Nightly regression: empty lifetime list in `for<lifetime-list>` doesn't parse anymore

[petrochenkov]

Code:

type A = for<> fn(); // error: trailing attribute after lifetime parameters

fn main() {}

Compiles on 1.13-stable,beta, fails on 1.14-nightly.

[TimNN]

I think this was a deliberate change, I'm pretty sure I remember a PR who's whole point was to disallow exactly that (although I can't finde the PR right now). (Forget that, see below).

[TimNN]

Apart from that, this was introduced between nightly-2016-09-30 and nightly-2016-10-03 (Changes).

[petrochenkov]

I strongly suspect this is an unintended side effect of #34764.
Prohibiting this deliberately doesn't make any sense, empty lists are allowed in (almost) all other possible contexts.

[TimNN]

Yeah, I probably mis-remembered :(

I agree that this was probably a side effect of #34764.

[est31]

In fact, this fails too:

type B = for<'a,> fn();