dropck: The drop order is now defined

[vorner]

So the examples are out of date, they no longer pose a problem. However,
there's still a case where it does, so let's use that instead.

Also, mention the definition of drop order in the text, as it is
related.

[Gankra]

Could you explain the motivation for rewriting all the examples? As far as I know nothing has been done that changes their correctness, and I don't think your changes at all make things more clear.

[vorner]

It's been a while, so I have to guess a bit. However, the problem with the examples I had is, they still don't compile as they are supposed not to compile. But the reason they do not compile is different and the reason stated on the page is outdated. It's not because one does not outlive the other, but because the order of drop is reversed. If you switch the order in the tuple/pattern (eg. use (days, inspector) instead of (inspector, days)) it now starts compiling. So the example is not relevant to the chapter at all any more.

https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=117cf19b8585f2c86f6f530b41a72fc9

I tried to modify the inspector example, but didn't come up with a way. So I created new ones that demonstrate the issue. If you have an idea how to modify the example in a way that doesn't compile in either order of the tuple, I'm all for it.

[Gankra]

Could you not just change the examples to let tup = (a, b) and using tup.0 and tup.1 (or perhaps introduce a struct just for better naming?)

[vorner]

Could you not just change the examples to let tup = (a, b) and using tup.0 and tup.1 (or perhaps introduce a struct just for better naming?)

OK, I'll try playing with it.

[vorner]

Just note for myself when I get to it again: the tuple seems to do the trick ‒ this compiles in both orders of the tuple: https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=7573478e37c067e48a1e728c499773ea

I'll try to return the examples to (close to) the original once I have a bit more time. Thanks for the suggestion.

[Gankra]

Just for clarity (named fields) the struct would probably be better, but ok, great!

[vorner]

I've returned to to much closer state to the original. It's now wrapped in a structure, to make the borrow checker not being able to track the lifetimes as separate and added references to the defined drop order of fields.

I've force-pushed into the branch, as the diff is shorter to master than to the previous version.

[Gankra]

Ok this looks great, thanks!

[Gankra]

thank you!!!!