2026-03-09— Last Week in Carbon

[josh11b]

News

• Carbon Copy No. 11: Generics Part III Parameterized Types published.

Updates on leads issues

• Decisions
  • #6768 how should function templates with explicit template arguments be exposed by C++ interop?
    • Use heuristics to identify arguments that must be template arguments.
    • Everything else defaults to being a non-template argument.
    • Some explicit marker to override that default.
  • #6739 Should we have a way to perform C++ default initialization?
    • Introduce an interface that initializes to a fully-formed state.
  • #6821 Should we provide an idiomatic way to express Unix file permissions?
    • Introduce octal literals with 0o prefix.
• New issues
  • #6830 Should import Cpp inline be part of the declaration list rather than the preamble?
    • Would allow inline C++ to be interleaved with Carbon code.
• Active discussions
  • (none)

Updates on proposals

• Decisions
  • #6358 C++ Interop: API importing and semantics: approved.
• New RFC proposals
  • #6825 C++ interop for overloaded functions and function templates
• Active discussions
  • (none)

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

• Conduct team updates:
  • none
• Moderation interventions last week:
  • A few accounts were banned after posting job-seeking messages on Discord, which are off-topic
  • Blocked: updating our community rules on Discord to make them more explicit in that regard
  • WIP: outreach for the next interview round of language community leaders in our neighborhood
• Activity in the past week:
  • 1 new contributor! 🎉
  • 13 authors have pushed 27 commits to trunk and 28 commits to all branches.
  • On trunk, 457 files have changed and there have been 8,082 additions and 7,290 deletions
  • We merged 28 PRs, opened 11 additional PRs, closed 4 issues, and opened 3 new issues.

Activity on toolchain

• #generics-and-templates Why does textual SemIR depict the function definition inside a generic but not the declaration?
  • It's an artifact; the declaration actually is in the scope of the generic.
• #toolchain Representing whether errors have been diagnosed
• 03-03 Interaction with context when reporting a diagnostic
• #generics-and-templates Are we worried about packing/unpacking generic params in the function body?
• #functions Would it be OK to have form-generic params after non-form-generic params in SemIR?
• #functions Is it OK if we model form generics with template actions that have side effects?
• #toolchain Converting a Carbon expression that represent a C++ type to a Clang expression
• #toolchain #6828 Syntax for new Any support in CARBON_KIND switch macro
• #toolchain We use llvm::formatv instead of std::format
• #toolchain Form generic implementation plan doc; will introduce some template-like actions that depend on each other's side effects
• #toolchain Building a witness table
• #toolchain Progress injecting a Carbon namespace into C++ for reverse interop
• #toolchain Looking for help fixing #6793
• #interop Handling C++ macros that interact with the surrounding code
• #llvm Fix for Clang crasher
• PRs:
  • pointer template params in #6810
  • Destroy in #6729
  • facet type in #6819
  • C++ macros in #6820 #6827
  • parameter patterns in #6815
  • explicit template arguments in #6814
  • lambdas in #6826
  • C++ overload resolution in #6829
  • C++ types impl Carbon interfaces in #6816
  • Clang format config in #6831
  • fuzzing in #6835
  • forms in #6817 #6812
  • AI skills in #6838 #6840 #6842 #6841
  • Update versions in #6818 #6846 #6848 #6847
  • infra in #6854 #6855
  • others in #6828 #6839
• New contributors:
  • @MK4070 made their first contribution in #6826

Activity on memory safety

• 03-03 What alias set should a C++ function be assumed to return when it receives and returns a const&, which may be a temporary (or value, coming from Carbon)?
  • Resolved to return a new alias set that is disjoint from everything else, called ^unsafe for C++ functions that receive and return pointers/references.
  • Calling the function would be an unsafe operation in Carbon.
• 03-05 Exploring thread safety in Safety Unit No. 37: Threads: Spawn
• 03-06 Considering more options for C++ returning pointer/reference, with ^unsafe(any) where invalidations no longer apply through an unsafe-annotated alias set.

Other active discussions

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

• #generics-and-templates 03-02
  • Canonicalizing facet values to dedupe specific interfaces in IdentifiedFacetTypes.
  • Using a FactualImplWitness instruction for finding a witness from a require decl breaks constant comparison with witnesses dervived through other means, as a LookupImplWitness.
• #typesystem Should we be worried about SemIR bloat for sum types?
• #strategy Carbon Copy Carbon: Proposal tracking #11 draft available for feedback.
• #build-help Is it normal to get a lot of "argument unused during compilation" errors when building a compilation database?
  • Yes, they're benign.
• #build-help Having trouble using Abseil flat_hash_map.
  • Don't use that, use common/map.h. Abseil license isn't compatible with ours.
• #infra Trouble with the new_proposal script.
• #pattern-matching Do we perform pattern matching before or after forming a specific? Should we make sure we get the same answer either way?
  • Constructively we'll probably have to.
• #strategy Link to Against Query Based Compilers
• #design-documentation Is there a reference for the pseudo-BNF we use?
• #editor-integrations What do we use to format Markdown?
  • prettier, but pinned to < 3.4.0 because of a bug.
• #strategy Should we include "C -> C++" as an example of a successor language (from #6833)?
• #generics-and-templates Attempting to implement use of Self in require declarations in a named constraint, which is creating cycles in the toolchain evaluation model.
  • After a number of attempts, belief this may be unimplementable in the current design and eval model: explainer doc
• 03-05 planning to add gemini.md and/or a skills directory, to help agent-based dev workflows.
• #infra More possible GitHub SHA trouble. Trying #6854 to stop relying on them.
• #syntax Has the placement of self been resolved?
  • Not yet, but very close.
• #getting-started How to get started with contributing, particularly from HPC perspective?
• #getting-started Looking for feedback on #6858, to add Windows support.
• #interop Deref interface for user smart pointer types
• #interop bound is from proposal proposal #5434

RSS Feed for "Last Week in Carbon" here