2015-08-10

Agenda

(Had a very full agenda; did not cover all of these topics.)

Non-zeroing move

nmatsakis: Felix landed a change, but this was shown to introduce leaks. He had a fix for that, but it's not valgrind clean (unclear if it's actually a new problem). So we disabled non-zeroing move for the time being; that needs to be backported to beta. Felix is on vacation this week, but hopes to get to the bottom of it when he comes back.
nmatsakis: Not fatal to the approach, just a hiccup.
PR: rust-lang/rust#27582

aturon: We narrowly missed the 1.3 beta train for stabilizing Duration. Would like to make an exception and backport the stabilization.
Some discussion about how to avoid this in the future. E.g. holding a go/no go meeting for stabilization >= 1 week before releases are cut.

Conference went very well!
Significant number of non-systems programmers, which was encouraging.
Logistics went blindingly well (people reported no obvious snags at all, despite everything being last-minute)
Leah gave her perspective

Codegen units is one possibility
It is possible to pass -O1 down from cargo through to rustc
YK: "Static check only" might be a better "fast check" default than -O0?
YK: cargo build --check using a metadata-only build might be better for the fastest build, and -O1 might be a better default for cargo period.
One thing that Servo wants is the ability to optimize particular crates even when doing a debug build.
brson: Would prefer to do this per-function.
wycats: But then the lib author has to annotate; this should be an application-level decision