2025-06-09 — Last Week in Carbon

[josh11b]

Updates on leads issues

• Decisions
  • (none)
• New issues
  • #5611 Should impls with maximal type structure be allowed to be declared final?
    • #generics-and-templates Allowing more impls to be final than our current rules allow.
• Active discussions
  • (none)

Updates on proposals

• Decisions
  • (none)
• New RFC proposals
  • #5606 Keep design documents current
    • #evolution adding TODOs to the design docs as part of a proposal that would change them
• Active discussions
  • #5545 Expression form basics
    • Several rounds of review, ready for leads

Updates on moderation across our community spaces (Discord and GitHub)

• 10 trained moderation & conduct team members + community lead, in theory available from UTC-7 to UTC+5.5.
  • WIP: Talking community journey with Zig folks.
  • WIP: resuming conversation about how we can be more intentional about providing good first issues for newcomers
• Last week’s stats across our spaces:
  • In the last 7 days in our repo: 8 authors pushed 35 commits to trunk, we changed 523 files with 9,778 additions and 10,939 deletions, with 36 merged PRs, 12 open PRs, and 3 new issues.

Activity on toolchain

• 06-03 #5601 Don't keep insts in eval block from failed deduce
• 06-05 Ids and value stores
• 06-05 Working around a problem with the current implicit conversion approach
• #toolchain #5607 #5625 removing min_prelude and no_prelude directories
• #toolchain syntax highlighting on GitHub
• #toolchain types accessors
• #toolchain InstKind and typed insts
• #toolchain Avoiding copies by delaying construction
• #toolchain Clarity or consistency with lit in file tests?
• #generics-and-templates FacetTypes with symbolic .Self in rewrite constraints are still considered concrete, which complicates performing Subst on them.
• #generics-and-templates Fix #5603 A crash due to LookupImplWitness failing to be resolved. The phase of the type was not being updated when constructing a FacetType that had symbolic (other than .Self) values in its rewrite constraints.
• #generics-and-templates An example where the facet type and an impl that can be found from the facet disagree about the value of an associated constant, and if/how it should be diagnosed. A test was added to #5617.
• #generics-and-templates Discussion of implementation challenge for replacing values in facet types for impl lookup comparison conflicting with cycle detection.
• PRs:
  • tests in #5578 #5547 #5549 #5540 #5602 #5596 #5612
  • lowering specifics for imported generics in #5583 #5475
  • value stores in #5576 #5589 #5626 #5628 #5621
  • #toolchain diagnostics in #5590 #5591 #5592
  • #toolchain dump-sem-ir-ranges in #5587 #5600 #5598 #5624
  • refactor #5585
  • crash fixes #5586 #5604 #5603
  • dependencies in #5605 #5593 #5622
  • test infra in #5595 #5610 #5623
  • generics in #5597
  • builtin & in #5613
  • import in #5616 #5618

Activity on C++ interop

• 06-03 Progress on in flight Clang PR #5543.
• 06-03 Discussion on SemIR knowledge sharing. Proposed ideas: Recorded talk, something similar to LLVM verifier, external documentation of SemIR, better typed inst documentation.

Other active discussions

These include Discord Chat, GitHub Discussions, or open discussion sessions.

• 06-03 Decided to use the same form types for parameters and returns
  • #syntax 06-04 06-05 Meant updating form literal syntax
• 06-03 06-04 Compile-time functions (like C++'s constexpr and consteval)
  • Expect to introduce another phase for values computed at compile time, but not available to type checking
  • Open question on how to handle side effects (memory allocation, diagnostics, logging) in functions that we are evaluating at compile time
• 06-03 Template constraint predicates
• 06-03 New definition of alias using forms, separate syntax for namespace aliases
• #standard-library Question about how to pass a slice to a function, but this is not designed or implemented yet.
• #infra Notes that PyLance extension in VSCode may cause performance problems.
• #general-design An example given of lack of progressive disclosure in java: public static void main, where you need to learn a lot of keywords to write hello world.
• #general-design Question about where member binding (from p3720) fits into the steps of member access. Answer that it replaces instance binding.
• #style Discussion on the use of auto in the toolchain implementation. General agreement to continue and document that we allow use of auto for “almost everything” except primitive types in local variables, without requiring it.
• #interop Imported C++ entities were not being stored correctly, so were not appear in textual semir correctly; fixed in #5616.
• #safety A paper on Spegion, an effect-based model for region based memory safety that does not rely on lexical relationships, and can be applied to C/C++ code without explicit lifetime annotations in order to detect use-after-free statically.
• #lambdas The “generic across forms” proposal is going to include deducing the return form for the return of a lambda written with =>.
• #interop How to store pointers into the Clang AST in the Carbon toolchain’s instructions and/or ValueStores. CanonicalValueStore suggested to help deduplicate pointers.
• #generics-and-templates Confirm that impl lookup in a generic function can’t use types defined in the lookup self type as concrete values in the function.
• #generics-and-templates We have renamed “template constant” to “concrete constant” in our terminology in the toolchain, though it hasn’t gone through a proposal.
• #generics-and-templates An example of deducing a generic binding that is used in a facet’s rewrite constraint value. Currently it does not work as deduce does not look into rewrite constraints, but it could.
• #generics-and-templates Discussion of implementing cycle detection, and why substitution will be required for cycle detection.

RSS Feed for "Last Week in Carbon" here