This PR includes a sequence of commits that gradually dismantles the `librustrt` `rtio` system -- the main trait previously used to abstract over green and native io. It also largely dismantles `libnative`, moving much of its code into `libstd` and refactoring as it does so.
TL;DR:
* Before this PR: `rustc hello.rs && wc -c hello` produces 715,996
* After this PR: `rustc hello.rs && wc -c hello` produces 368,100
That is, this PR reduces the footprint of hello world by ~50%.
This is a major step toward #17325 (i.e. toward implementing the [runtime removal RFC](https://github.com/rust-lang/rfcs/pull/230).) What remains is to pull out the scheduling, synchronization and task infrastructure, and to remove `libgreen`. These will be done soon in a follow-up PR.
Part of the work here is eliminating the `rtio` abstraction, which in many cases means bringing the implementation of io closer to the actual API presented in `std::io`.
Another aspect of this PR is the creation of two new, *private* modules within `std` that implement io:
* The `sys` module, which represents a platform-specific implementation of a number of low-level abstractions that are used directly within `std::io` and `std::os`. These "abstractions" are left largely the same as they were in `libnative` (except for the removal of `Arc` in file descriptors), but they are expected to evolve greatly over time. Organizationally, there are `sys/unix/` and `sys/windows/` directories which both implement the entire `sys` module hierarchy; this means that nearly all of the platform-specific code is isolated and you can get a handle on each platform in isolation.
* The `sys_common` module, which is rooted at `sys/common`, and provides a few pieces of private, low-level, but cross-platform functionality.
In the long term, the `sys` modules will provide hooks for exposing high-level platform-specific APIs as part of `libstd`. The first such API will be access to file descriptors from `std::io` abstractions, but a bit of design work remains before that step can be taken.
The `sys_common` module includes some traits (like `AsFileDesc`) which allow communication of private details between modules in disparate locations in the hierarchy; this helps overcome the relatively simple hierarchical privacy system in Rust.
To emphasize: the organization in `sys` is *very preliminary* and the main goal was to migrate away from `rtio` as quickly and simply as possible. The design will certainly evolve over time, and all of the details are currently private.
Along the way, this PR also entirely removes signal handling, since it was only supported on `librustuv` which was removed a while ago.
Because of the removal of APIs from `libnative` and `librustrt`, and the removal of signal handling, this is a:
[breaking-change]
Some of these APIs will return in public from from `std` over time.
r? @alexcrichton
Various miscellaneous changes pushing towards HRTB support:
1. Update parser and adjust ast to support `for<'a,'b>` syntax, both in closures and trait bounds. Warn on the old syntax (not error, for stage0).
2. Refactor TyTrait representation to include a TraitRef.
3. Purge `once_fns` feature gate and `once` keyword.
r? @pcwalton
This is a [breaking-change]:
- The `once_fns` feature is now officially deprecated. Rewrite using normal closures or unboxed closures.
- The new `for`-based syntax now issues warnings (but not yet errors):
- `fn<'a>(T) -> U` becomes `for<'a> fn(T) -> U`
- `<'a> |T| -> U` becomes `for<'a> |T| -> U`
`FnOnce` environments that fit within an `int` are passed to the closure by value. For some reason there was an assert that this would only happen if there were 1 or 0 free variables, but it can also happen if there are multiple variables that happen to fit.
Closes#18652
This commit implements processing these two attributes at the crate level as
well as at the item level. When #[cfg] is applied at the crate level, then the
entire crate will be omitted if the cfg doesn't match. The #[cfg_attr] attribute
is processed as usual in that the attribute is included or not depending on
whether the cfg matches.
This was spurred on by motivations of #18585 where #[cfg_attr] annotations will
be applied at the crate-level.
cc #18585
This resolves some issues that remained after adding support for monomorphizing unboxed closures in trans.
There were a few places where a set of substitutions for an unboxed closure type were dropped on the floor and later recalculated from scratch based on the def ID, but this failed spectacularly when the closure originated from a different param environment. The substitutions are now plumbed through end-to-end. Closes#18661
There was also a conflict in the meaning of the self param space within the body of the unboxed closure. Trans attempted to insert the unboxed closure type as the self type, but this could conflict with the self type from the param environment when an unboxed closure was used within a default method on a trait. Since the body of an unboxed closure cannot refer to its own self type or value, there's no need for it to actually use the self space. The downstream consumers of the substitutions in trans do not seem to need it either since they look up the type of the closure some other way, so I just stopped setting it. Closes#18685.
r? @pcwalton @nikomatsakis
This commit implements processing these two attributes at the crate level as
well as at the item level. When #[cfg] is applied at the crate level, then the
entire crate will be omitted if the cfg doesn't match. The #[cfg_attr] attribute
is processed as usual in that the attribute is included or not depending on
whether the cfg matches.
This was spurred on by motivations of #18585 where #[cfg_attr] annotations will
be applied at the crate-level.
cc #18585
Fixes#18567. `Struct{x:foo, .. with_expr}` did not walk `with_expr`, which allowed
using moved variables in some cases. The CFG for structs also built up with
`with_expr` happening before the fields, which is now reversed. (Fields are now
before the `with_expr` in the CFG)
Fixes#18567. Struct{x:foo, .. with_expr} did not walk with_expr, which allowed
using moved variables in some cases. The CFG for structs also built up with
with_expr happening before the fields, which is now reversed. (Fields are now
before the with_expr in the CFG)
This commit adds support for linting `extern crate` statements for stability
attributes attached to the crate itself. This is likely to be the mechanism used
to deny access to experimental crates that are part of the standard
distribution.
cc #18585
r? @aturon
`eq`, `ne`, `cmp`, etc methods now require one less level of indirection when dealing with `&str`/`&[T]`
``` rust
"foo".ne(&"bar") -> "foo".ne("bar")
slice.cmp(&another_slice) -> slice.cmp(another_slice)
// slice and another_slice have type `&[T]`
```
[breaking-change]
This branch cleans up overloaded operator resolution so that it is strictly based on the traits in `ops`, rather than going through the normal method lookup mechanism. It also adds full support for autoderef to overloaded index (whereas before autoderef only worked for non-overloaded index) as well as for the slicing operators.
This is a [breaking-change]: in the past, we were accepting combinations of operands that were not intended to be accepted. For example, it was possible to compare a fixed-length array and a slice, or apply the `!` operator to a `&int`. See the first two commits in this pull-request for examples.
One downside of this change is that comparing fixed-length arrays doesn't always work as smoothly as it did before. Before this, comparisons sometimes worked due to various coercions to slices. I've added impls for `Eq`, `Ord`, etc for fixed-lengths arrays up to and including length 32, but if the array is longer than that you'll need to either newtype the array or convert to slices. Note that this plays better with deriving in any case than the previous scheme.
Fixes#4920.
Fixes#16821.
Fixes#15757.
cc @alexcrichton
cc @aturon
This is a follow-up to [RFC PR #173](https://github.com/rust-lang/rfcs/pull/173). I was told there that changes like this don't need to go through the RFC process, so I'm submitting this directly.
This PR introduces `ToSocketAddr` trait as defined in said RFC. This trait defines a conversion from different types like `&str`, `(&str, u16)` or even `SocketAddr` to `SocketAddr`. Then this trait is used in all constructor methods for `TcpStream`, `TcpListener` and `UdpSocket`.
This unifies these constructor methods - previously they were using different types of input parameters (TCP ones used `(&str, u16)` pair while UDP ones used `SocketAddr`), which is not consistent by itself and sometimes inconvenient - for example, when the address initially is available as `SocketAddr`, you still need to convert it to string to pass it to e.g. `TcpStream`. This is very prominently demonstrated by the unit tests for TCP functionality. This PR makes working with network objects much like with `Path`, which also uses similar trait to be able to be constructed from `&[u8]`, `Vec<u8>` and other `Path`s.
This is a breaking change. If constant literals were used before, like this:
```rust
TcpStream::connect("localhost", 12345)
```
then the nicest fix is to change it to this:
```rust
TcpStream::connect("localhost:12345")
```
If variables were used before, like this:
```rust
TcpStream::connect(some_address, some_port)
```
then the arguments should be wrapped in another set of parentheses:
```rust
TcpStream::connect((some_address, some_port))
```
`UdpSocket` usages won't break because its constructor method accepted `SocketAddr` which implements `ToSocketAddr`, so `bind()` calls:
```rust
UdpSocket::bind(some_socket_addr)
```
will continue working as before.
I haven't changed `UdpStream` constructor because it is deprecated anyway.
