Deprecated problems on parameters

[FMorschel]

We now have two problems here. Passing values to deprecated parameters doesn't show any diagnostics, and there is no indication on completion either.

Repro:

void foo({@Deprecated('Not used') int? parameter}) {
  foo(parameter: 0);
  foo(parameter: 0);
}

I'd say this is probably worth the CP once fixed, since it may lead to people using deprecated parameters.

CC @bwilkerson @srawlins @lrhn

[FMorschel]

This was found while working on https://dart-review.googlesource.com/c/sdk/+/463840.

But some cases do show the diagnostic here, and I'm still trying to understand why:

FYI @mraleph and @scheglov, maybe related to #62012?

[FMorschel]

Noting that when fixing this, we need to be sure that the parameter doesn't trigger the diagnostic when referenced inside the function itself, as it was.

[FMorschel]

Also:

class Class {
  @Deprecated('Not used')
  int? parameter;
  Class({this.parameter});
}

void bar(Class c) {
  if (c case Class(:final parameter)) {}
  if (c case Class(parameter: final parameter)) {} // <- missing diagnostic too
}

Completion does work:

But how should the diagnostic happen on the :final parameter (1) case (@bwilkerson)?

[scheglov]

No, I don't think it is related to #62012, most probably this is because of a way the analyzer models invocations and types, where we sometimes jump to FunctionType and lose annotations on formal parameters. I don't think this is new issue.

[bwilkerson]

But how should the diagnostic happen on the :final parameter (1) case?

It isn't at all clear to me that is should. This is defining a pattern variable and the variable isn't deprecated. What's deprecated is the getter (and it seems reasonable to show that in the parameter: final parameter case, which it doesn't).

It also seems kind of reasonable to show that in the first case, but the problem is that the parameter there is kind of doing double duty. I don't think we want to imply that the variable is deprecated any more than we want to ignore that the getter is deprecated. I don't see a good solution. (I'll add it to the design doc I recently added.)

I noticed something else odd, so I opened #62080.