Subteam reports 2015-09-21

aturon · September 21, 2015, 6:41pm

Tools

Some of the highlights from the past two weeks:

MSVC backtraces now have line numbers where possible.
LLVM has been updated to the 3.7 release branch.
OpenBSD has a snapshot and a few fixes so it’s almost passing tests and "vanilla bootstrapping".
1.3 was released, be sure to thank @brson for his awesome work in preparing all the artifacts!
There’s a thread on internals about packaging Rust and Cargo with lots of good discussion.
A new RFC has been proposed for adding machine readable output to the test harness that rustc generates.

Libs

Last week marked the 1.3 stable/1.4 beta release. As part of this cycle, we went through a round of API stabilization and deprecation for the 1.4 release (after a cycle’s worth of collecting final comments). You can see the outcome here.

We will be selecting the next batch of APIs to put up for stabilization or deprecation (for the 1.5 release) this week. Please take a look at the "stabilization phase" dashboard below, and leave a comment if you’d like to nominate an API for this process.

On the RFC front, we’ve been continuing to churn on some controversial points. The wildcard dependencies RFC remains open for another round of final comments:

FCP PR #1241: Forbid wildcard dependencies on crates.io

In addition, there’s been substantial discussion about catch_panic (now recover) since it was merged, including new alternative proposals for more sophisticated bounds; check out the latest comments here.

Finally, a long-awaited RFC for building out our std::time APIs has been posted:

PR #1288: Improvements to the Time APIs

Please leave feedback on those RFCs to help drive the discussion forward!

Lang

Full report

We decided to accept a number of RFCs:

RFC #1199: SIMD groundwork.
RFC #1224: RFC process.
RFC #1237: Clarify extreme operator behavior.
RFC #1238: Non-parametric dropck.
RFC #1240: References into packed structs should be unsafe.

We decided to place the following RFCs into final comment period:

RFC #1228: Change the "placement new" syntax from in place { value } to place <- value. The discussion has centered about what uses for “placement new” will be most common; this syntax seems well-suited to uses that do not yield a result, such as vec <- elem, and less well-suited to those that do, such as let value = (arena <- value).
RFC #1245: Allow const fn functions to be unsafe.

Thanks!

–nmatsakis

Compiler