Skip to content

test(harness-config, harness-ops, harness-memory, plugin-quality): suites fail on Windows from path form and TMPDIR #6507

Closed

Description

@kyle-sexton

Part of #6020. Found while verifying #6506 (#6484) on Windows. Every failure below also fails on the base branch, so none comes from #6506.

Failing cases (Windows Bash only)

  • harness-config audit-engine.test.sh: case 4b (2 of 471 assertions).
  • harness-ops changelog-status.test.sh: cases [19], [32], [65] (3 of 77), path form (a D:/ path where the assertion expects the POSIX form, or the reverse).
  • harness-ops audit-install-state tests: the Python loader fails on the path it is handed.
  • harness-config check-hook-coverage test.
  • harness-memory nested-agents-check (6 of 36) and instruction-load-stats (3 of 50).
  • plugin-quality collect-categories test.

Cross-cutting cause

A TMPDIR in D:/... form breaks tests that put a curl shim on PATH from a temp directory: the drive colon splits the PATH entry, so the shim is never found and the real curl (or none) runs. With TMPDIR in POSIX form (/d/...) these tests resolve the shim.

Related: #6135 (the same failure class in source-control and playbooks).

Done when

Each listed suite passes on Windows Bash with TMPDIR in either form, or a case that cannot is skipped on Windows with a stated reason.

Activity

  1. added
    needs-triageNot yet classified. Floor until a type and one priority tier are set.
    on Oct 7, 2026
  2. github-actions commented on Oct 7, 2026

    @github-actions
    Contributor

    Thanks for the detailed write-up. I understood this as a Windows-Bash-only test failure report covering six suites across harness-config, harness-ops, harness-memory and plugin-quality, all failing on the base branch (not introduced by #6506), with a clear cross-cutting cause: a TMPDIR in D:/... form splits PATH at the drive colon, so a curl shim placed on PATH from a temp dir is never found. The "Done when" criteria (each suite passes under either TMPDIR form, or is skipped on Windows with a stated reason) are clear enough to act on without further scoping.

    This looks closely related to #6135, which reports the same failure class (path-form and line-ending mismatches on Windows Bash) in source-control and playbooks test suites, though it covers different plugins/suites so doesn't appear to be a duplicate.

    No missing information found on my end.

    Automated triage. A maintainer reviews it.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    needs-triageNot yet classified. Floor until a type and one priority tier are set.priority: mediumReal value, no hard deadline; normal backlog flow.

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions