You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Engine -> Framework roll stopped on framework text field test failures #122950
Flagging folks with engine commits in that range that seem text related: @flar@LongCatIsLooong@loic-sharma and @jason-simmons to take a look at the Skia rolls to see if there's anything suspicious in them.
Some of the failures are due to properties on Widgets which would not be a result of anything in Skia or any rendering changes. I'd strongly suspect the changes that affected the way accessibility is handled (single root assumption perhaps?)
These tests depend on execution order due to the interaction between debugDefaultTargetPlatformOverride and the Theme._kFallbackTheme singleton.
_kFallbackTheme is a lazily initialized static ThemeData. The fallback theme does not explicitly specify a platform, so the ThemeData constructor sets its platform based on the current value of defaultTargetPlatform.
Some of the tests in material/text_field_test.dart use TargetPlatformVariant to temporarily override the target platform. So the platform of the fallback theme will depend on which override was in effect the first time _kFallbackTheme was accessed and initialized.
Later tests may depend on platform-specific behaviors (for example, tests of haptics assume they are running on a platform that enables haptics). These tests will fail if they use the fallback theme and it is bound to an unexpected platform.
Is there a way for the test infrastructure to ensure that _kFallbackTheme is initialized for the Android platform, which is specified as the default for the test environment?
Or do tests need to avoid instantiating anything that would depend on the fallback theme?
This thread has been automatically locked since there has not been any recent activity after it was closed. If you are still experiencing a similar issue, please open a new bug, including the output of flutter doctor -v and a minimal reproduction of the issue.
Activity
flar commentedon Mar 18, 2023
Some of the failures are due to properties on Widgets which would not be a result of anything in Skia or any rendering changes. I'd strongly suspect the changes that affected the way accessibility is handled (single root assumption perhaps?)
jason-simmons commentedon Mar 18, 2023
This failure was caused by ordering dependencies in the
material/text_field_test.darttests and does not appear to be related to the engine roll.The tests started failing when the daily test order changed to
--test-randomize-ordering-seed=20230318.The tests run successfully with the #122942 roll and yesterday's seed value (
--test-randomize-ordering-seed=20230317)jason-simmons commentedon Mar 18, 2023
These tests depend on execution order due to the interaction between
debugDefaultTargetPlatformOverrideand theTheme._kFallbackThemesingleton._kFallbackThemeis a lazily initialized staticThemeData. The fallback theme does not explicitly specify a platform, so theThemeDataconstructor sets itsplatformbased on the current value ofdefaultTargetPlatform.Some of the tests in
material/text_field_test.dartuseTargetPlatformVariantto temporarily override the target platform. So the platform of the fallback theme will depend on which override was in effect the first time_kFallbackThemewas accessed and initialized.Later tests may depend on platform-specific behaviors (for example, tests of haptics assume they are running on a platform that enables haptics). These tests will fail if they use the fallback theme and it is bound to an unexpected platform.
@Hixie @goderbauer What is the best way to fix this?
Is there a way for the test infrastructure to ensure that
_kFallbackThemeis initialized for the Android platform, which is specified as the default for the test environment?Or do tests need to avoid instantiating anything that would depend on the fallback theme?
zanderso commentedon Mar 18, 2023
#122956 <- disables shuffling for this file.
goderbauer commentedon Mar 20, 2023
FYI @pdblasi-google The no-shuffle was recently removed in #122643 from this test.
goderbauer commentedon Mar 20, 2023
I am going to close this as a duplicate of #85160 where we track all tests that need their no-shuffle removed.
github-actions commentedon Apr 3, 2023
This thread has been automatically locked since there has not been any recent activity after it was closed. If you are still experiencing a similar issue, please open a new bug, including the output of
flutter doctor -vand a minimal reproduction of the issue.