freenode

← digests

Toolchains: faster debug info, AI review limits

Languages & Toolchains2026-11-18

LLVM dominated the day's toolchain discussion, with a much faster dsymutil prototype and several disputes over AI-generated contributions and review load. Git, Swift, Haskell, and Julia also moved on branch cleanup, floating-point semantics, and ecosystem tooling.

LLVM RFC proposes dsymutil2 with large speedups

An RFC on the LLVM Discourse presents a dsymutil2 prototype that uses techniques from lld and reports 10-15x speedups with lower memory use on Chromium-scale inputs. The proposal matters to anyone debugging large Mach-O builds, where today's dsymutil is a bottleneck.

LLVM considers pull request limits for non-committers

A separate LLVM RFC proposes limits on pull requests from non-committers, aimed at maintainer burnout from low-quality AI-generated submissions. The thread drew 61 messages from 38 participants, a sign of how widely the review-load problem is felt.

Git patch would find squash and rebase merges for deletion

A patch series on the Git mailing list adds --delete-integrated, extending --delete-merged, so Git can detect squash and rebase merges by file content rather than commit reachability. That would let branch cleanup catch integrated work that today's reachability check misses.

SpaceStation.jl announced as a Pluto notebook workspace

SpaceStation.jl was announced on the Julia Discourse as a workspace for Pluto notebooks, including AI-related features. Early users praised the tool while discussing how it will be maintained after the lead maintainer steps back.

Contributors challenge LLVM ban on AI-generated communication

LLVM contributors asked to reopen the policy forbidding AI-generated communication, arguing that the project council's rough-consensus approval ignored objections and left enforcement ambiguous. The short, heated thread puts both the process and the practical meaning of the ban back in dispute.

Swift draft explores relaxed floating-point operators

A Swift Evolution thread discusses a draft for relaxed floating-point operators, in the style of existing overflow operators such as &+, that would permit algebraic reordering to cut latency. The proposal trades strict floating-point semantics for speed in code that can accept the looser model.

Haskell Foundation TLC publishes a technical vision

The Haskell Foundation's technical leadership committee said it is operational and published a shared technical vision for the ecosystem, inviting proposals and funding. The document is meant to give contributors a common target for ecosystem work.

LLVM debates AI use on good first issues

A heated LLVM thread tries to clarify where AI assistance is allowed on good first issues, and confronts one contributor over a high volume of AI-assisted pull requests across projects. The discussion sits beside the broader fight over AI communication rules and maintainer load.