Implement `std::marker::Tuple`, use it in `extern "rust-call"` and `Fn`-family traits
Implements rust-lang/compiler-team#537
I made a few opinionated decisions in this implementation, specifically:
1. Enforcing `extern "rust-call"` on fn items during wfcheck,
2. Enforcing this for all functions (not just ones that have bodies),
3. Gating this `Tuple` marker trait behind its own feature, instead of grouping it into (e.g.) `unboxed_closures`.
Still needing to be done:
1. Enforce that `extern "rust-call"` `fn`-ptrs are well-formed only if they have 1/2 args and the second one implements `Tuple`. (Doing this would fix ICE in #66696.)
2. Deny all explicit/user `impl`s of the `Tuple` trait, kinda like `Sized`.
3. Fixing `Tuple` trait built-in impl for chalk, so that chalkification tests are un-broken.
Open questions:
1. Does this need t-lang or t-libs signoff?
Fixes#99820
fix a comment in UnsafeCell::new
There are several safe methods that access the inner value: `into_inner` has existed since forever and `get_mut` also exists since recently. So this comment seems just wrong. But `&self` methods return raw pointers and thus require unsafe code (though the methods themselves are still safe).
benchmark result:
```
$ cargo bench
Compiling div-euclid v0.1.0 (/me/div-euclid)
Finished bench [optimized] target(s) in 1.01s
Running unittests src/lib.rs (target/release/deps/div_euclid-7a4530ca7817d1ef)
running 7 tests
test tests::it_works ... ignored
test tests::bench_aaabs ... bench: 10,498,793 ns/iter (+/- 104,360)
test tests::bench_aadefault ... bench: 11,061,862 ns/iter (+/- 94,107)
test tests::bench_abs ... bench: 10,477,193 ns/iter (+/- 81,942)
test tests::bench_default ... bench: 10,622,983 ns/iter (+/- 25,119)
test tests::bench_zzabs ... bench: 10,481,971 ns/iter (+/- 43,787)
test tests::bench_zzdefault ... bench: 11,074,976 ns/iter (+/- 29,633)
test result: ok. 0 passed; 0 failed; 1 ignored; 6 measured; 0 filtered out; finished in 19.35s
```
benchmark code:
```rust
#![feature(test)]
extern crate test;
#[inline(always)]
fn rem_euclid(a:i32,rhs:i32)->i32{
let r = a % rhs;
if r < 0 {
// if rhs is `integer::MIN`, rhs.wrapping_abs() == rhs.wrapping_abs,
// thus r.wrapping_add(rhs.wrapping_abs()) == r.wrapping_add(rhs) == r - rhs,
// which suits our need.
// otherwise, rhs.wrapping_abs() == -rhs, which won't overflow since r is negative.
r.wrapping_add(rhs.wrapping_abs())
} else {
r
}
}
#[cfg(test)]
mod tests {
use super::*;
use test::Bencher;
use rand::prelude::*;
use rand::rngs::SmallRng;
const N:i32=1000;
#[test]
fn it_works() {
let a: i32 = 7; // or any other integer type
let b = 4;
let d:Vec<i32>=(-N..=N).collect();
let n:Vec<i32>=(-N..0).chain(1..=N).collect();
for i in &d {
for j in &n {
assert_eq!(i.rem_euclid(*j),rem_euclid(*i,*j));
}
}
assert_eq!(rem_euclid(a,b), 3);
assert_eq!(rem_euclid(-a,b), 1);
assert_eq!(rem_euclid(a,-b), 3);
assert_eq!(rem_euclid(-a,-b), 1);
}
#[bench]
fn bench_aaabs(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=rem_euclid(*i,*j);
}
}
res
});
}
#[bench]
fn bench_aadefault(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=i.rem_euclid(*j);
}
}
res
});
}
#[bench]
fn bench_abs(b: &mut Bencher) {
let d:Vec<i32>=(-N..=N).collect();
let n:Vec<i32>=(-N..0).chain(1..=N).collect();
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=rem_euclid(*i,*j);
}
}
res
});
}
#[bench]
fn bench_default(b: &mut Bencher) {
let d:Vec<i32>=(-N..=N).collect();
let n:Vec<i32>=(-N..0).chain(1..=N).collect();
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=i.rem_euclid(*j);
}
}
res
});
}
#[bench]
fn bench_zzabs(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=rem_euclid(*i,*j);
}
}
res
});
}
#[bench]
fn bench_zzdefault(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=i.rem_euclid(*j);
}
}
res
});
}
}
```
such code is copy from
https://github.com/rust-lang/rust/blob/master/library/std/src/f32.rs
and
https://github.com/rust-lang/rust/blob/master/library/std/src/f64.rs
using r+rhs.abs() is faster than calc it directly.
Bench result:
```
$ cargo bench
Compiling div-euclid v0.1.0 (/me/div-euclid)
Finished bench [optimized] target(s) in 1.01s
Running unittests src/lib.rs (target/release/deps/div_euclid-7a4530ca7817d1ef)
running 7 tests
test tests::it_works ... ignored
test tests::bench_aaabs ... bench: 10,498,793 ns/iter (+/- 104,360)
test tests::bench_aadefault ... bench: 11,061,862 ns/iter (+/- 94,107)
test tests::bench_abs ... bench: 10,477,193 ns/iter (+/- 81,942)
test tests::bench_default ... bench: 10,622,983 ns/iter (+/- 25,119)
test tests::bench_zzabs ... bench: 10,481,971 ns/iter (+/- 43,787)
test tests::bench_zzdefault ... bench: 11,074,976 ns/iter (+/- 29,633)
test result: ok. 0 passed; 0 failed; 1 ignored; 6 measured; 0 filtered out; finished in 19.35s
```
bench code:
```
#![feature(test)]
extern crate test;
fn rem_euclid(a:i32,rhs:i32)->i32{
let r = a % rhs;
if r < 0 { r + rhs.abs() } else { r }
}
#[cfg(test)]
mod tests {
use super::*;
use test::Bencher;
use rand::prelude::*;
use rand::rngs::SmallRng;
const N:i32=1000;
#[test]
fn it_works() {
let a: i32 = 7; // or any other integer type
let b = 4;
let d:Vec<i32>=(-N..=N).collect();
let n:Vec<i32>=(-N..0).chain(1..=N).collect();
for i in &d {
for j in &n {
assert_eq!(i.rem_euclid(*j),rem_euclid(*i,*j));
}
}
assert_eq!(rem_euclid(a,b), 3);
assert_eq!(rem_euclid(-a,b), 1);
assert_eq!(rem_euclid(a,-b), 3);
assert_eq!(rem_euclid(-a,-b), 1);
}
#[bench]
fn bench_aaabs(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=rem_euclid(*i,*j);
}
}
res
});
}
#[bench]
fn bench_aadefault(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=i.rem_euclid(*j);
}
}
res
});
}
#[bench]
fn bench_abs(b: &mut Bencher) {
let d:Vec<i32>=(-N..=N).collect();
let n:Vec<i32>=(-N..0).chain(1..=N).collect();
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=rem_euclid(*i,*j);
}
}
res
});
}
#[bench]
fn bench_default(b: &mut Bencher) {
let d:Vec<i32>=(-N..=N).collect();
let n:Vec<i32>=(-N..0).chain(1..=N).collect();
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=i.rem_euclid(*j);
}
}
res
});
}
#[bench]
fn bench_zzabs(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=rem_euclid(*i,*j);
}
}
res
});
}
#[bench]
fn bench_zzdefault(b: &mut Bencher) {
let mut d:Vec<i32>=(-N..=N).collect();
let mut n:Vec<i32>=(-N..0).chain(1..=N).collect();
let mut rng=SmallRng::from_seed([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,21]);
d.shuffle(&mut rng);
n.shuffle(&mut rng);
d.shuffle(&mut rng);
b.iter(||{
let mut res=0;
for i in &d {
for j in &n {
res+=i.rem_euclid(*j);
}
}
res
});
}
}
```
The new implementation doesn't use weak lang items and instead changes
`#[alloc_error_handler]` to an attribute macro just like
`#[global_allocator]`.
The attribute will generate the `__rg_oom` function which is called by
the compiler-generated `__rust_alloc_error_handler`. If no `__rg_oom`
function is defined in any crate then the compiler shim will call
`__rdl_oom` in the alloc crate which will simply panic.
This also fixes link errors with `-C link-dead-code` with
`default_alloc_error_handler`: `__rg_oom` was previously defined in the
alloc crate and would attempt to reference the `oom` lang item, even if
it didn't exist. This worked as long as `__rg_oom` was excluded from
linking since it was not called.
This is a prerequisite for the stabilization of
`default_alloc_error_handler` (#102318).
The signature for new was
```
fn new<F>(f: F) -> Lazy<T, F>
```
Notably, with `F` unconstrained, `T` can be literally anything, and just
`let _ = Lazy::new(|| 92)` would not typecheck.
This historiacally was a necessity -- `new` is a `const` function, it
couldn't have any bounds. Today though, we can move `new` under the `F:
FnOnce() -> T` bound, which gives the compiler enough data to infer the
type of T from closure.
poll_fn and Unpin: fix pinning
See [IRLO](https://internals.rust-lang.org/t/surprising-soundness-trouble-around-pollfn/17484) for details: currently `poll_fn` is very subtle to use, since it does not pin the closure, so creating a `Pin::get_unchcked(&mut capture)` inside the closure is unsound. This leads to actual miscompilations with `futures::join!`.
IMO the proper fix is to pin the closure when the future is pinned, which is achieved by changing the `Unpin` implementation. This is a breaking change though. 1.64.0 was *just* released, so maybe this is still okay?
The alternative would be to add some strong comments to the docs saying that closure captures are *not pinned* and doing `Pin::get_unchecked` on them is unsound.
