Adds support for .await under the existing async_await feature gate.
Moves macro-like await! syntax to the await_macro feature gate.
Removes support for `await` as a non-keyword under the `async_await`
feature.
[wg-async-await] Drop `async fn` arguments in async block
Fixes#54716.
This PR modifies the HIR lowering (and some other places to make this work) so that unused arguments to a async function are always dropped inside the async move block and not at the end of the function body.
```
async fn foo(<pattern>: <type>) {
async move {
}
} // <-- dropped as you "exit" the fn
// ...becomes...
fn foo(__arg0: <ty>) {
async move {
let <pattern>: <ty> = __arg0;
} // <-- dropped as you "exit" the async block
}
```
However, the exact ordering of drops is not the same as a regular function, [as visible in this playground example](https://play.rust-lang.org/?version=stable&mode=debug&edition=2015&gist=be39af1a58e5d430be1eb3c722cb1ec3) - I believe this to be an unrelated issue. There is a [Zulip topic](https://rust-lang.zulipchat.com/#narrow/stream/187312-t-compiler.2Fwg-async-await/topic/.2354716.20drop.20order) for this.
r? @cramertj
cc @nikomatsakis
This commit adds an `AsyncArgument` struct to the AST that contains the
generated argument and statement that will be used in HIR lowering, name
resolution and def collection.
Currently the pretty-printer calls `write!` for every space of
indentation. On some workloads the indentation level can exceed 100, and
a faster implementation reduces instruction counts by up to 7% on a few
workloads.
This commit makes two changes - separating the `NodeId` that identifies
an enum variant from the `NodeId` that identifies the variant's
constructor; and no longer creating a `NodeId` for `Struct`-style enum
variants and structs.
Separation of the variant id and variant constructor id will allow the
rest of RFC 2008 to be implemented by lowering the visibility of the
variant's constructor without lowering the visbility of the variant
itself.
No longer creating a `NodeId` for `Struct`-style enum variants and
structs mostly simplifies logic as previously this `NodeId` wasn't used.
There were various cases where the `NodeId` wouldn't be used unless
there was an unit or tuple struct or enum variant but not all uses of
this `NodeId` had that condition, by removing this `NodeId`, this must
be explicitly dealt with. This change mostly applied cleanly, but there
were one or two cases in name resolution and one case in type check
where the existing logic required a id for `Struct`-style enum variants
and structs.
There is no longer a need to append the string `", ..."` to a functions
args as `...` is parsed as an argument and will appear in the functions
arguments.
Support defining C compatible variadic functions
## Summary
Add support for defining C compatible variadic functions in unsafe rust with
`extern "C"` according to [RFC 2137].
## Details
### Parsing
When parsing a user defined function that is `unsafe` and `extern "C"` allow
variadic signatures and inject a "spoofed" `VaList` in the new functions
signature. This allows the user to interact with the variadic arguments via a
`VaList` instead of manually using `va_start` and `va_end` (See [RFC 2137] for
details).
### Codegen
When running codegen for a variadic function, remove the "spoofed" `VaList`
from the function signature and inject `va_start` when the arg local
references are created for the function and `va_end` on return.
## TODO
- [x] Get feedback on injecting `va_start/va_end` in MIR vs codegen
- [x] Properly inject `va_end` - It seems like it should be possible to inject
`va_end` on the `TerminatorKind::Return`. I just need to figure out how
to get the `LocalRef` here.
- [x] Properly call Rust defined C variadic functions in Rust - The spoofed
`VaList` causes problems here.
Related to: #44930
r? @ghost
[RFC 2137]: https://github.com/rust-lang/rfcs/blob/master/text/2137-variadic.md
Function signatures with the `variadic` member set are actually
C-variadic functions. Make this a little more explicit by renaming the
`variadic` boolean value, `c_variadic`.
It's present within `Token::Interpolated` as an optimization, so that if
a nonterminal is converted to a `TokenStream` multiple times, the
first-computed value is saved and reused.
But in practice it's not needed. `interpolated_to_tokenstream()` is a
cold function: it's only called a few dozen times while compiling rustc
itself, and a few hundred times across the entire `rustc-perf` suite.
Furthermore, when it is called, it is almost always the first
conversion, so no benefit is gained from it.
So this commit removes `LazyTokenStream`, along with the now-unnecessary
`Token::interpolated()`.
As well as a significant simplification, the removal speeds things up
slightly, mostly due to not having to `drop` the `LazyTokenStream`
instances.
It is currently a method of `Token`, but it only is valid to call if
`self` is a `Token::Interpolated`. This commit eliminates the
possibility of misuse by changing it to an associated function that
takes a `Nonterminal`, which also simplifies the call sites.
This requires splitting out a new function, `nonterminal_to_string`.
Because it's an extra type layer that doesn't really help; in a couple
of places it actively gets in the way, and overall removing it makes the
code nicer. It does, however, move `tokenstream::TokenTree` further away
from the `TokenTree` in `quote.rs`.
More importantly, this change reduces the size of `TokenStream` from 48
bytes to 40 bytes on x86-64, which is enough to slightly reduce
instruction counts on numerous benchmarks, the best by 1.5%.
Note that `open_tt` and `close_tt` have gone from being methods on
`Delimited` to associated methods of `TokenTree`.
`CrateRoot` -> `PathRoot`, `::` doesn't necessarily mean crate root now
`SelfValue` -> `SelfLower`, `SelfType` -> `SelfUpper`, both `self` and `Self` can be used in type and value namespaces now
Remove not used `DotEq` token
Currently libproc_macro does not use `DotEq` token.
https://github.com/rust-lang/rust/pull/49545 changed libproc_macro
to not generate `DotEq` token.
`pretty_print` takes a `Token` and `match`es on it. But the particular
`Token` kind is known at each call site, so this commit splits it into
five functions: `pretty_print_eof`, `pretty_print_begin`, etc.
This commit also does likewise with `print`, though there is one
callsite for `print` where the `Token` kind isn't known, so a generic
`print` has to stay (but it now just calls out to the various `print_*`
functions).
`Printer::word` takes a `&str` and converts it into a `String`, which
causes an allocation. But that allocation is rarely necessary, because
`&str` is almost always a `&'static str` or a `String` that won't be
used again.
This commit changes `Token::String` so it holds a `Cow<'static, str>`
instead of a `String`, which avoids a lot of allocations.
Joshua Netterfield reported an ICE when the unused-parentheses lint
triggered around an async block (#54752). In order to compose an
autofixable suggestion, the lint invokes the pretty-printer on the
unnecessarily-parenthesized expression. (One wonders why the lint
doesn't just use `SourceMap::span_to_snippet` instead, to preserve the
formatting of the original source?—but for that, you'd have to ask the
author of 5c9f806d.)
But then the pretty-printer panics when trying to call `<pprust::State
as PrintState>::end` when `State.boxes` is empty. Empirically, the
problem would seem to be solved if we start some "boxes" beforehand in
the `ast::ExprKind::Async` arm of the big match in
`print_expr_outer_attr_style`, exactly like we do in the
immediately-preceding match arm for `ast::ExprKind::Block`—it would
seem pretty ("pretty") reasonable for the pretty-printing of async
blocks to work a lot like the pretty-printing of ordinary non-async
blocks, right??
Of course, it would be shamefully cargo-culty to commit code on the
basis of this kind of mere reasoning-by-analogy (in contrast to
understanding the design of the pretty-printer in such detail that the
correctness of the patch is comprehended with all the lucid certainty
of mathematical proof, rather than being merely surmised by
intuition). But maybe we care more about fixing the bug with high
probability today, than with certainty in some indefinite hypothetical
future? Maybe the effort is worth a fifth of a shirt??
Humbly resolves#54752.
