fix(tests): stop snapshot tests failing on newer macOS, and hitting the Keychain - #31
Open
mpecan wants to merge 2 commits into
Open
fix(tests): stop snapshot tests failing on newer macOS, and hitting the Keychain#31mpecan wants to merge 2 commits into
mpecan wants to merge 2 commits into
Conversation
…he Keychain Two unrelated defects, both visible when running the suite locally. Snapshot tests compared images pixel-exact. SwiftUI renders gradients and antialiased edges differently across macOS versions, so all 8 failed on any OS other than the one that recorded the references — CI passes, a modern machine does not. Compare perceptually instead. The tolerance is measured, not guessed: the same icon drifts by at most ~10 deltaE across OS versions, while a different icon differs by ~128, so allowing deltaE <= 10 keeps a 12x margin. Reference images are unchanged, so CI keeps comparing against exactly what it did before. A guard test asserts a changed status colour is still rejected, so the tolerance cannot be loosened into asserting nothing. The test bundle is hosted by the app, so running tests launched the real menu bar: it read the session key from the Keychain (prompting the user, repeatedly, since the test binary is signed differently) and fetched live usage over the network. Skip bootstrap when running under XCTest. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The previous commit skipped menu bar bootstrap by having AppDelegate check for XCTestConfigurationFilePath — production code that knows it is under test. The app already has the hook needed: `--demo <mode>` routes to startWithoutBootstrap(), which sets up the menu bar without reading the Keychain or fetching usage. The scheme's test action now passes it, so AppDelegate is untouched. The cost is that the behaviour lives in the scheme, where it can be deleted without anyone noticing. TestHostLaunchTests asserts the test host was launched with a valid --demo argument, so removing it fails the suite instead of silently restoring the Keychain prompts. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jul 13, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Two defects visible when running the suite locally. Neither is caught by CI. No production files change.
Snapshot tests fail on newer macOS
All 8
MenuBarIconSnapshotTestscompared images pixel-exact. SwiftUI renders gradients and antialiased edges differently across macOS versions, so they fail on any OS other than the one that recorded the references. They pass on CI's runner and fail on macOS 26 / Xcode 26.Measured:
Now compared perceptually with
precision: 0.99, perceptualPrecision: 0.90(ΔE ≤ 10). That absorbs the rendering drift with a 12x margin before a different icon would pass. Reference images are unchanged, so CI compares against exactly what it did before.test_snapshotTolerance_rejectsChangedStatusColourasserts a changed status colour still fails, so the tolerance cannot be loosened into asserting nothing.Ruled out: system Dark/Light mode (renders are pixel-identical in both) and
ImageRenderer.colorMode(no mode reproduces the reference).Tests read the Keychain and hit the network
The test bundle is hosted by the app, so running tests launches it.
MenuBarManager.start()callsappModel.bootstrap(), which reads the session key from the Keychain and fetches live usage. The test binary is signed differently from the installed app, so macOS prompts for Keychain access on every run.The scheme's test action now passes
--demo safeUsage, routing the app to the existingstartWithoutBootstrap(), which sets up the menu bar without callingbootstrap().AppDelegateis untouched.That behaviour lives in the scheme, where it can be deleted unnoticed.
TestHostLaunchTestsasserts the test host was launched with a valid--demoargument and fails with an explanatory message if not.Verified by deleting
~/.claudemeter/usage.json, which the app writes on every successful fetch, and confirming it does not reappear across a full run.Result
61 tests pass. Previously 8 failed.
Not covered
--demo safeUsagealso injects demo usage data into the launched host app. Tests build their own objects and never read the host'sAppModel. A dedicated flag, or moving app logic into a framework so the test bundle needs no host, would both require production changes.🤖 Generated with Claude Code