Libs-API Meeting 2022-10-04

# Libs-API Meeting 2022-10-04 ###### tags: `Libs Meetings` `Minutes` **Meeting Link**: https://meet.jit.si/rust-libs-meeting-crxoz2at8hiccp7b3ixf89qgxfymlbwr **Attendees**: Amanieu, David, Josh Triplett, Mara, Thom, Chris, Urgau ## Agenda - Triage - Anything else? ## Triage ### FCPs 21 rust-lang/rust T-libs-api FCPs [joshtriplett (6)](https://rfcbot.rs/fcp/joshtriplett), [dtolnay (4)](https://rfcbot.rs/fcp/dtolnay), [Amanieu (8)](https://rfcbot.rs/fcp/Amanieu), [m-ou-se (9)](https://rfcbot.rs/fcp/m-ou-se), [BurntSushi (14)](https://rfcbot.rs/fcp/BurntSushi), [yaahc (15)](https://rfcbot.rs/fcp/yaahc) ### Nominated - [6 `rust-lang/rust` `I-libs-api-nominated` items](https://github.com/rust-lang/rust/issues?q=label:I-libs-api-nominated+is:open) - rust.tf/99939 *Sort tests at compile time, not at startup* - Just needs to be reviewed. Doesn't need libs-api approval. - rust.tf/100021 *made \`unreachable\_unchecked\` panic in debug mode* - Seems good, but shouldn't unwind. - Left a comment. - rust.tf/101698 *Constify \`TypeId\` ordering impls* - Seems like a hash or similar will always suffice for this. Which means this change is fine. - Nominated for compiler team to check. - rust.tf/101899 *#95295 causes unsoundness in multiple existing crates* - Amanieu: Letting a size greater than isize::MAX reach the global allocator is already UB, because of LLVM optimizations that understand allocations. - "library UB" vs "language UB"? Should Layout have a niche? - Is there a reason to have a niche? - LLVM could remove checks, if it knows that it's always < isize::MAX. - Problem is that we're changing the requirements for from_size_align_unchecked. - Adding a check to it for the new requirement would 'fix' the issue, but not desirable because it defeats the purpose of the unchecked version. - We could only check it in debug mode, by purposely causing an overflow (with inherit-overflow-checks) on >= isize::MAX. Doesn't fix anything for release mode though. - Alternative is to deprecate the existing function, add an assert in it, and have a new unchecked function with the new requirement. - Mara to comment. - rust.tf/101991 *Stabilize proc\_macro Span::source\_text* - Should it fail on any joined span? Or only for extreme cases like in the example. - Hard to answer. We should check what people do with joined spans / source_text. - The assert2 crate would use this with a joined span, to get the span of the entire expression. - We probably need some rule like 'it only works for tokens from the macro invocation itself'. - We consider the extreme case a bug that should be fixed, without breaking the assert2 case (stringifying an expression). - Is it a blocking bug? - Josh: Yes? - David, Mara: No - Of all the weird things people can do in macros, this one is very far down the list. - David to reply / resolve concerns. - rust.tf/102225 *Allow const iterator implementations* - This would mean committing to never add a non-const method to the Iterator trait. - Maybe partially (non-)const traits should just be a thing. (Just like `where Self: Sized` on methods for exclusion from trait objects.) - How const traits work is still experimental, so it's hard for us to make decisions around it, since it's not clear yet what it means. - As long as this has no stable effect, it's fine to merge. But these questions should be tracked for const trait stabilizations. - That's already tracked using `#[rustc_const_unstable]`. - [1 `rust-lang/rfcs` `I-libs-api-nominated` items](https://github.com/rust-lang/rfcs/issues?q=label:I-libs-api-nominated+is:open) - rust.tf/rfc3271 *RFC: Add more support for fallible allocations in Vec* - `try_` prefix already used on e.g. `try_insert` for maps, meaning something else. `try_push` could mean multiple things (try within capacity (without allocating?) - Mara: Feels like a similar issue as with lock poisoning: having two interfaces: one for manual error handling, one for just panicking/ignoring the issue (without verbosity like unwrap everywhere). Rust doesn't have the right tools to express this nicely. :( ### Waiting on team - [1 `rust-lang/rust` `T-libs-api` `S-waiting-on-team` items](https://github.com/rust-lang/rust/issues?q=label:T-libs-api+label:S-waiting-on-team+is:open) - rust.tf/99032 *Disable unwinding for \`catch\_unwind\` error payloads.* ### Needs decision None ### Stalled Tracking Issues - [30 `rust-lang/rust` `T-libs-api` `C-tracking-issue` items](https://github.com/rust-lang/rust/issues?q=label:T-libs-api+label:C-tracking-issue+is:open+sort:updated-asc) - rust.tf/79676 *Tracking Issue for \`backtrace\_frames\`* - rust.tf/42839 *Peer credentials on Unix socket* - rust.tf/55132 *Tracking issue for thread::Builder::spawn\_unchecked* - rust.tf/66893 *Tracking issue for atomic\_mut\_ptr* - rust.tf/72505 *Tracking issue for \`#!\[feature(const\_float\_classify)\]\`* ## Actions - [ ] Reply to all issues/PRs discussed in this meeting, or add them to the [open action items](https://hackmd.io/ovrbJj6CRduRgSA0Wzg2zg). _Generated by [fully-automatic-rust-libs-team-triage-meeting-agenda-generator](https://github.com/rust-lang/libs-team/tree/main/tools/agenda-generator)_