[Declaring Constructors] Evaluate support for prefer_final_parameters

[bwilkerson]

If the final modifier on a primary constructor's parameter indicates that the field should be final, then this lint will need to either be updated to not report on primary constructors, or it will need to be deprecated.

This issue is to collect feedback for which of those approaches we should take.

[eernstg]

I think the lint will have to be removed entirely (or disabled for language versions that support primary constructors). The reason is that final cannot be used to indicate that a parameter (itself) is immutable, it will only indicate that (1) this parameter is declaring, and (2) the implicitly induced instance variable for this parameter is final.

In every function that isn't a declaring constructor it is a compile-time error to have a declaring parameter, which means that final is an error on any parameter of, say, a top-level function, an instance or static method, a setter of any kind, a function literal, etc.

[bwilkerson]

... which means that final is an error on any parameter ...

Oh, I wasn't aware of that change. (The kick-off meeting is next week.)

Do we need to introduce a warning in the release before primary constructors is enabled to let users know that final won't be allowed in any parameter where it's currently being used? If so, we need to get that in quickly unless we want to delay the introduction of primary constructors.

[eernstg]

I think that would be very helpful! I'd also hope that there will be a bulk fix approach (dart --fix . or something like that), but there is no decision to invest time in doing that, so I can only hope it will be possible.

[bwilkerson]

@kallentu

[kallentu]

We won't be releasing primary constructors in version 3.11 (we're aiming for ~Spring 2026), so we have some time to make this change.
We don't need to rush to get this in the current version (3.10).