Libs Meetings
Minutes
Meeting Link: https://meet.jit.si/rust-libs-meeting-ujepnbwg2lzqgt6wvrndwimi
Attendees: Amanieu, The 8472, Mara, joboet
1 rust-lang/rust T-libs FCPs
This 'technically UB' doesn't really result in actual issues, it seems.
Not a recent regression, so might not be a good backport candidate.
Left a comment.
Discussed last time. Still need to respond. Mara to respond.
Discussed last time. Amanieu replied, saying the main issue is thread::current(), which is only used in a few places.
For LocalKey::with we already document it: https://doc.rust-lang.org/stable/std/thread/struct.LocalKey.html#method.with
(And also have a try_with.)
Only for thread::current() (and a few APIs that use that) we don't provide any panic docs (or a try_current() etc.).
Pre-main seems like it's rarely a problem, it's post-main that is more interesting.
Summary: https://github.com/rust-lang/rust/pull/111071#issuecomment-1576734594
Question 1: Change the default panic handler output format?
Possible answers:
3 seems like a nice option, because the current "at {message}, {location}" feels flipped/swapped.
Crater run could check how widely people depend on the exact output.
We can do an FCP to make it show up in TWiR.
Split this into a separate PR to focus first on just the panic_handler output, independent from assert_eq.
Question 2: Change the panic message of assert_eq?
Leave this for after the PR for question 1.
Discussed and responded last time. FCP close in progress. Removed label.
P-high
`+outline-atomics` violates "core shall not depend on libc"Resolutions: Either:
The 'can core depend on libc' discussion is a complicated one, probably needs an RFC for a resolution.
Amanieu to write up that RFC.
Might have already been broken but simply not resulted in a panic.
Tier 3, so not high priority.