Skip to content

Clear lab.5 release harness portability failures #706

Description

@shiny-code-bot

Objective

Clear the release-only timing and macOS portability failures exposed by validation run 31973347222 without masking product-contract failures.

Finish Line

The TypeScript streamed-history test is stable under release-runner load, the HTTP proxy listener does not panic on a nonblocking accepted socket, the project-validation interrupt fixture treats expected FIFO closure deterministically, and the guardian retry-history test completes without stack overflow. Focused tests pass repeatedly on macOS and the affected release jobs no longer fail for harness/resource reasons.

Current Status

State: waiting after completion of the first bounded harness stage. PR #716 merged on August 17, 2026 at main@780e5f32effbaeacf8a5dccdb0728b5a7bed5d1b from reviewed head 8a720ceb7a78312ef24482994a919b348fba2c25; all 17 checks passed with no review feedback or flaky retries.

The merged repair restores the codex_utils_cargo_bin::cargo_bin contract: absent cross-package CARGO_BIN_EXE_* variables now produce CargoBinError::NotFound instead of panicking inside deprecated assert_cmd code. Cargo and Bazel positive/missing-path regression coverage, dependency/lock checks, lint/format, strict convergence, and final Opus/Gemini reviews were clean. JetBrains inspection remained UNKNOWN because the worktree could not be opened by the IDE helper.

The helper stage conclusively unblocked all 14 #707 filter expressions. They expand to 25 cases: 9 pass and 16 expose real MultiAgentV2 product assertions/timeouts. The native #707 blocker relationship has therefore been removed; #707 is now the sole active Project Now lane.

The original #706 work remains open but parked: SDK runStreamed stability, two route-aware proxy failures, the project-validation FIFO closure race, and guardian retry-history stack overflow are still unaddressed.

Next action after the next bounded #707 product stage: return here, remeasure the four original portability clusters from then-current immutable main, and choose only the smallest coherent harness repair.

Acceptance Criteria

  • runStreamed history coverage passes repeatedly under constrained runner load without a blanket suite timeout increase.
  • Both route-aware proxy tests pass on macOS without WouldBlock or connection reset.
  • The interrupted project-validation FIFO test accepts only the expected closure race and still proves input recording.
  • The guardian retry-history test passes without stack overflow or reduced assertions.
  • Focused SDK and Rust tests, formatting, scoped lint, and relevant release jobs pass.

Relationships

Metadata

Metadata

Assignees

No one assigned

    Labels

    planDurable planning issueplan:waitingDurable plan parked pending a decision, event, or non-issue condition; not for PR QA

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions