Conditionally Mapped Type Leads to Confusing ts(2345) Error Message

從 issue 中的重現以及第 227 行附近所引用的 tests/cases/conformance/types/conditional/conditionalTypes2.ts 開始；比較列出的 TypeScript 版本和 Playground 連結中的診斷。完成的標準是，callB 錯誤不再呈現具有誤導性的 conditional mapped type，同時保留預期的 ts(2345) 診斷。

TypeScript Version:
3.9.6
4.1.0-dev.20200914

Search Terms:
2345
conditional

Code

declare type GetPropertyNamesOfType<T, RestrictToType> = {
	[K in keyof T]: T[K] extends RestrictToType ? K : never;
}[keyof T];

You can find this gnarly type here:
https://github.com/microsoft/TypeScript/blob/94d4023043dd032a8c96a59677eb63526b6b24b6/tests/cases/conformance/types/conditional/conditionalTypes2.ts#L227

The following consuming/using code correctly generates ts(2345) errors for the last two lines:

interface A {
	req: string;
	num: number;
	str: string;
}
interface B {
	req: string;
	num?: number;
	str?: string;
}

declare function callA(name: GetPropertyNamesOfType<A, string>): void;
declare function callB(name: GetPropertyNamesOfType<Required<B>, string>): void;

const prop = "someKey";
callA(prop); // ts(2345)
callB(prop); // ts(2345)

Expected behavior:
In typescript@3.9.5 & typescript@3.8 (and a few earlier versions I tried) the error message for both lines is:

Argument of type '"someKey"' is not assignable to parameter of type '"req" | "str"'.

Actual behavior:
In typescript@^3.9.6 this is reported for both lines:

Argument of type '"someKey"' is not assignable to parameter of type 'GetPropertyNamesOfType<A, string>'.

That's correct (nice even!) for the callA() line, but it's confusing for the callB() line.

(This reproduction is a single file that can run into this problem... it's not how I ran into it. I had a ts.Program with many 10s of *.ts files in the same compilation and the message was referring to some type that made no sense in the source file that was giving me the error.)

Playground Link:
https://www.typescriptlang.org/play?#code/CYUwxgNghgTiAEAXAngBwQcRIgCjA9ujCgHJQC2IAzgPIBmAKmiADwMA08AStYjAJZhEDfE3QA+eAF54AbwCwAKACQAbQDS8fgDt4AaxDJ8deAwC6ALlMaz8EAA9EIbcCrdeAoSLEIA-PE0rbRAANxAYAG4lAF9VAyMTcyjFJR0nGDooMAQAQTklZTgARysqPh0Ac2TlbQBXciD6gCNw6rKYUvLtKpjU7XTM7PgAIXyVYs6Bbuq68l9G8hbIgvb5+HbK5OilJVBIWAQ6Wu0hfnxdMCgICByACm0KECssXAIiUkfaRmYWHM4N7riACUVhC+H4wGSe2gcHgRxOiDOFyuEGG90ez2weEI4Q+lC+PhYPCKtX4cGALGG4n+XQqwNB4MhO0UYHOZXgqDe0ngACIqPhKOpDDzkpdrndOYQgRF4AB6WVIKi3ABMAGYACwAViBSjFqNuktQ0rlCsQSrVWp1iiAA

Related Issues:
I found reference to the gnarly type above in #32608 & #33568.
It looks to be the same "monster" as you see in #29505 too.