TypeError.toString() message has regressed for simple types

[jakemac53]

Consider the following test, which would have passed in older SDKs.

import 'package:test/test.dart';

void main() {
  test('TypeError "null as String" text does not mention type parameters', () {
    try {
      var x = null as String;
      print(x);
    } on TypeError catch (e) {
      expect(e.toString(),
          "type 'Null' is not a subtype of type 'String' in type cast'');
    }
  });
}

This now fails as follows:

Expected: 'type \'Null\' is not a subtype of type \'String\' in type cast\''
Actual: 'type \'Null\' is not a subtype of type \'String\' of \' in type cast\''

It looks like it is maybe trying to list some information about generic type parameters now, but String has no generic type parameters.

cc @lrhn

[jakemac53]

Note that in general I am not opposed to this message changing, we do have a golden test which will get broken, but I don't consider the exact message to be a part of the API. But I do think the new message is a regression from the old.

[lrhn]

This is probably a bug in the removal of CastError.
The message in a type cast, leading space and all, seems like it gets put into a place where something else is expected. Maybe the name of a variable being assigned to.
It's badly formatted and reads worse, so it's definitely not what we want to end up with.

Let's see if we can make sure it looks like before.

[kevmoo]

void main() {
  print(null as int);
}

Also the grammar is just...off...

Unhandled exception:
type 'Null' is not a subtype of type 'int' of ' in type cast'
#0      main (file:///Users/kevmoo/github/json_serializable.dart/checked_yaml/example/boo.dart:2:14)
#1      _delayEntrypointInvocation.<anonymous closure> (dart:isolate-patch/isolate_patch.dart:297:19)
#2      _RawReceivePort._handleMessage (dart:isolate-patch/isolate_patch.dart:192:26)

the of ' bit just seems...like a bad string template being applied!

[lrhn]

The call to create this message seems to happen somewhere in VM C++ code, so tagging as area-vm.

[a-siva]

#50756 should fix that message. I believe this regressed when CastError was deprecated and replaced by 'TypeError'