[M3gA-Mind]

Summary

When several parallel branches fail with no error handler in one superstep, the reported run error no longer depends on which branch finished first.

Root cause

finish_execution stored the failing node's structured EngineError into the run's shared terminal_error slot, first writer wins. In a parallel superstep the first writer is whichever activation completes first, so one graph and input could report different errors under different latencies. The graph runtime already names a failed step by the lowest active-set index, so the two could also disagree.

Fix

• The slot now holds a TerminalError { step, branch, error } (src/engine/terminal_error.rs).
• Handlers offer their error through terminal_error::offer, which keeps the lowest (superstep, active-set index). The branch index comes from the existing NodeContext::fork.

Rule (documented on offer and in the README's Reliability section)

The earliest superstep wins. Within one superstep the lowest active-set index wins. For a fan-out that is the edge declaration order, the same order parallel updates are folded in. Completion order never matters. A single failure is reported unchanged, and lane failures still belong to their gather.

Validation

On a Linux build host, on main f778023 (merges cleanly onto 8b22179):

• cargo test --workspace --all-features --locked: 2860 passed, 0 failed, 35 ignored (main's 35; no new ignores).
• clippy -D warnings: 0. fmt: clean. cargo +1.88.0 check --all-features (MSRV): passes. rustdoc -D warnings: passes.
• Red control: the new tests/parallel_failure_order_tests.rs, copied onto unmodified main, fails on the first schedule ("delays [10, 0] reported … unhandled failure at n2"). It passes with this change.

Tests

• src/engine/terminal_error_tests.rs:
  • the lowest index wins in either arrival order;
  • an earlier superstep is not replaced by a lower index in a later one;
  • a single failure passes through.
• tests/parallel_failure_order_tests.rs: two fan-out branches fail under opposite yield-count latencies, and both runs report the first-declared branch's error.

Found while writing the remaining fuzz targets for #46.

[coderabbitai]

Warning

Review limit reached

• Run on-demand review

_{This review includes 11 billable files and costs up to $2.75.}

Or wait 20 minutes for your next included review.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: f1a44c8e-caf5-4331-a068-f303aeca95fa

Commits

Reviewing files that changed from the base of the PR and between 8b22179 and 337370b.

Files selected for processing: 11

• README.md
• crates/tinyflows/src/engine.rs
• crates/tinyflows/src/engine/build.rs
• crates/tinyflows/src/engine/build/activation.rs
• crates/tinyflows/src/engine/build/handlers.rs
• crates/tinyflows/src/engine/build/outcome.rs
• crates/tinyflows/src/engine/resumable/continuation.rs
• crates/tinyflows/src/engine/run_state.rs
• crates/tinyflows/src/engine/terminal_error.rs
• crates/tinyflows/src/engine/terminal_error_tests.rs
• crates/tinyflows/tests/parallel_failure_order_tests.rs

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

[tinysweeper]

Tiny Sweeper review

Reviewed head: 337370b4cfaf2659027dce9fcf5d2d1b22a905cb

Verdict: Ready for maintainer review · Priority: none
Coverage: Reviewed 11 of 11 files.
Pull request purpose (title): fix(engine): report…

Diagnostics and supporting details

Full summary

Pull request purpose (title): fix(engine): report the same run error whatever order parallel failures finish in. Tiny Sweeper completed its review; deterministic results follow.

Review snapshot

Change surface	Files	Review signal	Count
Production	8	Active findings	0
Tests	2	Noted findings	0
Documentation	1	Resolved findings	0
Configuration	0	Pending checks/questions	0

Completeness: Complete
Test assessment: No supported feature-to-test mapping was available; this does not mean tests are absent or passed.

Features

• Added — Deterministic terminal-error slot with order-aware selection: Introduces `TerminalError`/`TerminalSlot` and `terminal_error::offer`, which keeps the failure with the lowest (superstep, active-set index) and ignores completion order, making the reported run error a function of the graph and input rather than of scheduling latency. A single failure passes through unchanged, and an earlier superstep's error is never replaced by a lower-indexed later one. (crates/tinyflows/src/engine/terminal_error.rs, crates/tinyflows/src/engine.rs use build::build_graph;)
• Internal refactor — Terminal-error slot type threaded through graph construction and handlers: `build_graph` and `register_handlers` take the new `TerminalSlot` type instead of the raw mutex of `Option<EngineError>`, so the ordering rule is enforced at the single point of insertion rather than at each call site. No change to the error string channel that parallel paths already consume. (crates/tinyflows/src/engine/build.rs pub(super) fn build_graph(, crates/tinyflows/src/engine/build/handlers.rs pub(super) fn register_handlers(, crates/tinyflows/src/engine/build/outcome.rs where)
• Modified — README documents the parallel-failure reporting rule: The Reliability section now states that when several parallel branches fail under `stop` in one super-step, the error of the branch earliest in that step's active set (edge order for a fan-out) is reported, never the first to finish, and that an earlier super-step's failure always wins — making the observable contract explicit for users. (README.md Rust 2024 · MSRV 1.85 · `#![forbid(unsafe_code)]` · GPL-3.0-or-later.)

Tests

No supported feature-to-test mapping was produced. Test execution is not inferred.

Findings

No active actionable findings.

Before merge

None.

Agent review details

critique

• Conclusion: Success
• Scope reviewed: all assigned evidence
• Lane summary: Reviewed 11 files; 0 findings. _The code index is behind this pull request (indexed at `16cd898e2a25`), so retrieved context may be out of date._

security

• Conclusion: Success
• Scope reviewed: all assigned evidence
• Positive: The security lane answered with no findings across the 7 files it reviewed; the remaining files were excluded as prose or build-output-shaped rather than left unreviewed for cause.
• Lane summary: Reviewed 7 files; 0 findings. 4 files were not security-reviewed: README.md (prose or tabular data), crates/tinyflows/src/engine/build/activation.rs (vendored or build output), crates/tinyflows/src/engine/build/handlers.rs (vendored or build output), crates/tinyflows/src/engine/build/outcome.rs (vendored or build output). _The code index is behind this pull request (indexed at `16cd898e2a25`), so retrieved context may be out of date._

tests

• Conclusion: Success
• Scope reviewed: all assigned evidence
• Positive: The tests lane answered with no findings and characterized the deterministic error-selection policy as correct, pinned on both ordering paths, and safely mergeable.
• Lane summary: This change introduces a deterministic error-selection policy for parallel branches that fail under `stop`: the earliest superstep wins, and within a superstep the lowest active-set index wins. The implementation is correct, the tests pin the behaviour on both ordering paths, and the new `TerminalSlot` type isolates the selection logic cleanly. The change is safe to merge. _The code index is behind this pull request (indexed at `16cd898e2a25`), so retrieved context may be out of date._

commits

• Conclusion: Success
• Scope reviewed: all assigned evidence
• Positive: The commits lane answered with no findings, reporting no sensitive content in the committed files.
• Lane summary: No sensitive content in committed files. _The code index is behind this pull request (indexed at `16cd898e2a25`), so retrieved context may be out of date._

description

• Conclusion: Success
• Scope reviewed: all assigned evidence
• Lane summary: The description matches the diff: the shared error slot becomes `TerminalSlot`/`TerminalError`, failures are offered through `terminal_error::offer` with `(superstep, active-set index)` ordering sourced from `NodeContext::fork`, the two entry points unwrap `held.error`, and the README documents the rule. The listed unit and integration tests are present, and the new files stay well inside the repository's 500-line and `_tests.rs`-naming rules. Nothing in the change looks like it would misbehave for a reachable input. _The code index is behind this pull request (indexed at `16cd898e2a25`), so retrieved context may be out of date._

Evidence and run details

• Models: ladder/vectors-oai3, deepseek-v4.1-flash, deepseek-v4-flash
• Spend: $0.025082
• Completion usage: incomplete; some provider completion costs are unavailable.
• Tokens: 748732 input · 82368 output · 269948 cached · 529 embedding
• Updated: 2026-11-11T11:37:53Z

Head	State	Pass summary
337370b4cfaf	ready for maintainer review	0 active finding(s), 0 resolved finding(s) (at 2026-11-11T11:37:53Z)

[tinysweeper]

tinysweeper found nothing blocking. Approving.

Diagnostics

Total spend: $0.0251