2025-05-26 — Last Week in Carbon #5544

josh11b · 2025-05-27T17:17:51Z

josh11b
May 27, 2025
Maintainer

Sorry for the delay! This week's issue is coming out one day late due to a US Holiday on Monday.

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.
Last week’s stats across our spaces:
- An account posting spam in our GitHub repository was banned and reported to GitHub, hiding all corresponding comments.
- In the last 7 days in our repo: 9 authors pushed 32 commits to trunk, we changed 449 files with 11,924 additions and 15,507 deletions, with 32 merged PRs, 8 open PRs, 1 closed issue and 5 new ones.

05-20 Fuzzer bugs fixed
05-20 #toolchain Some work to do to modernize tests
- Use --dump-sem-ir-ranges=only on new tests
- Some iteration on refining what output is included
05-20 #generics-and-templates How to avoid redundant recomputation in #5506?
#toolchain Still iterating on #5314
#toolchain Anyone want to fix a crash?
#toolchain Specifying target in tests to make deterministic
#toolchain Figuring out the failing test with ASAN which doesn't include custom crash dumps
- Some sources of randomness in the execution from sort and hashing
- Documenting which functions can cause tables to reallocate
PRs:
- comments in #5499 #5498 #5497
- import in #5483 #5400 #5517
- instruction formatting in #5495 #5501 #5491 #5494 #5505 #5516 #5503
- golden tests in #5477 #5507 #5508 #5518 #5512 #5525 #5509
- CHECK in #5504 #5519
- lowering for globals in #5492
- importing C++ inline functions in #5427
- compiler output in #5520 #5521
- impl diagnostics in #5420
- others in #5500 #5515 #5526 #5527

05-20 How to propagate parameters for Clang to generate the AST. Decided to keep as is for now reevaluate when pcms are used.
05-20 Using Decl IDs vs. Pointers. Decided on using a side table containing pointers, which would eventually hold lazy decl ptrs.
05-20 Primitive types. Learning from Rust experience. Decided to start with a few types and not have platform dependent aliases for now.
#interop How to represent the mapping from Carbon functions to Clang declarations.
#interop How to generate C++-to-Carbon bridge function.

#infra Considering removing the devcontainers due to issue #5496 (implemented in #5500)
#variadics Will the ... [[each Cat]] each t: each auto syntax we discussed be dropped?
- Yes, it's been subsumed by the in-progress design of "forms"
#syntax Should we plan on supporting destructuring assignment?
- If so, should it be assignment, or should it initialize an unformed object, or should we have syntax for both?
#lambdas Should => deduce the form, not just the type?
#getting-started Discussion of writing an HTTP server in Carbon, and the interop/migration strategy.
- Also a report of fake online tutorial for doing so (probably LLM-generated).
#strategy Created a playlist of Carbon talks on YouTube.
#general-design 05-22 Discussion of possible ways to make the pattern/declaration syntax closer to the underlying semantics.
- Dropping var as sugar for let var?
- Introducing val as a counterpart for ref?
- Side discussion: let's immutability guarantee is not as strong as some people expected (unless there's compiler enforcement). Is that a problem?
- Filed #5522, #5523, #5524 to get resolution on these issues
#generics-and-templates Resolving rewrite constraints
- Need directionality to avoid undecidability
- Only okay to have a cycle if there is an assignment to a concrete value
#generics-and-templates Constraints that conflict with applicable impls
05-19 What patterns can match what return forms?