We're suddenly getting very persistent flake in the integration tests for path_provider on macOS:
00:00 +0: getTemporaryDirectory
══╡ EXCEPTION CAUGHT BY FLUTTER TEST FRAMEWORK ╞════════════════════════════════════════════════════
The following PathNotFoundException was thrown running a test:
Cannot open file, path =
'/Users/chrome-bot/Library/Caches/dev.flutter.plugins.pathProviderExample/temporaryDirectory' (OS
Error: No such file or directory, errno = 2)
When the exception was thrown, this was the stack:
#0 _File.throwIfError (dart:io/file_impl.dart:779:7)
#1 _File.openSync (dart:io/file_impl.dart:572:5)
#2 _File.writeAsBytesSync (dart:io/file_impl.dart:744:31)
#3 _File.writeAsStringSync (dart:io/file_impl.dart:772:5)
#4 _verifySampleFile (file:///Volumes/Work/s/w/ir/x/w/packages/packages/path_provider/path_provider/example/integration_test/path_provider_test.dart:124:8)
#5 main.<anonymous closure> (file:///Volumes/Work/s/w/ir/x/w/packages/packages/path_provider/path_provider/example/integration_test/path_provider_test.dart:15:5)
<asynchronous suspension>
#6 testWidgets.<anonymous closure>.<anonymous closure> (package:flutter_test/src/widget_tester.dart:192:15)
<asynchronous suspension>
#7 TestWidgetsFlutterBinding._runTestBody (package:flutter_test/src/binding.dart:1953:5)
<asynchronous suspension>
<asynchronous suspension>
(elided one frame from package:stack_trace)
The test description was:
getTemporaryDirectory
════════════════════════════════════════════════════════════════════════════════════════════════════
00:00 +0 -1: getTemporaryDirectory [E]
Test failed. See exception logs above.
The test description was: getTemporaryDirectory
I don't think anything changed on our side, so it may be a change on the bots (e.g., a macOS update with different behavior). It looks like this may be a latent issue in the test, as it seems to be assuming the directory will already exist even though we don't guarantee that. What's odd though is that it looks like path_provider_darwin has the same structure still, and isn't failing.
Still, we should make the test create the directory if necessary, and see if that resolves it.
We're suddenly getting very persistent flake in the integration tests for path_provider on macOS:
I don't think anything changed on our side, so it may be a change on the bots (e.g., a macOS update with different behavior). It looks like this may be a latent issue in the test, as it seems to be assuming the directory will already exist even though we don't guarantee that. What's odd though is that it looks like
path_provider_darwinhas the same structure still, and isn't failing.Still, we should make the test create the directory if necessary, and see if that resolves it.