Impl period newsletter #3

aturon · November 8, 2017, 6:11am

impl period newsletter

The impl period

The “impl period” consists of 37 working groups, all helping push our 2017 roadmap to completion! There’s a global Gitter channel (with IRC bridge), and dedicated channels for each working group.

This effort is intended for people with all experience levels and interests — there really is something for everyone, and the WG leads are putting a ton of work into mentoring. If you’re not sure what working group is right for you, drop by the main chatroom and introduce yourself, and we’ll help you find your way!

Highlights

A lot has happened since the last newsletter! We had a very successful “impl days” sprint at Rust Belt Rust, which included:

nmatsakis and spastorino pairing to get the very first end-to-end piece of NLL working.
acrichto getting the pure-wasm (no emscripten needed) backend cooking.
withoutboats finalizing their design of the “failure” crate; blog announcement coming soon.
aturon worked out a new overall structure for the home page content.
aturon, ag_dubs, steveklabnik and acrichto worked out the core ideas for a new book on async in Rust.
carols10cents hammering away on alternative registries.
arielb1, nmatsakis, and aturon mind-melding on the trait system
aturon and nmatsakis maybe finding a route to sound specialization??
nmatsakis and aturon laying out a sprint to the finish line.
acrichto building a star destroyer.
surely more that I missed (pleae add in the comments!)

But that’s just a small fraction of what’s been happening. See the shoutouts below for a longer (but still not comprehensive!) list.

Where help is most needed

WG-rand

We have a new issue tracker, regarding status of design & implementation

WG-libs-blitz

Reqwest needs more feedback.
Rayon needs more docs.

WG-infra-crater

Always looking for more participants on this vital piece of infrastructure!

Shoutouts

WG-libs-cookbook

Covers 03.10.2017 - 29.10.2017.

WG-libz-blitz

Covers 05.10.2017 - 28.10.2017. BurntSushi/same-file

@BurntSushi towards 1.0

BurntSushi/walkdir

@BurntSushi towards 2.0
@nivkner add example for converting to I/O error

alexcrichton/flate2-rs

@oyvindln Documentation - Mention rust backend & update miniz and flate info
@ovibos Fix typo in Status documentation in mem.rs
@Frommi Rust backend miniz_oxide

carllerche/mio

@carllerche Bump version to v0.6.11
@FuGangqiang fix doc
@jtescher Fix small typos
@alexcrichton Avoid runtime panics in IoVec on Windows
@GabrielMajeri Fix Android CI
@cramertj Sync with Fuchsia

nikomatsakis/rayon

seanmonstar/num_cpus

sebasmagri/env_logger

steveklabnik/semver

@muzikmoe Libz blitz/turn references to other semver items int links
@muzikmoe Libz blitz/use question mark instead of unwrap in examples

WG-libs-guidelines

Covers October 25 — October 29

Brand new guidelines

WG-libs-simd

@alexcrichton for proposing stabilization path and keeping the trains running
@gnzlbg for awesome work on adding intrinsics for ARM

WG-dev-tools-bindgen

0.31.0 release! Check out the release notes and list of everyone who helped make it happen here: https://users.rust-lang.org/t/bindgen-automatically-generate-rust-ffi-bindings-to-c-and-c-libraries/12126/4
- (And a couple point releases since then with minor bug fixes)
@fitzgen removed the dependency on the unmaintained syntex [crate]
- and build times were cut in half! ~90 seconds to ~45 seconds!
@fitzgen made it so that our csmith fuzzer will automatically run creduce on test cases that uncover bugs in bindgen
@fitzgen fixed translation of constant integers that were greater than u32::MAX
@fitzgen made determining which structs need an explicit virtual table pointer into a fixed-point analysis
@pepyakin enabled emitting derive(Clone) when targetting Rust >= 1.21, now that Clone is a builtin item
@jrmuizel extended our IR to keep track of functions’ and methods’ linkage; one day we hope to emit C trampolines for inlined functions and methods
@seemyvest improved our error messages and handling when given unreadable files or folders as input headers
@photoszzt fixed a bug where bindgen would erroneously emit derive(Hash) for flexibly sized arrays
@hallfox added the ability to skip deriving Clone and Copy for specific types
@fitzgen made bindgen's type sizedness computation into a fixed-point algorithm
@seemyvest added the ability to skip deriving Hash for specific types
@manaskarekar made it so that bindgen will rustfmt its generated bindings not only when writing them into a file, but also when printing to stdout
@emilio fixed a bug where C++ references were getting template type parameters that they shouldn’t
@photoszzt fixed a bug in the impl ::std::fmt::Debug for … blocks that bindgen emits where an unmangled name was used when it should have been mangled

WG-compiler-incr

We have seen some incredible contributions to incremental compilation since the last newsletter:

@vitiral has given the testing framework for incremental compilation hashes a comprehensive revamp which makes writing the respective tests much more concise and involved updating a ton of existing tests!
@theotherjimmy has continued to power through multiple refactorings and improvements, like adding the ensure() primitive to the query system, turning vtable computation into a query, and cleaning up some of the core functionality of the query system.
@wesleywiser has finished their work of turning trans-time trait selection into a query and then gone on to implement “eval-always queries”, a new kind of query that optimizes certain workload patterns.
@SimonSapin has adapted legacy dependency tracking code in variance computation to the new red-green system and volunteered to add and maintain two large test cases from the Servo project on http://perf.rust-lang.org.
@Xanewok has cleaned up a long-standing FIXME around crate disambiguators and has brought some of the stable infrastructure from incr. comp. to the RLS.
@alexcrichton has adapted the unused-mut-nodes pass to the query system and made part of the compilation pipeline more multi-threaded which will help ThinLTO as well as incr. comp.
@michaelwoerister has fixed some bugs in the incr. comp. hashing code, added some documentation to the important force_from_dep_node() procedure, re-implemented -Zincremental-info after red-green tracking had broken it, switched incr. comp. hashing from BLAKE2 to 128 bit SipHash for increased performance, fixed some bugs in the HIR hashing code, updated some of the diagnostics framework to avoid emitting duplicate error messages during incr. comp., and done some preliminary work for caching MIR and type-information.
And finally, @aturon has put out a call to help benchmark incremental compilation which has received an awesome amount of feedback so far!

WG-compiler-front and WG-compiler-middle

Completed:

dyn Trait syntax is available on nightly thanks to @petrochenkov
the crate keyword can now be used to describe visibility of an item, thanks to @zackmdavis
we now have an (opt-in, for now) unreachable-pub lint for detecting things declared as pub but no reachable from outside the crate, thanks to @zackmdavis
we now permit '``_ to explicitly elide a lifetime, thanks to @cramertj
default match bindings are available on nightly thanks to @tschottdorf
- also thanks to @tamird for advertising them better when an error occurs
we now have #[non_exhaustive] working on nightly, thanks to @davidtwco
@sunjay landed a number of refactorings towards Generic Associated Types

In progress:

@pietroalbini is working on use import groups (use foo::{a, bar::{b, c}})
@slo1 has been working on supporting foo.rs modules that permit foo/bar.rs submodules
@chrisvittal is getting very close to supporting impl Trait in argument position (fn foo(x: impl Iterator<Item = u32>))
@sunjay has been working on extending the parser to support Generic Associated Types
@qwz is pushing on permitting the elision of lifetimes in impl headers
@cramertj has been making progress towards the newer semantics for impl Trait in return position
@durka has been working on trait aliases

WG-compiler-nll