Add Gankro's table to nomicon/src/phantom-data.md

[Rufflewind]

Original: #30069 (comment)

Testing confirms that:

• PhantomData<fn() -> T> does not actually enable drop checking.
• PhantomData<fn(T) -> T> is neither variant nor contravariant.

[rust-highfive]

r? @steveklabnik

(rust_highfive has picked a reviewer for you, use r? to override)

[steveklabnik]

@bors: r+

neat, thank you! ❤️

[bors]

📌 Commit 49fa66c has been approved by steveklabnik

[bluss]

@pnkfelix Does PhantomData<T> have a special role in drop checking now? If we compare with PhantomData<*const T>, don't both those have the same drop check implications nowadays?

[Rufflewind]

This is what I used to test whether the drop checker is active:

use std::marker::PhantomData;structInspector<'a>(&'a());impl<'a>DropforInspector<'a>{fndrop(&mutself){}}fnwrap<T>(_:T) -> PhantomData<T>{PhantomData}fnmain(){let(i, inspector);
i = ();
inspector = wrap(Inspector(&i));}

Right now this fails to compile. But if I use PhantomData<*const T> then it compiles successfully.

[bluss]

Nice!

You said:

PhantomData<fn() -> T> does enable drop checking.

The code doesn't demonstrate this, did you mean the opposite?

[Rufflewind]

The code doesn't demonstrate this, did you mean the opposite?

Yeah, I meant the opposite. The documentation is correct, but I messed up the commit message / pull request -.-

Rebasing…

[steveklabnik]

@bors: r+ rollup

nice catch all, sorry I missed it! 😓

[bors]

📌 Commit 1f75085 has been approved by steveklabnik

[bluss]

Thanks for the clarification and the new and better docs!