Skip to content

docs: post-0.8.0 roadmap status and CLAUDE.md for the 0.8 API - #275

Merged
unbalancedparentheses merged 1 commit into
mainfrom
docs/post-080-refresh
Nov 16, 2026
Merged

unbalancedparentheses merged 1 commit into
mainfrom
docs/post-080-refresh

Conversation

@unbalancedparentheses

Copy link
Copy Markdown
Contributor

ROADMAP: 0.8.0 marked released (with qualification summary); release numbering fixed (0.9 introspection, 0.10 start + dynamic supervisor, 0.11 admission — 1.0 item 3 said 0.10); 0.8 features moved to Implemented; post-0.8 cleanups listed. CLAUDE.md: 0.8 Handler signature (PendingReply), reply tokens and jobs, ExitReason Killed/NoProc, send_request with Deadline, the death-point rule. Docs only.

@unbalancedparentheses
unbalancedparentheses marked this pull request as ready for review November 16, 2026 12:35
@github-actions

Copy link
Copy Markdown

🤖 Gemini Code Review

Looks good! The documentation updates in CLAUDE.md and ROADMAP.md accurately reflect the v0.8.0 API and lifecycle changes (PendingReply/ReplyTo, pipe/pipe_blocking/emit_at, send_request, new ExitReason variants, and single-point actor cleanup).


Automated review by Gemini · gemini-4-argon · custom prompt

@github-actions

Copy link
Copy Markdown

🤖 Claude Code Review

Review: PR #275 (docs: post-0.8.0 updates to CLAUDE.md and ROADMAP.md)

This PR only changes documentation, so there are no correctness or concurrency risks in code. The updates are mostly accurate and consistent with each other. A few items need tightening.

CLAUDE.md

  1. Inconsistent PendingReply signature.

    • The Handler<M> table row (around line 61) says handle(msg, reply: PendingReply<R>, ctx).
    • The lifecycle section (around line 80) says PendingReply<'_, M::Result>.
    • Pick one form. The lifetime version is presumably the real one, and readers (and Claude) will copy whichever they see first.
    • The PendingReply<R> / ReplyTo<R> row has the same problem.
  2. stopped() skip condition is imprecise (line ~81).

    • "skipped on kill or an untrapped exit signal" implies any untrapped exit signal kills the actor.
    • Under the OTP semantics this release adopts, an untrapped Normal exit signal is ignored.
    • Suggested wording: "skipped on kill or an untrapped non-normal exit signal".
  3. Kill vs Killed (line ~71). Listing both without explanation invites confusion. Add a short gloss, e.g. "Kill (the signal sent) / Killed (the reason observers see)". Also note that the enum is #[non_exhaustive], so match arms need a wildcard.

  4. Hard-coded file path in the "Death has one point (actor_cleanup.rs)" sentence.

    • ROADMAP.md, in this same PR, says the large actor files are being split.
    • Give the full crate-relative path, or point only to semantics.md, so the reference doesn't go stale.
  5. Non-type entry in the "Type" table. ctx.pipe / pipe_blocking / emit_at are methods, not types. Either move them under "Common Patterns" or name the relevant job/handle type.

  6. Request examples (line ~97).

    • Please confirm that the 5 s default timeout matches the code; it's an easy number to drift.
    • It would also help to show what rx is and how it's awaited.
    • The Response<T> row ("tasks: .await, threads: .unwrap()") may no longer describe send_request's return type.
  7. Unverified claim on line 38. The docs/MIGRATION.md row now promises an "old-call → new-call table", but this PR doesn't touch that file. Please verify the table exists.

ROADMAP.md

  1. "0.10 admission package" → "0.11" (line ~131). Good catch; this now matches the planned-order list.

  2. D1-D3 bookkeeping is consistent. They now appear in three places:

    • moved into 0.8 in the planned order,
    • noted as shipped under 0.9,
    • added to the Implemented table.
  3. Dynamic supervisor placement. It moved from "Next feature" to "0.10 designs". That is consistent with the planned order, but the old "next feature" framing is gone without comment. If this is a deliberate reprioritization behind introspection, a one-line note would help readers who remember the earlier plan.

  4. Unanchored claim in the v0.8.0 bullet. "run it on 24-hour canaries" is a factual claim; consider linking the client PRs.

  5. Minor: CHANGELOG. CLAUDE.md says to keep Unreleased current. A docs-only change probably doesn't need an entry, but confirm Unreleased was reset after the 0.8.0 release.

Once the PendingReply signature inconsistency (CLAUDE.md #1) and the stopped() wording (CLAUDE.md #2) are fixed, this is fine to merge.


Automated review by Claude · claude-opus-5-5 · custom prompt

@github-actions

Copy link
Copy Markdown

🤖 Codex Code Review

No findings. This PR only updates CLAUDE.md and ROADMAP.md; it changes no executable code and introduces no runtime, concurrency, or memory-safety changes. The documentation updates look consistent with the existing API and guide.


Automated review by OpenAI Codex · gpt-6.1-sol · custom prompt

@github-actions

Copy link
Copy Markdown

🤖 Kimi Code Review

Non-blocking documentation issue

  • ROADMAP.md, “0.10: starting and supervising, OTP-complete” (around line 111): The phrase “designs approved: unified-start.md; dynamic supervisor in design: dynamic supervisor (rank 12a) #245” can imply that both designs are approved. This conflicts with the earlier status saying the dynamic supervisor’s design and TLA+ model are still in review. Consider:

    “unified start approved; dynamic supervisor design and TLA+ model in review in design: dynamic supervisor (rank 12a) #245”

Otherwise, the updates are internally consistent, including moving D1–D3 and the admission package to their correct releases. Since this PR changes documentation only, there are no runtime correctness, concurrency, performance, or memory-safety implications.


Automated review by Kimi (Moonshot AI) · kimi-k3.1-pro · custom prompt

@github-actions

Copy link
Copy Markdown

🤖 Grok Code Review

Documentation-only PR. The 0.8 status updates are consistent (D1–D3 moved into 0.8, admission package 0.10 → 0.11 in both the release plan and the 1.0 requirements). No code, concurrency, or lifecycle implementation to review.

A few doc nits:

  1. CLAUDE.md type table (~line 71) lists both Kill and Killed with only “OTP killed/noproc”. That will be read as duplicates. One clause on the difference (untrappable signal you send vs reason observers see) would match how the rest of this cheat sheet is used.

  2. Handler signature is written two ways. The table (~line 61) uses PendingReply<R> with no lifetime; the lifecycle section (~line 80) uses PendingReply<'_, M::Result>. Pick the real signature so generated #[actor] impls and hand-written Handler impls don’t diverge.

  3. CancellationToken (~line 68) now says it only cancels timers, and actors stop via ctx.stop() / an internal flag. That contradicts the previous public description. Confirm no public path still ties actor lifetime to a token; if any does, this line is wrong. Worth a row in the 0.7 → 0.8 migration table this file now claims exists (MIGRATION.md is not in this diff).

  4. Death vs stopped (~lines 81–83). “Callers answered before any user code” and “stopped skipped on kill / untrapped exit” are compatible only if stopped runs after that fence on the paths where it runs. Say that explicitly so nobody answers callers from stopped or assumes it always runs.

  5. request “5 s timeout” (~line 98) is easy to treat as a hard contract. Mark it as the default if send_request / Deadline can override it, and mention Deadline in the type table since the new example depends on it.


Automated review by Grok (xAI) · grok-4.7 · custom prompt

@github-actions

Copy link
Copy Markdown

🤖 DeepSeek Code Review

The PR is documentation-only, so no code/concurrency bugs to flag. The updates are mostly clear and align with the 0.8 release, but there are a few consistency and clarity issues:

  • ROADMAP.md, “What’s shipping” v0.8.0 bullet: The bullet omits the D1–D3 determinism fixes, but the same PR adds “Simulator determinism D1–D3 (0.8)” to the backlog table and the later “0.8 (released 2026-11-15)” line in the 1.0 section explicitly includes “the D1-D3 determinism fixes.” Add D1–D3 to this bullet to avoid contradicting the rest of the doc.

  • ROADMAP.md, “What’s shipping” v0.8.0 bullet: #271 is grouped with API cleanup: “...and the API cleanup (#240, #271)”. The previous wording had #240 for implementation/review fixes and #271 for release metadata/publication. Conflating them misattributes the PR. Suggest separating: “...and the API cleanup (#240); release metadata/publication in #271.”

  • ROADMAP.md, “0.10 designs” line: (#245) is plain text, while other PR references in the file are markdown links (e.g. [#245](https://github.com/lambdaclass/spawned/pull/245)). Make it a link for consistency.

  • CLAUDE.md, Handler<M> table row: The table uses PendingReply<R>, but the Actor Lifecycle section directly above uses PendingReply<'_, M::Result>. Use the same type parameter for consistency, e.g. PendingReply<'_, M::Result> (or define R as M::Result in the row).

  • CLAUDE.md, Common Patterns: let rx = actor.send_request(...) uses rx, which typically implies a channel receiver. Since send_request returns a future/response, a name like response or reply_fut would be clearer.

  • Optional (CLAUDE.md, CancellationToken row): The phrase “actors stop through ctx.stop() / an internal stop flag” is a bit vague. It’s fine if the intent is to say actors are no longer cancelled via CancellationToken, but consider clarifying that termination also occurs via exit signals/kill to avoid implying that ctx.stop() is the only path.

No other issues noted.


Automated review by DeepSeek · deepseek-v4-pro · custom prompt

@unbalancedparentheses
unbalancedparentheses added this pull request to the merge queue Nov 16, 2026
Merged via the queue into main with commit ed30336 Nov 16, 2026
55 checks passed
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