2025-07-07 — Last Week in Carbon #5780

josh11b · 2025-07-07T21:40:00Z

josh11b
Jul 7, 2025
Maintainer

Decisions
- #5413 Does struct field order matter for impl lookup and type structure?
  - Resolved: "different types but convertible"
New issues
- (none)
Active discussions
- (none)

Decisions
- #5661 Progressive disclosure principle
  - Approved & merged!
- #5448 Carbon/C++ Interop: Primitive Types proposal
  - was approved, now merged
New RFC proposals
- #5776 Replace impl fn with override fn
  - Implementing leads issue decision in #5253
- #5689 Semantic Identity and Order-Dependent Resolution for Rewrite Constraints
Active discussions
- #5434 ref parameters, arguments, returns and val returns
  - added using ref to specify value representation
- #5606 Keep design documents current
  - When does this happen in the proposal review process?

Moderation and conduct teams are still the same size, available in theory from UTC-7 until UTC-5.5: 10 trained members altogether + community lead (hands-off trainer, coach, guide)
- RESUMED: conversation about how to improve our code of conduct framework (scope, conditions under which members should step back from their duties, voting mechanism, sustainability, etc.)
- WIP: Transparency report for the last quarter (2025Q2)
Peaceful past week again, with:
- In the last 7 days in our repo: 7 authors pushed 47 commits to all branches, 21 of which to trunk, we changed 106 files with 2,189 additions and 5,060 deletions, 21 merged PRs and 10 open PRs, we closed 3 issues and opened 2 new ones.

06-30 Builtin conversion between facet types is not recognized as an implementation of ImplicitAs.
#generics-and-templates What needs to be done on #5673
#toolchain #5756 Adding value type parameter to ValueStoreChunk.
#toolchain Connection between id types and the type of values stored in the corresponding value store, and more about the value store change
#toolchain Turning numbers into InstIds when debugging
#toolchain When you get a NameRef vs. a BindSymbolicName.
PRs:
- partial in #5744
- value store parameterization #5754 #5755 #5756 #5757 #5758 #5759 #5761
- import vtable_ptr in #5762
- mangling in #5764
- docs #5751 #5752 #5749 #5770
- others in #5747 #5729 #5768 #5765

07-01 Discussed how to call C++ thunks.
- The C++ thunk should only have i32, i64, pointers in its signature.
- The C++ thunk should be mapped to a SemIR::Function to allow Lower to similarly create the llvm::Function when lowering a call.
- SemIR should include the materialization of values and converting them to pointers. We should type-check calls to the Thunk in SemIR
- Function mangled name should be custom: "<C++ mangled name>.<some suffix>", using AsmLabelAttr.
07-01 Pointer conversions (virtual / multiple inheritance).
- Consider whether we want to support some conversions in Carbon or create a different type of Thunk that only does conversion.
- We need to design how multiple inheritance is mapped to Carbon.
We should keep working on per function thunks rather than per call thunks.

06-30 #generics-and-templates Facet type .Self canonicalization
06-30 What is going on with .Self on the right of an impls?
#language-questions How does Carbon's pointers and ownership model compare to Mojo's pointers and ownership model?
#pointers-and-references You can take the address of ref.
#typesystem Interaction between type conversions and forms
#typesystem Choice types
#typesystem Adding more type information shouldn't change observed behavior
#infra #lambdas was renamed #functions on Discord
#infra Unexpected bazel behavior
#summit Pre-reading on Ante and Rust for safety discussions