app to test on my Apple Watch (Series 9) so I removed it and thought to re-add it. However, after removing it, it no longer shows under my iPhone for me to add it as a possible run destination. I have tried numerous actions: Restart all devices Repair my watch Use hotspot so everything is on the same network Delete and reinstall Xcode None of these actions seem to be working. Does anyone have a way to get my watch to show up?

There are many situations where debug breaking points are not the right tool. What I often need is a log that shows what gets called, with how much frequency, and in which order. There are so many times where I have to add print statements, re compile, re run, see that I didn't include all the print statements I needed, and repeat a couple of times. My proposal (with zero knowledge of technical feasibility), is that much like we have break points, add "print break points" for logging without modifying code. It would be nice too, if I could also, add a "timer break point", where I can tag a function definition, and every time it gets called, it's are timed. Again, no idea if this is possible, but it would be extremely helpful for me, and I'm guessing to many other devs as well. Thoughts?

Hello Apple Developer Community, I’ve got a problem with the Xcode Build feature. Because when I would like to build / run my app, my iPhone won’t show up. I’ve already enabled Developer Mode and have set it up in Xcode and did the setup again. I don’t know what I can do next. Are there any possible solutions to this? Thanks for the help 😁. additional Information: Xcode Version: 26.5 Mac: MacBook Air M3 · macOS 26.5 iPhone: iPhone 16 · iOS 26.5

Hello, everyone! I've discovered a bug similar to libstdc++/115939 in GCC. This bug exists for a while. Did someone face it too? I'd be glad to accept any advise. Thank you in advance!

Hey everyone, this is a difficult topic to pin down on a certain component but I don't want to assign blame anyway. I want it to go away. Looks like people started to see this somewhere at Xcode 15 but it keeps happening. Xcode 26.5 as of this writing. For Apple people reading this, I posted a bug/feedback FB20206100 including a ready-to-compile (or rather fail-to-compile) example project. The gist is: If you open a Package.swift file with Xcode and that package contain a sub-package, Xcode will insert -suppress-warnings into the build of said sub-package. This means if you add "swiftSettings: [.unsafeFlags(["-warnings-as-errors"])]" into that sub-packages Package.swift file, it wont compile. Why? Because: error: Conflicting options '-warnings-as-errors' and '-suppress-warnings' When you look into the fairly recently accepted swift evolution propsal 0443 they basically say "That's fine. Nothing to see here." or as the proposal put it: Interaction with -suppress-warnings This proposal deliberately excludes -suppress-warnings and its group-based counterpart from the new unified model. We retain the behavior of the existing -suppress-warnings flag but forbid its usage with the new options. The following rules will be applied: It is forbidden to combine -suppress-warnings with -Wwarning or -Werror. The compiler will produce an error if these options are present in the command line together. It is allowed to be combined with -no-warnings-as-errors. The current compiler behavior permits the usage of -no-warnings-as-errors or -warnings-as-errors -no-warnings-as-errors with -suppress-warnings. We will maintain this behavior. I don't care who is right or wrong here but the bottom line is that if you use sub-packages you are effectively being errored out of using "-warnings-as-errors" for (IMHO) no good reason. Could Please someone look at this and say "Yeah, that doesn't look right, let's fix this". Please Thank You Roddi P.S.: Yes, I know there have been posts before that are lamenting this. But I thought, Let's try a fresh start here.

Hullo, I wanted to turn a class in a struct and move intelligence in a protocol extension, but when I try to call inits in the extension I get 'self' used before 'self.init' call or assignment to 'self' How do I call self.init or assign to self before updating the values? Of course init does no accept mutating.

Xcode 26.5 downloads darwin-arm64 Claude Agent binary on Intel Macs, causing exec failure misreported as code-signing error (workaround verified) On Intel Macs, Xcode 26.5 (build 17F42) downloads the darwin-arm64 build of the Claude Agent rather than the darwin-x64 build that the vendor publishes at the same version. The arm64 binary fails to exec on Intel with NSPOSIXErrorDomain Code=86 ("Bad CPU type in executable") and Xcode's Intelligence subsystem then surfaces this in the UI as a code-signing / sandboxing failure. Replacing the downloaded binary with the parallel darwin-x64 build, and updating the agent's Info.plist checksum and URL to match, results in a fully functional Claude Agent in Xcode 26.5 on the same Intel Mac. This was verified end to end on the affected hardware. The agent downloader in Xcode should select the URL matching the host CPU architecture (e.g. by inspecting the result of uname -m or the equivalent in Foundation). This appears to be an isolated fix to whichever component constructs the agent's download URL in Xcode's Intelligence subsystem. Secondary suggestion: the user-facing error message "the code signing identity for the agent did not match expectations, or the agent violated sandboxing rules" should be revised when the underlying cause is an exec failure (EBADARCH or otherwise). The current wording is misleading and has produced considerable misdirected investigation in the developer community. USER-SIDE WORKAROUND (VERIFIED) The following restores Claude Agent functionality on this Intel Mac running Xcode 26.5. It may need re-applying after Xcode updates the agent, since the buggy download logic will run again. Quit Xcode. Download the parallel darwin-x64 binary from the vendor's release bucket (the same URL as in the Info.plist, with "darwin-arm64" replaced by "darwin-x64"). Make the destination writable, copy in the new binary, restore read-only mode. Recompute SHA-512 and update Info.plist's checksum and url fields with plutil. Relaunch Xcode. Workaround for Intel Mac users (verified) This restores a working Claude Agent in Xcode 26.5 on Intel until Xcode'sdownloader is fixed. Quit Xcode fully first. 1. Quit Xcode osascript -e 'tell application "Xcode" to quit' sleep 3 2. Download the correct-arch binary directly from Anthropic's release bucket curl -fL -o /tmp/claude-x64\ttps://storage.googleapis.com/claude-code-dist-86c565f3-f756-42ad-8dfa-d59b1c096819/claude-code-releases/2.1.118/darwin-x64/claude chmod +x /tmp/claude-x64 3. Verify it's the right arch and properly signed file /tmp/claude-x64 # → Mach-O 64-bit executable x86_64 codesign --verify --verbose /tmp/claude-x64 4. Replace the arm64 binary Xcode downloaded (destination is read-only by default) AGENT_DIR=~/Library/Developer/Xcode/CodingAssistant/Agents/XcodeVersions/17F42/claude chmod u+w "$AGENT_DIR/claude" cp -p /tmp/claude-x64 "$AGENT_DIR/claude" chmod a-w "$AGENT_DIR/claude" 5. Update Info.plist so its checksum and URL match the swapped binary NEW_CHECKSUM=$(shasum -a 512 "$AGENT_DIR/claude" | awk '{print $1}') plutil -replace checksum -string "$NEW_CHECKSUM" "$AGENT_DIR/Info.plist" plutil -replace url -string "https://storage.googleapis.com/claude-code-dist-86c565f3-f756-42ad-8dfa-d59b1c096819/claude-code-releases/2.1.118/darwin-x64/claude" "$AGENT_DIR/Info.plist" 6. Relaunch Xcode; the Claude Agent now works on Intel. Caveats Adjust 17F42 and 2.1.118 to whatever your Xcode build and agent version actually are. Check with ls ~/Library/Developer/Xcode/CodingAssistant/Agents/XcodeVersions/ and the version key in the Info.plist. The workaround may not survive an Xcode update. If Xcode refreshes the agent (new Claude Code version, new Xcode point release), the buggy download logic will run again and overwrite the x86_64 binary. The same procedure restores working state. Worth saving the commands above as a small shell script so you don't have to re-derive them next time. ENVIRONMENT Hardware: MacBook Pro 2019, Intel Core i9 (x86_64) macOS: 26.5 Xcode: 26.5 (build 17F42) Agent: Claude Code 2.1.118 Account: Claude Pro (claude.ai OAuth) I have tried unsuccessfully to request via 'Apple Feedback Assistant' that the above problem be looked into. This was their response: "Thank you for your feedback, it is noted. Engineering has determined that there are currently no plans to address this issue. This feature relies on the dedicated Neural Engine found on Macs with Apple silicon (M1 or newer) to provide the necessary performance and capabilities. As a result, systems that do not include this hardware are not supported for this specific workflow. You can close this feedback by selecting "Close Feedback" via the Actions button found above. This Feedback will no longer be monitored, and incoming messages will not be reviewed. " Clearly no one from the senior engineering team has looked at this or given it any consideration and they are not interested in making a trivial fix within Xcode so that it selects the correct Claude binary for download. It is frustrating that Apple's response cites the requirement for a 'neural engine found in M-series Macs' , when this is very clear to any developer that Claude does not make use of this feature and it self-evidently works on x86 hardware without it . If anyone has any ideas how to get a response from a helpful senior Xcode developer, I would be most interested. I can understand Apple are reluctant to put any effort into old hardware, however they still provide an up to date Xcode for the x86 platform and I don't think it is asking too much to fix a simple bug which would make life easier.

FB22844728 Xcode 26.5, macOS 26.5, MacStudio M1 64 GB The disaster of S-L-O-W T-Y-P-I-N-G that plagued Xcode several years ago returns. I guess all the ultra-advanced LLM neuroneirowebs still can't advise those at Apple who develop the Xcode app how to provide this uber-exotic, unheard-of functionality: JUST TYPING THE CODE. This disaster doesn't affect all projects, but it does affect the exact one I'm working on, and I need to type a lot of code there. Significant time is already wasted. I performed all the recommended steps like relaunching, restarting, and deleting deprived data with no effect. Hope something can be done about it..

Swift Package Manager fails while downloading Firebase SDK. Error: Authentication failed because no credentials were provided. Package: https://github.com/firebase/firebase-ios-sdk.git Already tried: Reset package cache Removed DerivedData Re-added GitHub account Reinstalled Xcode

I'm trying to publish two frameworks I wrote in Swift on GitHub. https://github.com/nickasd/BoardGameKit https://github.com/nickasd/BoardGameKitHost I want to add the first framework as a dependency to one of my Xcode projects. I click on the + button in the Package Dependencies tab, select BoardGameKit and click Add Package. A sheet opens and after a couple seconds, while the sheet is still open, Xcode crashes. When I try the same steps with the second framework, Xcode survives and finally shows me a confirmation sheet, and clicking on Add Package works as expected. Both repositories look exactly the same to me, except for their name of course. I even force pushed the default, empty Swift package created in Xcode with the Framework template to both repositories, and still only the first one makes Xcode crash, even after deleting all DerivedData folders and restarting Xcode. What could be the issue? I created FB22832333.

我发现在 M1 芯片的 MacMini上使用 Xcode26.3打包出来的 App 在 iphone6设备 ios 12.5.8 的系统下 安装或者直接调试都会直接 启动后 立即 崩溃，但是在 intel 芯片的 MacBook pro 上使用 Xcode 26.3 在相同的设备相同的系统下，就可以正常安装和调试，请问这是为什么？

I've found several references for this functionality being removed, but no viable workaround. When I'm trying to fine tune a layout the constraint lines, boundary boxes, etc. get in the way and I can't see what I'm doing. For doing pixel-perfect layouts in a storyboard it gets really tedious. This borders on an accessibility issue. Does anyone know when or if this is ever coming? It works fine with iOS apps.

Hello, I can understand why you do not ship the MetalToolchain with the default Xcode installation any more due to the relatively low usage and high download size. That said, every time Xcode runs an auto update it wipes MetalToolchain and breaks my local development build. It would be nice if the updates would be smart enough to honor the fact that. I have already run: "xcodebuild -downloadComponent MetalToolchain" and include that in the update, rather than deleting the module. Thanks, Chris

In a recent Xcode update (maybe 26.5, maybe all 26 versions), Xcode started inserting the closing quote, parenthesis, or bracket. It happens like this. I type, say, an opening double quote ("). Then as soon as I type another character, it adds another double quote just ahead of the insertion point. Now, I have everything in the Settings->Editing->Completion section turned off. I also have virtually everything in the Indentation section turned off. Is this just a bug? I've reported it (FB22693037). But I'm surprised it wasn't caught by internal testing. I can't express how much I hate this behavior.

The process of copying shared cache symbols from my Apple Watch has been horrendous. I cannot even get past 2% complete before this process fails. Every device is on the same network. I have trusted my machine. Developer mode is enabled on the watch. I have successfully paired the phone and watch. I managed to install an Watch app onto the watch. But the shared cache symbols were never copied from the watch. I gathered several commands at the terminal to help me. rm -rf ~/Library/Developer/Xcode/DerivedData rm -rf ~/Library/Developer/CoreDevice sudo killall -9 remoted CoreDeviceService 2>/dev/null || true xcrun devicectl manage ddis update xcrun devicectl list devices Clear CoreDevice state. With Xcode quit: rm -rf ~/Library/Developer/Xcode/DerivedData rm -rf ~/Library/Caches/com.apple.dt.Xcode Reset Xcode's Local Network permission. tccutil reset Local-Network com.apple.dt.Xcode tccutil reset Bluetooth com.apple.dt.Xcode xcrun devicectl device info hardware --device MY_DEVICE_ID xcrun devicectl list devices ping -c 3 MY_IP_ADDRESS Confirm no VPN/proxy/filter is lurking scutil --nwi ifconfig | grep -E "utun|ipsec|tap|tun" xcrun devicectl device info details --device MY_DEVICE_ID xcrun devicectl manage pair --device MY_DEVICE_ID I keep obtain this type of error: Unable to copy shared cache files Domain: com.apple.dt.deviceprep Code: -33 User Info: { DVTErrorCreationDateKey = "2026-05-20 23:41:00 +0000"; NSLocalizedRecoverySuggestion = ""; "com.apple.dt.DVTCoreDevice.operationName" = dtfetchsymbols; } -- The operation couldn’t be completed. No route to host Domain: NSPOSIXErrorDomain Code: 65 Failure Reason: No route to host -- System Information macOS Version 26.5 (Build 25F71) Xcode 26.5 (24943) (Build 17F42) Timestamp: 2026-05-20T16:41:00-07:00 Does anyone have ideas to help me overcome this nightmare?

Hi! Here is a brief introduction at first. I am developing a framework and our project has following structure: All frameworks are static ones. PublicAPI framework contains the only declarations that are visible to users, while ModuleX frameworks are internal modules. During build process we archive all targets and libtool all internal module binaries into PublicAPI one. This way we distribute our framework as a single xcframework package and hide all internal modules from users. Here is the problem part. All these frameworks exist in separate projects that have single static library target. I want to add new targets with different settings, say SWIFT_ACTIVE_COMPILATION_CONDITIONS or else. So project would look like this Also I would like both regular and extended targets to build frameworks with same module name, so that neither users (that use PublicAPI framework) nor I would have to change imports in my code. My idea was to set PRODUCT_MODULE_NAME in extended targets to the same value as in regular ones. But when I build PublicAPIExtended target I get an error Unable to find module dependency 'Module1' although derived data contains Module1Extended.framework with Module1 module inside. What did I do wrong? Is it possible to build project like this at all? And another thing - I also tried to build a scheme containing both PublicAPI and PublicAPIExtended targets and got different error even prior to build: error: Multiple commands produce '<ExtractAppIntentsMetadata Path(_str: "/Users/antonprotko/Library/Developer/Xcode/DerivedData/test-acncglphufkosyhjzbwrbzalsnlh/Build/Products/Debug") bar/Metadata.appintents>' note: Target 'bar copy' (project 'bar'): ExtractAppIntentsMetadata note: Target 'bar' (project 'bar'): ExtractAppIntentsMetadata Can I fix it somehow? Here is example project with similar targets - foo/foo_copy and bar/bar_copy share the same sources with different compilation conditions. bar targets depend on foo targets and building bar_copy fails with Unable to find module dependency: 'foo' https://drive.google.com/file/d/1Cj5cIw8U0LJ5Xt9ZnhbKIo240kbrizLE/view?usp=sharing