TypeCheckType: Fix some bugs in the any syntax checker

[AnthonyLatsis]

Resolves #65026, resolves #65027, resolves #72588.

Diff counts are mostly due to tests.

[AnthonyLatsis]

@swift-ci please smoke test macOS

[AnthonyLatsis]

@slavapestov @hborla ping

[AnthonyLatsis]

I have incorporated the tests I deleted here into test/type/explicit_existential.swift and test/type/explicit_existential_swift6.swift

[AnthonyLatsis]

@swift-ci please smoke test Linux

[AnthonyLatsis]

This is unfortunate although definitely not a ExistentialTypeSyntaxChecker problem.

[AnthonyLatsis]

@swift-ci please smoke test Linux

[AnthonyLatsis]

@swift-ci please test source compatibility release

[AnthonyLatsis]

@hborla ping

[AnthonyLatsis]

@slavapestov ping

[slavapestov]

You might be able to just pass down the outermost location instead of "backtracking" here, but it's fine the way it is

[AnthonyLatsis]

Thanks for sharing your thoughts on this, Slava.

You might be able to just pass down the outermost location

Is this roughly what you meant (passing tildeLoc down such that if valid here, then the ~ necessarily pertains to the given identifier type)?

PreWalkAction walkToTypeReprPre(TypeRepr *T) override {
  // ...
  if (auto *inverse = isa<InverseTypeRepr>(T)) {
    this->tildeLoc = inverse->getTildeLoc();
  } else if (this->tildeLoc.isValid()) {
    if (!T->isParenType() &&
        !isa<MetatypeTypeRepr>(T) &&
        !isa<DeclRefTypeRepr>(T)) {
      this->tildeLoc = SourceLoc();
    }
  }
  // ...
  
  checkDeclRefTypeRepr(declRefTR);
  
  this->tildeLoc = SourceLoc();

  // ...
}

[AnthonyLatsis]

I’m going to try to get this into the release as is before it’s too late, but I’m still very much willing to discuss this alternative.

[AnthonyLatsis]

@swift-ci please smoke test