I don't like that predicate is getting called twice on each route, but that can be fixed. What I am not sure about is whether it is safe to assume that the predicate result won't change between checking it as a next route vs as a current route?

@chunhtai I have made the changes, PTAL.
It still has a problem of testing the route as next, but I am not sure if there even is a different way in this case.
This way does give more control to the developer, but I feel like the majority of cases (judging from the issue) is still going to be

Navigator.popUntil(
   context,
  (Route<dynamic> route) => route.isFirst,
  (Route<dynamic> route) => route.isFirst ? true : null,
);

to just return a value to the first route.

I think the current approach should be fine, if they really want a special case, they can write a helper function. I want to avoid having conflict that people want a slightly different variant of your use case and we would need to end up adding a different popUntil API.

I think you are right. I am worried that there can be a predicate that when tested as the next route will return a different result to when it was tested as the current route. Do you think this can be an issue?

@chunhtai I have reverted the changes, please take a look.

Regarding your last question:

I believe that any predicate that for some reason checks route.isCurrent (since it will now be false for all but the first checked route) or any other predicate that only becomes true when this route is the top one will break because of these changes.

Maybe I am overthinking this. Do you think this is an issue?

I think I found an example of such a predicate:

To make the test pass, I had to leave this check:

This works, but it now has an issue with checking each route twice. Once as the next route, and once as the current route. This happens because checking willHandlePopInternally before the pop returns false, so we pop, but then if we don't check the predicate the second time, we will pop an extra time.

I see now, I will take another look