Hmm it looks like the test failures are caused by this change. Could you take a look?

TheTooltipThemeData tests needed some adjustments. I also added an additional assert in the new tooltip test to prevent regressions in the text style merging logic.

There are Google test failures of two types:

1. Image diff tests where the text style has changed very slightly. I think we can approve these.
2. Text.style that was previously non-null but now is null. Any idea what caused that and what we can do about it?

Text.style that was previously non-null but now is null. Any idea what caused that and what we can do about it?

Before this PR, the tooltip'sText.rich widget received its text style explicitly via thestyle parameter. However, since we've introduced aDefaultTextStyle ancestor thatText.rich can inherit from, explicitly passing the same text style has become redundant.

There are two ways we can resolve these failures:

1. Modify the tests so that, instead of checking thestyle parameter directly on theText widget, we retrieve the firstDefaultTextStyle ancestor of theText and check itsstyle. This change should not cause any breakages in apps outside of tests because, let’s be real, does anyone actually read thestyle parameter of aText widget outside of tests?

   - final TextStyle textStyle = tester.widget<Text>(find.text(tooltipText)).style!;+ final Finder defaultTextStyle = find.ancestor(+   of: find.text(tooltipText),+   matching: find.byType(DefaultTextStyle),+ );+ final TextStyle textStyle = tester.widget<DefaultTextStyle>(defaultTextStyle.first).style;

2. We can pass the same text style to both theDefaultTextStyle widget and theText.rich widget. This won't change the behavior but will resolve the test failures.

Text.style that was previously non-null but now is null. Any idea what caused that and what we can do about it?

Before this PR, the tooltip'sText.rich widget received its text style explicitly via thestyle parameter. However, since we've introduced aDefaultTextStyle ancestor thatText.rich can inherit from, explicitly passing the same text style has become redundant.

There are two ways we can resolve these failures:

1. Modify the tests so that, instead of checking thestyle parameter directly on theText widget, we retrieve the firstDefaultTextStyle ancestor of theText and check itsstyle. This change should not cause any breakages in apps outside of tests because, let’s be real, does anyone actually read thestyle parameter of aText widget outside of tests?

   - final TextStyle textStyle = tester.widget<Text>(find.text(tooltipText)).style!;+ final Finder defaultTextStyle = find.ancestor(+   of: find.text(tooltipText),+   matching: find.byType(DefaultTextStyle),+ );+ final TextStyle textStyle = tester.widget<DefaultTextStyle>(defaultTextStyle.first).style;

2. We can pass the same text style to both theDefaultTextStyle widget and theText.rich widget. This won't change the behavior but will resolve the test failures.

This by definition is a breaking change (it affects registered tests). Arguably the tests could have been improved so it does not rely on implementation details but I would recommend avoiding the breaking change for now, by defaulting toDefaultTextStyle.

As for the failing golden image tests, it could be a flake or it could be a legitimate typeface change introduced by.merge(tooltipTheme.textStyle).merge(widget.textStyle). Let's see if it persists when you push another commit.

This by definition is a breaking change (it affects registered tests). Arguably the tests could have been improved so it does not rely on implementation details but I would recommend avoiding the breaking change for now, by defaulting to DefaultTextStyle.

I agree, done!

As for the failing golden image tests, it could be a flake or it could be a legitimate typeface change introduced by.merge(tooltipTheme.textStyle).merge(widget.textStyle). Let's see if it persists when you push another commit.

Unfortunately Google tests are still failing and I'm afraid what you're saying about the typefaces changing might be the case. Would that also be considered a breaking change?

Looks like the tests are passing, thank you for your patience 🙏