Deny common lints by default for lib{std,extra}

[alexcrichton]

Closes #6963

I imagine though that this will bounce in bors once or twice... Because attributes can't be cfg(stage0)'d off, there's temporarily a lot of new stage0/stage1+ code.

[alexcrichton]

Fixed the warnings related to windows, so should be ready for another go.

[brson]

I suspect that this approach is going to be hard to maintain because of the differences in stage0 vs. stage1. What I would prefer is to have a -Werror (but not necessarily that syntax) equivalent that can be passed to rustc in stage1+, leaving stage0 as warnings.

[alexcrichton]

@brson, I just modified it to apply the -D flags via the makefile instead of using attributes inside of the crate itself. This way anything other than stage1/2 don't have the flags.

[alexcrichton]

I was also mucking around in the makefiles and managed to go ahead and rename the rc extensions to rs extensions

[brson]

@alexcrichton I didn't expect all the stage0 conditional definitions when using the -D flag in only stage1+. What is it that requires the #[cfg(stage0)] stuff here?

[brson]

Also, thanks for taking my feedback.

[alexcrichton]

Most of them have to do with unsafe blocks, and the reason is that the unsafe checker was changed.

In stage0, this is invalid code:

use std::libc;

fn callback(_: ~fn()) {}

unsafe fn foo() {
  do callback {
    libc::exit(1);
  }
}

fn main() {
}

Whereas it's valid in stage1. I'm not sure if this is an unintended bug from when the unsafe checker changed, but in stage0 the owned function is required to explicitly have another unsafe block, whereas in stage1 the top-level unsafe satisfies the un-safeness requirement.

This means that to prevent errors in stage1 and so that code compiles in stage0, there have to be two versions.

[alexcrichton]

@brson, for some reason github is reordering the commits again, I just rebased (to fix conflicts), but the head of the pull request is 8394895 instead of the other

[pnkfelix]

@brson what do you think about turning off warnings entirely for the stage0 build, so that we don't get annoyed by bugs in lint that are not in the snapshot version of the compiler?

[pnkfelix]

(or at least add [or just document?] an easy way, perhaps a configure flag, to let one easily opt into ignoring all warnings in the stage0 build.)

[pnkfelix]

(I'll move my comments to #6963.)

[alexcrichton]

I added a commit to turn off warnings in stage0 in #7113, I think that's a good idea as well!