flutter
Some tests fail locally on macOS
While trying to fix up test failures on another branch, I noticed that some of them failed even without my changes and when running on Flutter dev:
I don't know if these only occur locally, only for me, or generally on macOS (macOS bots are currently disabled), but it'd be good to get them working to simplify running tests locally (and knowing whether you've broken anything).
1. inspector_controller_test
This fails because an isolate event arrives and is sent to the message bus, but the message bus is null.
Q: Should the test set up a message bus, or should _sendToMessageBus handle a null bus?
2. dart_cli_profile_test
Exception: There are no registered methods for service "_flutter.listViews
This is not running on CI because it's in an integration folder, but the rest are in integration_tests.
Q: Should this be in integration_tests?
Q: Why isn't the service registered? 🤷♂️
3. memory_screen_test
Exception: Getting current working directory failed, path = '' (OS Error: No such file or directory, errno = 2)
Unclear why this _fs.currentDirectory is throwing this error for me, but a try/catch around the code and falling back to the temp directory solves it.
Q: Is it reasonable to commit the try/catch?
4. timeline_flame_chart_test
Golden failure (no obvious difference to my eyes - perhaps slight differences between platforms and these were generated on Linux?).
Q: Is this expected to pass on macOS? If not, should we skip when platform != Linux?
5. treemap_test
Expected: exactly one matching node in the widget tree Actual: _TextFinder:<zero widgets with text "new CastIterable. [0.2 KB]" (ignoring offstage widgets)> Which: means none were found but one was expected
This fails on:
expect(find.text('new CastIterable. [0.2 KB]'), findsOneWidget);
Though if I run the test on a device and add a breakpoint, I can see that text in the last item of the breadcrumb, so it's not clear why this isn't found.
From a discussion with Danny, some of these failures might be race conditions, and some are tests that we're not running on the bots (but probably should).
