2025-06-30 — Last Week in Carbon

[josh11b]

Updates on leads issues

• Decisions
  • #5522 Immutable value semantic returns questions from ref parameters, arguments, returns and val returns #5434
    • val for return form that produces value expression in the caller
    • var to explicitly request return form that produces initializing expression in the caller (usually the default)
  • #5010 Precedence diagram is missing partial operator
    • For now, const and partial have same precedence and can't be combined
• New issues
  • (none)
• Active discussions
  • (none)

Updates on proposals

• Decisions
  • #5448 Carbon/C++ Interop: Primitive Types proposal
    • Approved, not yet submitted
• New RFC proposals
  • (none)
• Active discussions
  • #5545 Expression form basics:
    • 06-26 Are composite forms more tied to actual expression syntax, or to tuple/struct values?
  • #5689 Semantic Identity and Order-Dependent Resolution for Rewrite Constraints
    • #generics-and-templates order sensitivity in applying rewrite rules

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.
  • DONE: Summary of the Zig community leader interviews shared with the moderation team.
  • WIP: Started marking more “good first issues” on GitHub
  • WIP: Conversation around mindful wording for our communication to The World™
• Last week’s stats across our spaces:
  • In the last 7 days in our repo: 10 authors pushed 38 commits to trunk, we changed 513 files with 23,495 additions and 8,094 deletions, 38 merged PRs and 4 open PRs, we closed 6 issues and opened 1 new one.

Activity on toolchain

• #toolchain const only has minimal implementation in the toolchain so far
• #toolchain #5728 Interaction between imports, runtime phase, and errors
• #toolchain #5734 #5733 Recent changes broke codegen, will use LLVM verifier to catch future problems
• #toolchain Refactoring association of types to value stores
• #toolchain Destruction model interaction with generics
• PRs:
  • test updates in #5705 #5707 #5708 #5691 #5704 #5716 #5722
  • C++ class as a parameter or return by value in #5709
  • tuple pattern matching in #5697
  • generics work in #5710 #5702 #5712 #5730 #5713
  • C++ interop in #5715 #5737
  • lowering of specific functions in #5688 #5684 #5723 #5732
  • prelude for tests in #5703 #5741
  • for loops in #5698 #5699 #5746
  • output when tests crash in #5739 #5740
  • partial in #5736
  • vtables in #5472
  • value stores in #5743
  • destroy in #5678 #5748
  • hashtable benchmark in #5735
  • others in #5721 #5731

Activity on C++ interop

• 06-24 Discussed tasks that could be worked on in parallel: Non null pointers, `#define` integer literals constants.
• 06-24 We need to write down the design for Carbon string literals, so it can be implemented and implement C++ interop for string literals to allow many system apis.
• 06-24 Primitive Types proposal looks in good shape.
• #interop When and how to generate the thunk for calling a C++ function.
  • Probably call into Clang Sema.
• #interop How to represent a C++-call thunk in SemIR

Other active discussions

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

• #interop #5743 Continued discussion about how to support reverse-lookup in CanonicalValueStore using a subset of the value fields.
• #control-flow It might make sense to split out an AsIterate interface to convert to something iterable.
  • Rust does something similar.
• #generics-and-templates #typesystem 06-26 More discussion of sum type redesign
• #strategy Carbon contributor summit in Toronto July 15-17
  • Not a conference -- the focus is on deep collaboration, similar to open discussion meetings.
  • There will also be a video option for those who can't attend in person, and a "trip report"-style summary afterwards.
  • No change to our decision process -- formal decisions will be made on GitHub proposals with documented rationale etc.
• #generics-and-templates The value of X after let X:! i32 = 0 is X symbolically, not 0.
  • #generics-and-templates What do we call the relationship between X and 0?
• #typesystem Is (F(), F()).0 an initializing expression?
• #safety Paper Memory Safety is Merely Table Stakes, about how interop breaks safety
• #typesystem Can ImplicitAsPrimitive return a non-primitive form?
• #generics-and-templates Associated constant values don't convert during impl lookup
• #generics-and-templates .Self can be used within a nested FacetType
• #generics-and-templates Storing the index of .Self to distinguish different values, in a way that works with importing
• 06-23 impl lookup when doing facet type conversion
  • We’re fine with the facet type’s assoc constant value taking preference over the final impl’s when there is a conflict. The priority is just that the function sees a consistent single value for the associated constant, whichever it is.
• 06-23 Use of anchor with named constraints
  • Will want to support both kinds of variance for parameters
• 06-24 (a, b) anchor will allow some conversions that add/remove const and partial.
  • The .type property of a form value will return the unqualified type
• 06-24 Implicit conversions that add/remove const and partial are allowed for some categories, and preserve category
• 06-24 What keywords to use to control variance
  • implicit_into and implicit_from.
• 06-23 06-24 Could potentially repeat the variance keyword on the interface parameter when either impling the interface (like ref) or using those names in the interface body (like each)
• #toolchain const is prefix not postfix operator, with link to rationale
• #toolchain Form variance are expected to be our solution to allowing operators to work on const types without having to define extra impls

RSS Feed for "Last Week in Carbon" here