This defines the `_mm256_blendv_pd` and `_mm256_blendv_ps` intrinsics.
The `_mm256_blend_pd` and `_mm256_blend_ps` intrinsics are not available
as LLVM intrinsics. In Clang they are implemented using the
shufflevector builtin.
Intel reference: https://software.intel.com/en-us/node/524070.
Do not report errors from regionck if other errors were already reported
Do not report errors from regionck if other errors were already reported during the lifetime of this inferencer. Fixes#30580.
r? @arielb1
Add Pass manager for MIR
A new PR, since rebasing the original one (https://github.com/rust-lang/rust/pull/31448) properly was a pain. Since then there has been several changes most notable of which:
1. Removed the pretty-printing with `#[rustc_mir(graphviz/pretty)]`, mostly because we now have `--unpretty=mir`, IMHO that’s the direction we should expand this functionality into;
2. Reverted the infercx change done for typeck, because typeck can make an infercx for itself by being a `MirMapPass`
r? @nikomatsakis
Call str::to_owned in String::from and uninline it
Call str::to_owned in String::from and uninline it
These methods were already effectively equal, but now one calls
the other, and neither is marked inline.
String::from does not need to be inlined, it can be without it just like
str::to_owned and String::clone are.
Fixes#32163
Fix name resolution in lexical scopes
Currently, `resolve_item_in_lexical_scope` does not check the "ribs" (type parameters and local variables). This can allow items that should be shadowed by type parameters to be named.
For example,
```rust
struct T { i: i32 }
fn f<T>() {
let t = T { i: 0 }; // This use of `T` resolves to the struct, not the type parameter
}
mod Foo {
pub fn f() {}
}
fn g<Foo>() {
Foo::f(); // This use of `Foo` resolves to the module, not the type parameter
}
```
This PR changes `resolve_item_in_lexical_scope` so that it fails when the item is shadowed by a rib (fixes#32120).
This is a [breaking-change], but it looks unlikely to cause breakage in practice.
r? @nikomatsakis
Add AVX broadcast and conversion intrinsics
This adds the following intrinsics:
* `_mm256_broadcast_pd`
* `_mm256_broadcast_ps`
* `_mm256_cvtepi32_pd`
* `_mm256_cvtepi32_ps`
* `_mm256_cvtpd_epi32`
* `_mm256_cvtpd_ps`
* `_mm256_cvtps_epi32`
* `_mm256_cvtps_pd`
* `_mm256_cvttpd_epi32`
* `_mm256_cvttps_epi32`
The "avx" codegen feature must be enabled to use these.
std: Fix tracking issues and clean deprecated APIs
This PR fixes up a number of discrepancies found with tracking issues (some closed, some needed new ones, etc), and also cleans out all pre-1.8 deprecated APIs. The big beast here was dealing with `std::dynamic_lib`, and via many applications of a large hammer it's now out of the standard library.
Removes all unstable and deprecated APIs prior to the 1.8 release. All APIs that
are deprecated in the 1.8 release are sticking around for the rest of this
cycle.
Some notable changes are:
* The `dynamic_lib` module was moved into `rustc_back` as the compiler still
relies on a few bits and pieces.
* The `DebugTuple` formatter now special-cases an empty struct name with only
one field to append a trailing comma.
Forbid items with the same name from appearing in overlapping inherent impl blocks
For example, the following is now correctly illegal:
```rust
struct Foo;
impl Foo {
fn id() {}
}
impl Foo {
fn id() {}
}
```
"Overlapping" here is determined the same way it is for traits (and in fact shares the same code path): roughly, there must be some way of substituting any generic types to unify the impls, such that none of the `where` clauses are provably unsatisfiable under such a unification.
Along the way, this PR also introduces an `ImplHeader` abstraction (the first commit) that makes it easier to work with impls abstractly (without caring whether they are trait or inherent impl blocks); see the first commit.
Closes#22889
r? @nikomatsakis
cover more linux targets in libstd cargobuild
libstd/build.rs checked the target name against `"unknown-linux"`... That doesn't necessarily apply to all Linux targets as some toolchains for embedded targets will change that `unknown` field to some vendor name. Some shifting around was needed since Android is also a Linux target.
r? @alexcrichton
lint: mark associated types as live for the dead_code pass
Associated types of trait impls were being excluded from the live list. So types that only appeared in trait impls were being marked as dead code.
Add missing documentation examples for BinaryHeap.
As part of the ongoing effort to document all methods with examples,
this commit adds the missing examples for the `BinaryHeap` collection
type.
This is part of issue #29348.
r? @steveklabnik
impl blocks.
For example, the following is now correctly illegal:
```rust
struct Foo;
impl Foo {
fn id() {}
}
impl Foo {
fn id() {}
}
```
"Overlapping" here is determined the same way it is for traits (and in
fact shares the same code path): roughly, there must be some way of
substituting any generic types to unify the impls, such that none of the
`where` clauses are provably unsatisfiable under such a unification.
Closes#22889
This commit introduces the idea of an "impl header", which consists of
everything outside the impl body: the Self type, the trait
reference (when applicable), and predicates from `where` clauses. This
type is usable with the type folding machinery, making it possible to
work with impl headers at a higher and more generic level.
Forbid glob-importing from a type alias
This PR forbids glob-importing from a type alias or trait (fixes#30560):
```rust
type Alias = ();
use Alias::*; // This is currently allowed but shouldn't be
```
This is a [breaking-change]. Since the disallowed glob imports don't actually import anything, any breakage can be fixed by removing the offending glob import.
r? @alexcrichton
Add a link validator to rustbuild
This commit was originally targeted at just adding a link checking script to the rustbuild system. This ended up snowballing a bit to extend rustbuild to be amenable to various tools we have as part of the build system in general.
There's a new `src/tools` directory which has a number of scripts/programs that are purely intended to be used as part of the build system and CI of this repository. This is currently inhabited by rustbook, the error index generator, and a new linkchecker script added as part of this PR. I suspect that more tools like compiletest, tidy scripts, snapshot scripts, etc will migrate their way into this directory over time.
The commit which adds the error index generator shows the steps necessary to add new tools to the build system, namely:
1. New steps are defined for building the tool and running the tool
2. The dependencies are configured
3. The steps are implemented
In terms of the link checker, these commits do a few things:
* A new `src/tools/linkchecker` script is added. This will read an entire documentation tree looking for broken relative links (HTTP links aren't followed yet).
* A large number of broken links throughout the documentation were fixed. Many of these were just broken when viewed from core as opposed to std, but were easily fixed.
* A few rustdoc bugs here and there were fixed
rustc: Add an i586-pc-windows-msvc target
Similarly to #31629 where an i586-unknown-linux-gnu target was added, there is
sometimes a desire to compile for x86 Windows as well where SSE2 is disabled.
This commit mirrors the i586-unknown-linux-gnu target and simply adds a variant
for Windows as well.
This is motivated by a recent [Gecko bug][ff] where crashes were seen on 32-bit
Windows due to users having CPUs that don't support SSE2 instructions. It was
requested that we could have non-SSE2 builds of the standard library available
so they could continue to use vanilla releases and nightlies.
[ff]: https://bugzilla.mozilla.org/show_bug.cgi?id=1253202
rustdoc: correct src-link url
It would have been nice for htmldocck to catch this, especially since there are rustdoc tests specifically for checking src-links.
fixes#32113
WriteConsoleW can fail if called with a large buffer so we need to slice
any stdout/stderr output.
However the current slicing has a few problems:
1. It slices by byte but still expects valid UTF-8.
2. The slicing happens even when not outputting to a console.
3. panic! output is not sliced.
This fixes these issues by moving the slice to right before
WriteConsoleW and slicing on a char boundary.
`assert_eq!(-9223372036854775808isize as u64, 0x8000000000000000);`
fails on 32 bit and succeeds on 64 bit. These commits don't change that behavior.
The following worked before my changes, because the discriminant was
always processed as `u64`.
Now it fails, because the discriminant of `Bu64` is now `0` instead of `-9223372036854775808`. This is more in line with the above assertion's code, since
`-9223372036854775808isize as u64` on 32 bit yielded `0`.
```rust
enum Eu64 {
Au64 = 0,
Bu64 = 0x8000_0000_0000_0000
}
```
Small grammar fix in Guessing Game
When it was Option.expect(), there was an .ok().expect(), but now that it uses Result.expect(), there's only one method, not two.
Fixes#31912
Add missing documentation examples for BTreeSet.
As part of the ongoing effort to document all methods with examples,
this commit adds the missing examples for the `BTreeSet` collection
type.
This is part of issue #29348.
r? @steveklabnik
Remove final note from testing chapter.
The information that documentation tests cannot be run in binary crates is already given at the beginning of the section.
Clarify documentation of hash::SipHasher
The docs were making assertions/recommendations they shouldn't have. This clarifies them and adds some helpful links.
Fixes#32043.
r? @sfackler
Prefer 'associated function' over 'static method' in msg.
TRPL seems to refer to 'static functions' as 'associated functions'.
This terminology should be used consistently.
"can be built on Ref::map"… how?
Now that `std::cell::Ref::filter_map` and `RefMut::filter_map` are deprecated, using them gives a warning like:
```
script/dom/element.rs:754:9: 754:24 warning: use of deprecated item: can be built on Ref::map, #[warn(deprecated)] on by default
```
But it’s not at all obvious *how* the functionality can be built on `Ref::map`. This PR adds to the warning message a crates.io URL for a crate that does.
Fix a regression in import resolution
This fixes#32089 (caused by #31726) by deducing that name resolution has failed (as opposed to being determinate) in more cases.
r? @nikomatsakis
Distinguish fn item types to allow reification from nothing to fn pointers.
The first commit is a rebase of #26284, except for files that have moved since.
This is a [breaking-change], due to:
* each FFI function has a distinct type, like all other functions currently do
* all generic parameters on functions are recorded in their item types, e.g.:
`size_of::<u8>` & `size_of::<i8>`'s types differ despite their identical signature.
* function items are zero-sized, which will stop transmutes from working on them
The first two cases are handled in most cases with the new coerce-unify logic,
which will combine incompatible function item types into function pointers,
at the outer-most level of if-else chains, match arms and array literals.
The last case is specially handled during type-checking such that transmutes
from a function item type to a pointer or integer type will continue to work for
another release cycle, but are being linted against. To get rid of warnings and
ensure your code will continue to compile, cast to a pointer before transmuting.
Optimize some functions in std::process
* Be sure that `read_to_end` gets directed towards `read_to_end_uninitialized` for all handles
* When spawning a child that guaranteed doesn't need a stdin, don't actually create a stdin pipe for that process, instead just redirect it to /dev/null
* When calling `wait_with_output`, don't spawn threads to read out the pipes of the child. Instead drain all pipes on the calling thread and *then* wait on the process.
Functionally, it is intended that nothing changes as part of this PR
---
Note that this was the same as #31613, and even after that it turned out that fixing Windows was easier than I thought! To copy a comment from over there:
> As some rationale for this as well, it's always bothered me that we've spawned threads in the standard library for this (seems a bit overkill), and I've also been curious lately as to our why our build times for Windows are so much higher than Unix (on the buildbots we have). I have done basically 0 investigation into why, but I figured it can't help to try to optimize Command::output which I believe is called quite a few times during the test suite.
