Ben Kimock

@bors try @rust-timer queue

@craterbot run mode=build-and-test

@craterbot abort

@craterbot run mode=build-and-test start=master#61223975d46f794466efa832bc7562b9707ecc46+rustflags=-Copt-level=3 end=try#d073071d77ce0f93b4fd8cc567a1e2b9e1b22126+rustflags=-Copt-level=3

@craterbot run mode=build-and-test start=master#61223975d46f794466efa832bc7562b9707ecc46+rustflags=-Copt-level=3 end=try#d073071d77ce0f93b4fd8cc567a1e2b9e1b22126+rustflags=-Copt-level=3 p=1 crates=https://crater-reports.s3.amazonaws.com/pr-121282/retry-regressed-list.txt

All the build-fail results in the re-run are from the MIR inliner's dependency on crate hash. I don't know if that makes the crates qualify as flaky or rustc. Only...

Here's a paste of my notes, I'll update this later: ``` Regressions (I will file issues): reg/faktory/0.12.1 # the int -> ptr transmute is in a dependency, atomic-option: https://github.com/reem/rust-atomic-option/pull/5 reg/gazebo/0.8.1...

The LLVM LangRef documents this about coldcc: > Furthermore the inliner doesn’t consider such function calls for inlining. So I don't think that does the post-mono-only inlining that I described.

@bors try @rust-timer queue

As the author of many of std's UB-detecting debug assertions and the pointer alignment checks, I've thought about this problem a lot. > if the debug assertion triggers only as...