proc_macro_plugin: Wrap nonexistent filename in <>
I'm not sure how big of an issue this can become in practice, but `FileMap`s made from something that's not a file are supposed to wrap the file name in `<>`.
For an example fix, see kevinmehall/rust-peg@332fd4dbae. There, it caused cargo to always recompile a crate using rust-peg, even when nothing was changed, because cargo sees that the dummy file doesn't exist.
Fix regression involving custom derives on items with `$crate`
The regression was introduced in #37213.
I believe we cannot make the improvements from #37213 work with the current custom derive setup (c.f. https://github.com/rust-lang/rust/issues/37637#issuecomment-258959145) -- we'll have to wait for `TokenStream`'s API to improve.
Fixes#37637.
r? @nrc
macros: improve reexports
This PR
- avoids building multiple module graphs for a crate that is referenced by multiple `extern crate` items,
- registers `#[no_link] extern crate`s to avoid loading the same crate metadata twice,
- stability checks `#[no_link] extern crate`s,
- [breaking-chage]: `#[no_link] #[macro_use] extern crate syntax;` is allowed on stable today
- fixes `$crate` in `#[macro_reexport]`ed macros,
- [breaking-change] for `#[feature(macro_reexport)]` (technically)
- allows selective macro importing (i.e. `#[macro_use(foo, bar)]`) from custom derive crates, and
- refactors the crate metadata to support re-exported macros in arbitrary modules (not yet needed).
r? @nrc
Document the question mark operator in reference and the book's syntax index
The question mark operator will be stabilized for the Rust 1.13 release (unfortunately). Even though I don't like the operator, it still should be documented in the syntax index in the book and in the reference.
Maybe there are people who also want to change the book's chapters on error handling, depending on their views of what idiomatic error handling is, now that the operator is stable, but I don't want to and I'd prefer to keep this PR focused on the reference and syntax index only.
Please also apply this PR to the beta branch of rust.
[6/n] rustc: transition HIR function bodies from Block to Expr.
_This is part of a series ([prev](https://github.com/rust-lang/rust/pull/37408) | [next](https://github.com/rust-lang/rust/pull/37676)) of patches designed to rework rustc into an out-of-order on-demand pipeline model for both better feature support (e.g. [MIR-based](https://github.com/solson/miri) early constant evaluation) and incremental execution of compiler passes (e.g. type-checking), with beneficial consequences to IDE support as well.
If any motivation is unclear, please ask for additional PR description clarifications or code comments._
<hr>
The main change here is that functions and closures both use `Expr` instead of `Block` for their bodies.
For closures this actually allows a honest representation of brace-less closure bodies, e.g. `|x| x + 1` is now distinguishable from `|x| { x + 1 }`, therefore this PR is `[syntax-breaking]` (cc @Manishearth).
Using `Expr` allows more logic to be shared between constant bodies and function bodies, with some small such changes already part of this PR, and eventually easing #35078 and per-body type tables.
Incidentally, there used to be some corners cut here and there and as such I had to (re)write divergence tracking for type-checking so that it is capable of understanding basic structured control-flow:
``` rust
fn a(x: bool) -> i32 {
// match also works (as long as all arms diverge)
if x { panic!("true") } else { return 1; }
0 // "unreachable expression" after this PR
}
```
And since liveness' "not all control paths return a value" moved to type-checking we can have nice things:
``` rust
// before & after:
fn b() -> i32 { 0; } // help: consider removing this semicolon
// only after this PR
fn c() -> i32 { { 0; } } // help: consider removing this semicolon
fn d() { let x: i32 = { 0; }; } // help: consider removing this semicolon
fn e() { f({ 0; }); } // help: consider removing this semicolon
```
[3/n] rustc: unify and simplify managing associated items.
_This is part of a series ([prev](https://github.com/rust-lang/rust/pull/37401) | [next](https://github.com/rust-lang/rust/pull/37404)) of patches designed to rework rustc into an out-of-order on-demand pipeline model for both better feature support (e.g. [MIR-based](https://github.com/solson/miri) early constant evaluation) and incremental execution of compiler passes (e.g. type-checking), with beneficial consequences to IDE support as well.
If any motivation is unclear, please ask for additional PR description clarifications or code comments._
<hr>
`ImplOrTraitItem`/`impl_or_trait_item` have been renamed to `AssociatedItem`/`associated_item`.
The common fields from (what used to be) `ty::ImplOrTraitItem`'s variants have been pulled out, leaving only an `AssociatedKind` C-like enum to distinguish between methods, constants and types.
The type information has been removed from `AssociatedItem`, and as such the latter can now be computed on-demand from the local HIR map, i.e. an extern-crate-enabled `TraitItem | ImplItem`.
It may be moved to HIR in the future, if we intend to start using HIR types cross-crate.
`ty::ExplicitSelfCategory` has been moved to `rustc_typeck` and is produced on-demand from the signature of the method, and a `method_has_self_argument` field on `AssociatedItem`, which is used to indicate that the first argument is a sugary "method receiver" and as such, method call syntax can be used.
_match: correct max_slice_length logic
The logic used to be wildly wrong, but before the HAIR patch its wrongness was in most cases hidden by another bug.
Fixes#37598.
r? @nikomatsakis
Marking the 'no-stack-check' codegen option as deprecated (Issue #34915)
Attempts to finish resolving issue #34915. Based on pull request #35156, which was closed due to inactivity.
macros 1.1: Allow proc_macro functions to declare attributes to be mark as used
This PR allows proc macro functions to declare attribute names that should be marked as used when attached to the deriving item. There are a few questions for this PR.
- Currently this uses a separate attribute named `#[proc_macro_attributes(..)]`, is this the best choice?
- In order to make this work, the `check_attribute` function had to be modified to not error on attributes marked as used. This is a pretty large change in semantics, is there a better way to do this?
- I've got a few clones where I don't know if I need them (like turning `item` into a `TokenStream`), can these be avoided?
- Is switching to `MultiItemDecorator` the right thing here?
Also fixes https://github.com/rust-lang/rust/issues/37563.
Vendor all rustbuild dependencies in this repo
This commit vendors all crates.io dependencies into the rust-lang/rust repository using the `cargo-vendor` tool. This is done in an effort to make rustbuild distro-ready by ensuring that our source tarballs are self-contained units which don't need extraneous network downloads.
A new `src/vendor` directory is created with all vendored crates, and Cargo, when using rustbuild, is configured to use this directory. Over time we can deduplicate this directory with the actual src tree (e.g. src/librustc_serialize, src/liblibc, src/libgetopts, ...). For now though that's left to a separate commit.
Document convention for using both fmt::Write and io::Write
Using a trait's methods (like `Write::write_fmt` as used in `writeln!` and other macros) requires importing that trait directly (not just the module containing it). Both `fmt::Write` and `io::Write` provide compatible `Write::write_fmt` methods, and code can use `writeln!` and other macros on both an object implementing `fmt::Write` (such as a `String`) and an object implementing `io::Write` (such as `Stderr`). However, importing both `Write` traits produces an error due to the name conflict.
The convention I've seen renames both of them on import, to `FmtWrite` and `IoWrite` respectively. Document that convention in the Rust documentation for `write!` and `writeln!`, with examples.
Point to type argument span when used as trait
Given the following code:
``` rust
struct Foo<T: Clone>(T);
use std::ops::Add;
impl<T: Clone, Add> Add for Foo<T> {
type Output = usize;
fn add(self, rhs: Self) -> Self::Output {
unimplemented!();
}
}
```
present the following output:
``` nocode
error[E0404]: `Add` is not a trait
--> file3.rs:5:21
|
5 | impl<T: Clone, Add> Add for Okok<T> {
| --- ^^^ expected trait, found type parameter
| |
| type parameter defined here
```
Fixes#35987.
Include type of missing trait methods in error
Provide either a span pointing to the original definition of missing
trait items, or a message with the inferred definitions.
Fixes#24626. Follow up to PR #36371.
If PR #37369 lands, missing trait items that present a multiline span will be able to show the entirety of the item definition on the error itself, instead of just the first line.
Replace FNV with a faster hash function.
Hash table lookups are very hot in rustc profiles and the time taken within `FnvHash` itself is a big part of that. Although FNV is a simple hash, it processes its input one byte at a time. In contrast, Firefox has a homespun hash function that is also simple but works on multiple bytes at a time. So I tried it out and the results are compelling:
```
futures-rs-test 4.326s vs 4.212s --> 1.027x faster (variance: 1.001x, 1.007x)
helloworld 0.233s vs 0.232s --> 1.004x faster (variance: 1.037x, 1.016x)
html5ever-2016- 5.397s vs 5.210s --> 1.036x faster (variance: 1.009x, 1.006x)
hyper.0.5.0 5.018s vs 4.905s --> 1.023x faster (variance: 1.007x, 1.006x)
inflate-0.1.0 4.889s vs 4.872s --> 1.004x faster (variance: 1.012x, 1.007x)
issue-32062-equ 0.347s vs 0.335s --> 1.035x faster (variance: 1.033x, 1.019x)
issue-32278-big 1.717s vs 1.622s --> 1.059x faster (variance: 1.027x, 1.028x)
jld-day15-parse 1.537s vs 1.459s --> 1.054x faster (variance: 1.005x, 1.003x)
piston-image-0. 11.863s vs 11.482s --> 1.033x faster (variance: 1.060x, 1.002x)
regex.0.1.30 2.517s vs 2.453s --> 1.026x faster (variance: 1.011x, 1.013x)
rust-encoding-0 2.080s vs 2.047s --> 1.016x faster (variance: 1.005x, 1.005x)
syntex-0.42.2 32.268s vs 31.275s --> 1.032x faster (variance: 1.014x, 1.022x)
syntex-0.42.2-i 17.629s vs 16.559s --> 1.065x faster (variance: 1.013x, 1.021x)
```
(That's a stage1 compiler doing debug builds. Results for a stage2 compiler are similar.)
The attached commit is not in a state suitable for landing because I changed the implementation of FnvHasher without changing its name (because that would have required touching many lines in the compiler). Nonetheless, it is a good place to start discussions.
Profiles show very clearly that this new hash function is a lot faster to compute than FNV. The quality of the new hash function is less clear -- it seems to do better in some cases and worse in others (judging by the number of instructions executed in `Hash{Map,Set}::get`).
CC @brson, @arthurprs
