T-lang meeting agenda

Attendance

  • Team members: nikomatsakis, scottmcm, pnkfelix
  • Others:

Pending proposals

  • Discuss policy proposals
    • Switch to compiler-team process: meta-design-meeting once a month
    • At start of that meeting, review active project groups; this also allows for them to be added to one of the month's meetings
    • Handling blocking concerns, and requiring a "second" for them. (Does not indicate agreement with the concern, but indicates desire to hear more about the concern.)
    • Will handle these via FCP on a lang-team issue
  • "MCP: Deref Patterns" lang-team#77

Project board review

Review the Lang team project board and get updates

Nominated RFCs

  • No nominated RFCs this time.

P-high issues on rust-lang/rust

  • "regression 1.50: deny after forbid breaks build" rust#80988
  • "beta regression: forbid(warnings) interacts poorly with common derives." rust#81218
    • Fix is in PR #81556
      • the effect is that if you have a forbid(group) and later do a allow(lint) where lint is in group, you get a FCW but the allow takes effect
    • Notes for future:
      • If we convert a lint into a group, it should be exempted
      • We should move swiftly to make this a hard error as there wasn't much breakage
        • As soon as we address derive triggering it

Nominated PRs and issues on rust-lang/rust

  • "Tracking issue for RFC 2535, 2530, 2175, "Or patterns, i.e Foo(Bar(x) | Baz(x))"" rust#54883
  • "Lint for unused borrows as part of UNUSED_MUST_USE" rust#76894
    • Pending action item: nikomatsakis
  • "CString lifetime warning incorrectly fires" rust#78691
    • Needs the lang team to make a call on this.
    • False warning on lint for this line
      • gecko_profiler_register_thread(CString::new("foo").unwrap().as_ptr());
      • the CString::new() is being converted to a raw pointer; that raw pointer is valid until end of statement
    • Whether this is a false positive depends on what gecko_profiler_register_thread
    • What is expected rewrite?
      • Put CString::new into a let, but that is also quite possibly wrong (e.g., )
    • Lint on:
      • let x = CString::new("foo").unwrap().as_ptr(); – good, guaranteed wrong
      • let x = Some(CString::new("foo").unwrap().as_ptr());
      • let x = Foo { field: CString::new("foo").unwrap().as_ptr() };
      • let x = (CString::new("foo").unwrap().as_ptr(), ...)
    • Not lint on:
      • let x = foo(CString::new("foo").unwrap().as_ptr());
      • let x = foo(Some(CString::new("foo").unwrap().as_ptr()));
      • let x = foo(x.as_ptr()) where foo is the identity function
    • Resolution: josh to comment

Current output:

warning: getting the inner pointer of a temporary `CString`
 --> src/main.rs:9:69
  |
9 |         gecko_profiler_register_thread(CString::new("foo").unwrap().as_ptr());
  |                                        ---------------------------- ^^^^^^ this pointer will be invalid
  |                                        |
  |                                        this `CString` is deallocated at the end of the statement, bind it to a variable to extend its lifetime
  |
  = note: `#[warn(temporary_cstring_as_ptr)]` on by default
  = note: pointers do not have a lifetime; when calling `as_ptr` the `CString` will be deallocated at the end of the statement because nothing is referencing it as far as the type system is concerned
  = help: for more information, see https://doc.rust-lang.org/reference/destructors.html
  • "expand/resolve: Turn #[derive] into a regular macro attribute" rust#79078
    • In FCP for 3 more days, no need to review. Can merge after FCP and after petrochenkov fixes merge conflicts
  • "Stabilize unsafe_op_in_unsafe_fn lint" rust#79208
    • Nominated for discussion, would default to allow
    • Deferring the question of whether to change the default/recommended style
    • Action item: nikomatsakis to request stabilization report
  • ""!" is the only type that is unsafe to raw-ptr-deref into a "_" pattern" rust#79735
  • "Unsafe checking skips pointer dereferences in unused places" rust#80059
    • Observation:
      • initialization is inherently a control-flow based analysis, which implies
        • panic!(); *foo, the *foo is dead code, it's hard to say whether (we care that) foo is initialized or not
      • let foo: bool; let _ = *foo; doesn't involve dead code
    • Consensus:
  • "Promote 2018 warn lints to deny in Rust 2021" rust#80165
    • Conclusion from previous week: we will break these out into separate PRs
    • Action item: Ryan to create those PRs
    • Question: should we warn on these lints in prior editions?
      • Action item: Niko to write up pros/cons
  • "make const_err a future incompat lint" rust#80394
    • Ralf's grand plan to make const-err less painful
    • This step: makes it a future incompatibility lint to have constants that cannot be evaluated
    • Mark's summary
    • Josh to leave comment saying "go for it"
      • Done
  • "relax adt unsizing requirements" rust#80726
    • lcnr added a write-up
    • struct Foo<T: ?Sized> { x: T }
    • Arc<Foo<[u8; 32]>> want to permit coercion to Arc<Foo<[u8]>>
      • Arc<T> can be coerced to Arc<U> if T: Unsize<U> or something like that
    • struct Foo2<T: ?Sized> { a: T, x: T }
      • Foo2<[u8]> would affect type of both fields
    • struct A<T, U: ?Sized>(T, B<T, U>);
    • struct B<T, U: ?Sized>(T, U);
    • Propose feature-gating – on principle, to give time
    • Felix to leave comment
  • "Visit more targets when validating attributes" rust#80920
    • Validating more cases where we were failing to check for unexpected attributes
    • e.g., #[inline] const T: u32, these were being silently ignored before
    • We now get an error, breaking change, crater run, 3 crates affected
    • Affected crates:
      • olm-rs-1.0.0 – real regression on the latest version
    • Conclusion:
      • Move forward (we may want to revisit our policy but not on this instance)
      • Ping authors of olm-rs crate (and perhaps other affected crates) – done
    • Action item:
  • "don't gratuitously error on tests returning Result with lifetime" rust#80934
  • "Stabilize remaining integer methods as const fn" rust#80962
  • "beta regression: forbid(warnings) interacts poorly with common derives." rust#81218
    • Already addressed above in P-high issues
  • "Allow specifying alignment for functions" rust#81234
  • "Support forwarding caller location through trait object method call" rust#81360
  • "Allow casting mut array ref to mut ptr" rust#81479
Select a repo