Repository navigation
fix(engine): a lane activation never writes the node's top-level slot - #195
Merged
M3gA-Mind merged 2 commits intoNov 11, 2026
Merged
Conversation
Lane slots are never cleared between loop passes, and the gather counted every slot under its predecessors against the scatter's lane_count, taken from the first lane_count anywhere in state. A pass with 3 lanes followed by one with 2 released [20, 21, 12], where 12 is the first pass's third lane. The scatter now stamps each lane envelope with its activation step, every lane slot records `_scatter` and `_scatter_step`, and the gather counts only slots from its own scatter's current activation, reading lane_count from that scatter. Slots from checkpoints written before this carry no stamp and count as before while no stamped slot exists. Also corrects the reason given for refusing a gather with no scatter: it releases empty and silently drops its predecessors' output.
Four paths in a scatter lane wrote `nodes.<id>` instead of the lane's slot: a cancel seen at the node boundary, a cancel during retry or backoff, a cancel while waiting to re-enter, and a Reenter (a gate or approval polling), which also re-activated with a bare id and so left its lane. A cancelled lane also kept routing down its static edge, so its successor ran outside the lane. A cancelled lane now marks its lane slot "cancelled" and stops; a re-entering lane writes "pending" notes to its lane slot and re-activates with its own lane envelope. The gather counts neither as arrived, and gate and approval read their poll count from the lane's slot.
Collaborator
Author
|
Warning Review limit reached
This review includes 17 billable files and costs up to $4.25. Or wait 4 minutes for your next included review. View limit details
|
Tiny Sweeper reviewVerdict: Incomplete · Priority: none Diagnostics and supporting detailsFull summaryPull request purpose (title): fix(engine): a lane activation never writes the node's top-level slot. Tiny Sweeper reviewed this change across 5 lane(s) and found 0 active actionable finding(s). Detailed lane evidence and any incomplete work are listed below. Unreviewed paths
Partial review reasons
Review snapshot
Completeness: Incomplete FeaturesNone identified with supported citations. TestsNo supported feature-to-test mapping was produced. Test execution is not inferred. FindingsNo active actionable findings. Before merge
Agent review detailscritique
security
tests
commits
description
Evidence and run details
|
![tinysweeper[bot]](https://avatars.githubusercontent.com/in/4526585?s=60&v=4){kind=small}
tinysweeper
Bot
added
the
priority: p3
M3gA-Mind
added a commit
to M3gA-Mind/tinyflows
that referenced
this pull request
Nov 11, 2026
Resolves the CHANGELOG conflict with the engine fixes that landed on main (tinyhumansai#192-tinyhumansai#195): both Unreleased entries are kept.
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on
fix/gather-current-pass-lanes: merge that first. This branch contains its commit until then.Summary
Inside a
scatterlane, an activation must write onlynodes.<id>.lanes.<lane>, never the node's top-levelnodes.<id>. Four paths broke that. This fixes them and keeps the gather counting correctly.Root cause per path
engine/build/activation/gates.rs): wroteitems_update(&node.id, &[], None)whether or not the activation was in a lane. It also returned a plainUpdate, so the static edge scheduled the successor outside the lane, where it wrote its own top-level slot.activation.rs): the same write and the same static-edge leak.outcome.rs): wroteitems_update_with_meta(&node.id, …)and re-activated withCommand::goto([node.id]), a bare id. So the re-entered activation lost its_laneenvelope and ran outside the lane. Cancel while waiting to re-enter had the same write.gateandapprovalreadpolls, and a gate itsfromspawn's tickets, from the top-level slot. Inside a lane,max_pollsnever bounded them, andgate { from: [spawn] }never found the lane's tickets.Fix
engine::state::cancelled_result: in a lane it writes the lane slot with status"cancelled"and goes toEND. Outside a lane the behaviour is unchanged."pending", and re-activates through aSendcarrying the same lane envelope, including the scatter step from the stacked fix.gatherskips"pending"and"cancelled"slots when counting arrivals.NodeContext::slotreturns the lane slot when present, otherwise the top-level slot. Gate and approval read through it.Validation
On a Linux build host, merged onto main 8b22179:
cargo test --workspace --all-features --locked: 2866 passed, 0 failed, 35 ignored (no new ignores).tests/lane_top_level_writes_tests.rson its base (the stacked fix alone) fails all three tests. Boundary cancel and retry cancel fail with "lane body node … has a top-level slot". The re-entering gate fails with "graph exceeded the recursion limit of 200 steps", because its re-activation left the lane. All three pass with this change.-D warnings, fmt,cargo +1.88.0 check --all-features, rustdoc-D warnings.Tests
tests/lane_top_level_writes_tests.rs(mockfeature):backoff_ms: 0) and during backoff (backoff_ms: 40) inside a lane;ticketsandfromforms. It asserts no top-level writes at any checkpoint, that the gather never releases while a lane is pending, and that results come in lane order.Found while writing #46's fuzz targets.