2025-07-14 — Last Week in Carbon

[geoffromer]

Updates on leads issues

• Decisions
  • (none)
• New issues
  • (none)
• Active discussions
  • (none)

Updates on proposals

• Decisions
  • (none)
• New RFC proposals
  • #5781: do...while loops similar to the C++
• Active discussions
  • #5434: ref parameters, arguments, returns and val returns
    • Discussions seem to have converged.

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

• 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)
  • RFC: Transparency report for the last quarter (2025Q2)
• Peaceful past week again, with:
  • In the last 7 days in our repo: 6 authors pushed 19 commits to trunk, we changed 356 files with 9,622 additions and 6,961 deletions, 19 merged PRs and 5 open PRs, we closed 1 issue and opened 2 new ones.

Activity on toolchain

• 07-07 Numbering .Self references
• 07-08 Single clean up block or one per code path with destructors?
  • Going with a single one to match Clang, and since duplicating is easier than deduplicating.
• 07-08 #toolchain #5793 Generics for vtables
• #toolchain Implementing "destroy." Don't yet have a way to specify an impl for all tuples.
• #toolchain #5798 Naming functions in impls in SemIR
• PRs:
  • pattern match in #5771
  • field order matters for impl lookup in #5778
  • C++ import in #5769 #5783 #5767 #5773 #5787 #5794
  • destruction in #5785 #5790
  • carbon-busybox in #5795
  • clangd-tidy in #5763 #5782
  • benchmark runner in #5797
  • others in #5779 #5786 #5791 #5792

Activity on C++ interop

• 07-08 To get a pointer to pass the thunk we should follow the `&` operator logic.
• 07-08 Return value might emit an emplacement new expression on the thunk side.

Other active discussions

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

• #strategy All meetings are cancelled the week of July 14 due to the summit.
• #infra Continued discussion of "No install marker at path" errors.
  • Turns out to be due to non-hermetic behavior of Bazel symlinks. Fixed in #5795.
• #getting-started #control-flow Discussion of do ... while proposal #5781.
• #pointers-and-references Can you implement an interface differently for ref T than for T?
  • Only if the interface is form-generic.
  • How do you write an associated constant that uses such a specialization?
• #typesystem Does the "alternatives are refinements" model apply to all sum types or just choice types?
• #infra Looking for other opinions on whether to push back on changes like #5774.
• #summit Posted agenda and other information for the summit.
  • Video call info for the summit is in the project calendar.
• #infra autoupdate_testdata.py sometimes warns about "discarding analysis cache", but it shouldn't have to do that.
• #infra Still seeing bogus clangd-tidy errors in CI, even after #5782.
  • Known issue, but shouldn't block merges because it's not in the merge queue.
• #syntax Should we drop fn destroy and use impl as Destroy instead?
  • More verbose, but no more so than ImplicitAs and operators.
  • Otherwise having fn destroy desugar to impl as Destroy could lead to confusing conflicts with explicit impl as Destroy.
  • However, fn destroy flags that something additional is happening here: the function body will be synthesized by the compiler.
• #getting-started Is it possible to directly use the toolchain directly, without bazel run?
  • Yes, you can use it in-place via bazel-bin/toolchain/install/prefix_root/bin/carbon, or generate an installable tarball with the //toolchain/install:carbon_toolchain_tar_rule build target.
• 07-07 When do named constraints have to be complete? If we want to support using using incomplete constraints in more places, that means having a separate "resolution" step and a separate "resolved facet type" to represent the transition when they are required to be complete.
• 07-10 Transitioning from bound to more expressive lifetimes
• 07-10 Deep vs. shallow immutability. Decided shallow immutability since it allows more expressivity.
  • Have workaround for Rust compatibility, which needs deep immutability, using a library type
• 07-10 Do we need both immutability and const?
• 07-10 Making progress on unformed state doc

RSS Feed for "Last Week in Carbon" here