Improve SubSupConflict with a named and an anonymous lifetime parameter #42701
Hello!
This fixes#42701.
## UPDATE 01
Tests are producing different results between different env builds.
This inconsistency might take a long time to investigate and fix. So, be patient
## UPDATE 02
Changed an `FxHashMap` with a `BTreeMap`. Inconsistency seems to be resolved for now.
rustbuild: Disable ThinLTO for libtest
Right now ThinLTO is generating bad dwarf which is tracked by #45511, but this
is causing issues on OSX (#45768) where `dsymutil` is segfaulting and failing to
produce output.
Closes#45768
Fix MIR CopyPropagation errneously propagating assignments to function args
Compiling this code with MIR CopyPropagation activated will result in printing `5`,
because CopyProp errneously propagates the assignment of `5` to all `x`:
```rust
fn bar(mut x: u8) {
println!("{}", x);
x = 5;
}
fn main() {
bar(123);
}
```
If a local is propagated, it will result in an ICE at trans due to an use-before-def:
```rust
fn dummy(x: u8) -> u8 { x }
fn foo(mut x: u8) {
x = dummy(x); // this will assign a local to `x`
}
```
Currently CopyProp conservatively gives up if there are multiple assignments to a local,
but it is not took into account that arguments are already assigned from the beginning.
This PR fixes the problem by preventing propagation of assignments to function arguments.
Make saturating u128 -> f32 casts the default behavior
... rather than being gated by `-Z saturating-float-casts`. There are several reasons for this:
1. Const eval already implements this behavior.
2. Unlike with float->int casts, this behavior is uncontroversially the right behavior and it is not as performance critical. Thus there is no particular need to make the bug fix for u128->f32 casts opt-in.
3. Having two orthogonal features under one flag is silly, and never should have happened in the first place.
4. Benchmarking float->int casts with the -Z flag should not pick up performance changes due to the u128->f32 casts (assuming there are any).
Fixes#41799
Implement arbitrary_self_types
r? @arielb1
cc @nikomatsakis
Partial implementation of #44874. Supports trait and struct methods with arbitrary self types, as long as the type derefs (transitively) to `Self`. Doesn't support raw-pointer `self` yet.
Methods with non-standard self types (i.e. anything other than `&self, &mut self, and Box<Self>`) are not object safe, because dynamic dispatch hasn't been implemented for them yet.
I believe this is also a (partial) fix for #27941.
Addressed issues raised in #44286. (`OccupiedEntry::replace_entry`)
This commit renames the `replace` function to `replace_entry`, and
creates a seperate `replace_key` function for `OccupiedEntry`. The
original `replace` function did not solve the use-case where the
key needed to be replaced, but not the value. Documentation and
naming has also been updated to better reflect what the original
replace function does.
Improve SliceExt::binary_search performance
Improve the performance of binary_search by reducing the number of unpredictable conditional branches in the loop. In addition improve the benchmarks to test performance in l1, l2 and l3 caches on sorted arrays with or without dups.
Before:
```
test slice::binary_search_l1 ... bench: 48 ns/iter (+/- 1)
test slice::binary_search_l2 ... bench: 63 ns/iter (+/- 0)
test slice::binary_search_l3 ... bench: 152 ns/iter (+/- 12)
test slice::binary_search_l1_with_dups ... bench: 36 ns/iter (+/- 0)
test slice::binary_search_l2_with_dups ... bench: 64 ns/iter (+/- 1)
test slice::binary_search_l3_with_dups ... bench: 153 ns/iter (+/- 6)
```
After:
```
test slice::binary_search_l1 ... bench: 15 ns/iter (+/- 0)
test slice::binary_search_l2 ... bench: 23 ns/iter (+/- 0)
test slice::binary_search_l3 ... bench: 100 ns/iter (+/- 17)
test slice::binary_search_l1_with_dups ... bench: 15 ns/iter (+/- 0)
test slice::binary_search_l2_with_dups ... bench: 23 ns/iter (+/- 0)
test slice::binary_search_l3_with_dups ... bench: 98 ns/iter (+/- 14)
```
Accept interpolated patterns in trait method parameters
Permit this, basically
```rust
macro_rules! m {
($pat: pat) => {
trait Tr {
fn f($pat: u8) {}
}
}
}
```
it previously caused a parsing error during expansion because trait methods accept only very restricted set of patterns during parsing due to ambiguities caused by [anonymous parameters](https://github.com/rust-lang/rust/issues/41686), and this set didn't include interpolated patterns.
Some outdated messages from "no patterns allowed" errors are also removed.
Addresses https://github.com/rust-lang/rust/issues/35203#issuecomment-341937159
This commit renames the `replace` function to `replace_entry`, and
creates a seperate `replace_key` function for `OccupiedEntry`. The
original `replace` function did not solve the use-case where the
key needed to be replaced, but not the value. Documentation and
naming has also been updated to better reflect what the original
replace function does.
Fix checking of auto trait bounds in trait objects.
Any auto trait is allowed in trait object bounds. Fix duplicate check of type and lifetime parameter count, which we were [emitting twice](https://play.rust-lang.org/?gist=37dbbdbbec62dec423bb8f6d92f137cc&version=stable).
Note: This was the last use of `Send` in the compiler, meaning after a new `stage0` we could remove the `send` lang item.
rustbuild: Re-enable ThinLTO for MIPS
Now that the upstream LLVM bug is now fixed this commit cherry-picks the commit
onto our LLVM and then re-enables the ThinLTO paths for MIPS.
Closes#45654
rustc: add item name to deprecated lint warning
It can sometimes be difficult to know what is actually deprecated when you have `foo.bar()` and `bar` comes from a trait in another crate.
Prefer libproc_macro APIs to libsyntax ones in the quasi-quoter.
The shift to using `proc_macro`'s own APIs in `proc_macro::quote`, both in the implementation of the quasi-quoter and the Rust code it generates to build `TokenStream`s at runtime, greatly reduces the dependency on `libsyntax`, with the generated runtime code being completely free of it.
This is a prerequirement for introducing more abstraction/indirection between `proc_macro` and compiler implementation details (mainly those from `libsyntax`), which I want to attempt.
cc @alexcrichton @jseyfried @nrc
Right now ThinLTO is generating bad dwarf which is tracked by #45511, but this
is causing issues on OSX (#45768) where `dsymutil` is segfaulting and failing to
produce output.
Closes#45768
... rather than being gated by -Z saturating-float-casts.
There are several reasons for this:
1. Const eval already implements this behavior.
2. Unlike with float->int casts, this behavior is uncontroversially the
right behavior and it is not as performance critical. Thus there is no
particular need to make the bug fix for u128->f32 casts opt-in.
3. Having two orthogonal features under one flag is silly, and never
should have happened in the first place.
4. Benchmarking float->int casts with the -Z flag should not pick up
performance changes due to the u128->f32 casts (assuming there are any).
Fixes#41799
