Repository navigation
chore(deps): .NET SDK 10.0.401, CI actions and NuGet refresh - #237
Conversation
Pin the SDK to 10.0.401 with an explicit workloadVersion 10.0.401.1 so the workload set no longer drifts independently of the SDK. Every workflow now reads the same root global.json through setup-dotnet's global-json-file input instead of repeating a 10.0.3xx channel. SalmonEgg/global.json only duplicated the Uno.Sdk pin; MSBuild picks the nearest file, so the two copies could silently split the GUI project and src/tests onto different SDKs. It is removed and all directories share the root file. The stable Apple workload for 10.0.401 requires Xcode 27, so the iOS gates move to the xcode-27 runner (public preview) and the Windows jobs to windows-2025-vs2026 for Visual Studio 18.9.3+. Actions are bumped to checkout v7.0.1, setup-dotnet v6.0.0, upload-artifact v7.0.2, download-artifact v8.0.2 and codeql-action v4.38.2, all pinned by SHA. The published bridge gate moves to acpremote 1.10.0. Docs (README zh/en, BUILD_GUIDE, release-guide) and the workflow guard tests follow the new pins. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- Microsoft.Graph 6.7.0 -> 6.8.0 - Microsoft.Identity.Client(.Extensions.Msal) 4.90.0 -> 4.90.1 - Microsoft.WindowsAppSDK 1.8.260804001 -> 1.8.260921001; WinUI stays at 1.8.260803003, the exact version this meta-package pins in its nuspec - Porta.Pty 2.2.2 -> 2.2.3, Websocket.Client 5.5.0 -> 5.5.1 - coverlet.MTP 10.0.1 -> 10.1.0, Moq 4.20.72 -> 4.21.0 Windows App SDK 2.x is intentionally deferred until Uno 6.7 and MSIX packaging are validated on it. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3bc66e9ce1
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| run: | | ||
| python3 -m venv "$RUNNER_TEMP/acp-bridge-venv" | ||
| "$RUNNER_TEMP/acp-bridge-venv/bin/pip" install acpremote==1.9.0 | ||
| "$RUNNER_TEMP/acp-bridge-venv/bin/pip" install acpremote==1.10.0 |
There was a problem hiding this comment.
acpremote==1.10.0 is not published on PyPI—the project's release history currently ends at 1.9.0—so this pip install fails before the published-bridge cancellation gate can run on every CI invocation. Pin this back to an existing release (or publish/configure the intended index) so the required transport validation remains executable. PyPI release history
Useful? React with 👍 / 👎.
sdk.workloadVersion makes the MSBuild workload resolver require that exact workload set, so every project fails with MSB4242 on a fresh runner that has not installed it, including jobs that need no workload at all. It also rejects the --skip-manifest-update flag our workload install steps use. Workload manifests follow the SDK feature band again, as on main; BUILD_GUIDE records why the field is intentionally absent. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Uno.Sdk 6.7.30 still defaults SupportedOSPlatformVersion to 14.2 for iOS, but the Xcode 27 Apple workload (Microsoft.iOS.Sdk 27.0) that ships with the 10.0.401 SDK band fails the build below 15.0. iOS 15 runs on exactly the devices iOS 14 did, so no supported hardware is dropped. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 63b291af17
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| # https://github.com/dotnet/macios/releases/tag/dotnet-10.0.1xx-xcode26.5-10284 | ||
| dotnet workload install ios --version 10.0.300.3 --disable-parallel --no-http-cache | ||
| # https://github.com/dotnet/macios/releases/tag/dotnet-10.0.1xx-xcode27.0-10722 | ||
| dotnet workload install ios --disable-parallel --no-http-cache |
There was a problem hiding this comment.
When this acceptance job runs after a newer iOS manifest is advertised, this bare install downloads that newer manifest by default; dotnet workload install only avoids manifest downloads with --skip-manifest-update. global.json pins only the SDK, while this workflow fixes Xcode at 27.0, so a later workload can become incompatible and make the installed-iOS gate fail independently of the PR. Add --skip-manifest-update as in the platform iOS gate (or explicitly install the matching workload set) to keep the SDK, workload, and Xcode as one toolchain contract.
AGENTS.md reference: AGENTS.md:L104-L104
Useful? React with 👍 / 👎.
The xcode-27 runner image does not register Simulator with LaunchServices, so 'open -a Simulator' fails before the iOS acceptance run can install the app. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Xcode 27 removed Developer/Applications/Simulator.app and ships Contents/Applications/DeviceHub.app instead, so the path lookup from the previous commit still found nothing. Accept either layout and list the apps actually present when neither exists. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Xcode 27 host is slower than macos-26, where installing the Debug interpreter bundle already took about 130 seconds, so the default 180-second command timeout expired mid-install. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
On the Xcode 27 / iOS 27 simulator the product logs its startup marker and then nothing else, while the same flow on iOS 26.5 logs the next startup step within two seconds. The app log cannot say which thread or frame is stuck. When the gate fails after install, take a host stack sample of the product process, the device's unified log for it, and any crash report before cleanup terminates it, and upload them. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The 10.0.401 feature band ships two workload sets. 10.0.401.0 carries the Apple workload 26.5.10318, which requires Xcode 26.6; 10.0.401.1 moves to 27.0.10722, which requires Xcode 27. Installing without a version picked the newest set and forced both iOS jobs onto the xcode-27 public-preview image, where the iOS 27 simulator took over five minutes to install the app and then wedged simctl and the product's own startup. Pin --version 10.0.401 on the two iOS workload installs, the same per-job pin main used with 10.0.300.3, and return to the GA macos-26 image with its Xcode 26.6. The guard test now requires the pin. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
That floor was only required by the Xcode 27 Apple workload (27.0), which the iOS jobs no longer install; the pinned 26.5 workload accepts Uno.Sdk's 14.2 default, as on main. This reverts commit 63b291a. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
On macos-26 with Xcode 26.6 the installed-product test passed, but xcodebuild test-without-building took 458 s end to end (main on Xcode 26.5: 241 s), overrunning the 420 s wait. Raise it to 600 s and give the reverse verification matching headroom. Cleanup then crashed with EPERM from killpg: macOS reports that, not ESRCH, when the process group holds only zombies. Treat it as gone. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
This branch was successfully deployed
1 participant
Summary
global.jsonviasetup-dotnetglobal-json-file. Removed the duplicateSalmonEgg/global.jsonso every directory shares one SDK/Uno.Sdk pin.xcode-27(required by the 10.0.401 Apple workload; image is public preview), Windows →windows-2025-vs2026.Windows App SDK 2.x is intentionally deferred: it needs Uno 6.7 + MSIX validation on Windows.
Tested (local Linux arm64, SDK 10.0.401)
Full test suite: 6315 passed / 0 failed / 7 skipped
Windows CI-form MSIX restore: resolves WindowsAppSDK 1.8.260921001 + WinUI 1.8.260803003
WASM smoke gates (Debug, as in CI): all passed
coverlet.MTP 10.1.0 emits cobertura reports
run-acp-published-bridge-gate.pywith acpremote 1.10.0: passedClean SDK 10.0.401 install (no workloads): builds succeed;
workloadVersionin global.json reproduces the MSB4242 seen in the first CI run, so it is not pinnedNot verified locally
🤖 Generated with Claude Code