Skip to content

Cheaper returned-object ownership in Go and Dart - #751

Closed
kixelated wants to merge 1 commit into
mainfrom
quest/m1/ffi-speed/object-ownership
Closed

kixelated wants to merge 1 commit into
mainfrom
quest/m1/ffi-speed/object-ownership

Conversation

@kixelated

@kixelated kixelated commented Nov 13, 2026 •

Copy link
Copy Markdown
Collaborator

Continues quest/m1/ffi-speed/object-ownership.md.

Problem

Returned Go and Dart objects pay substantial automatic ownership costs. The existing local profiling checkpoint revises the quest around those measurements; no binding implementation has changed.

Approach

Stopped implementation as requested because the revised plan conflicts with the maintainer's 2026-11-12 decision on kixel-ai/rust2#730: close takes the value through Rust's TAKEN state and the bindings' own hold counts go. The quest instead recommends retaining Go's hold count and stopping cleanup at close. Its Required section also waits for checks-raise, whose PR is still open.

Impact

  • No public API or wire changes. This checkpoint changes only the quest's profiling evidence and plan.

Alternatives

Retaining Go hold counts could reduce the measured cleanup cost but contradicts the settled close policy. Recommendation: retain that policy and revise this quest after checks-raise lands, using fresh measurements to establish whether a compatible optimization remains.

Waiting

  • Maintainer: the quest's Go recommendation conflicts with the settled close policy. Should its plan be rewritten to retain Rust-like close and remeasure after checks-raise lands, or should Go receive an explicit exception? Recommended: rewrite the plan and retain Rust-like close. No exception or implementation was selected by this agent.
  • kixel-ai/rust2: kixel-ai/rust2#730 must land before the quest's requested post-change measurements can run.

Follow-ups

None added. The existing quest remains unfinished.

Validation

Implementation stopped at the conflicting plan. quest gate just ci was canceled while queued because no implementation proceeds until the conflict is resolved. No tests ran, no local review ran, and the PR remains draft.

The pre-existing local profiling commit 5505681fe90b596686410a2a53d991efbc5385e7 is preserved in the checkout but was not pushed; the PR head remains the claim commit. Its measurements were not reproduced by this agent.

Models

Size: M
Implementer: Claude Code, claude-opus-5-5, unknown effort
Implementer: Codex, gpt-6.1-sol, high effort (handoff audit only; no implementation changes)

(Written by GPT-6.1 Sol with high effort)

@coderabbitai

coderabbitai Bot commented Nov 13, 2026

Copy link
Copy Markdown

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

  • Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

@kixelated

Copy link
Copy Markdown
Collaborator Author

Releasing this claim without merging. Its plan kept Go's own hold count, which conflicts with the maintainer's 2026-11-12 decision on kixel-ai/rust2#730: close is Rust-like, through the runtime's TAKEN state, and the bindings' hold counts go. The quest already says to profile only after checks-raise lands, since #730 rewrites the same handle word. #730 will rewrite the quest to build on Rust-like close and to require checks-raise, so it becomes ready again once #730 merges. This branch held only its empty claim commit.

(Written by Opus 5.5 with high)

@kixelated kixelated closed this Nov 13, 2026
@kixelated
kixelated deleted the quest/m1/ffi-speed/object-ownership branch November 13, 2026 07:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant