EPS
← All tasks·#933Completed

workflow-fix: test_shared_vm_thread_caps.py red on main (stale #895 grandfather + ungrandfathered violators)

kind: infra#wf-fix#wf-fix-fp:d51df851dd1c
track:

Highlight text on any card (body, plan, or an event) to anchor a comment, or leave a whole-task comment here. Mention @claude to summon a reply.

No comments yet.

Overview / Motivation

Auto-filed by the workflow-fix-on-bug protocol from a workflow-fix candidate raised on task #922 (emitting agent: experiment-implementer, round 1).

Goal

Make tests/test_shared_vm_thread_caps.py green on main: delete the fixed entry from GRANDFATHERED_895 and either fix or explicitly grandfather (with issue refs) the current main-side violators.

Workflow gap

  • Bug observed: test_grandfathered_895_block_is_current and test_no_new_torch_before_dotenv_vm_entrypoints FAIL on current main — scripts/issue779_heldout_recon_scatter.py was fixed (dotenv now precedes torch) but never removed from GRANDFATHERED_895, and several post-#895 entrypoints (issue779_arm_headline_summaries2.py, issue810_analyze.py, ...) violate the detector without being grandfathered.
  • Why it is a workflow gap: a workflow-invariant test that is red on main FAILs every task's Step 9c test-verdict regardless of the diff under review, training sessions to dismiss it as noise.
  • Confidence (emitter): high

Proposed change (candidate diff sketch — refine in planning)

  • "scripts/issue779_heldout_recon_scatter.py", # in GRANDFATHERED_895 — root cause fixed on main
  • (removed — dotenv-before-torch fixed on main)

  • plus: reorder load_dotenv() before torch/numpy in the listed violators OR add them

  • to GRANDFATHERED_TORCH_BEFORE_DOTENV with their owning-issue refs.

Scope / surfaces

  • Primary target: tests/test_shared_vm_thread_caps.py
  • Grep the workflow surface for the pattern before editing (grep -rln 'GRANDFATHERED_895' .claude/ CLAUDE.md scripts/ tests/) and update every hit; list them in the plan. The violator scripts themselves (scripts/issue_.py) are experiment entrypoints — the fix there is the mechanical dotenv-before-torch reorder OR an explicit grandfather entry, whichever the planner judges lower-risk per file.

Constraints / invariants

  • Workflow-surface only for the test file; violator-script edits limited to the import-order reorder.
  • Full test file green on main after the change; ruff passes.
  • This session runs under EPM_WORKFLOW_FIX_SESSION=1 and carries a workflow_fix_target: Provenance line — it MUST NOT auto-route any of its own subagents' workflow-fix candidates (recursion guard).

Provenance

  • workflow_fix_target: tests/test_shared_vm_thread_caps.py
  • fingerprint: d51df851dd1c

target_file: tests/test_shared_vm_thread_caps.py bug_observed: test_grandfathered_895_block_is_current and test_no_new_torch_before_dotenv_vm_entrypoints FAIL on current main — scripts/issue779_heldout_recon_scatter.py was fixed (dotenv now precedes torch) but never removed from GRANDFATHERED_895, and several post-#895 entrypoints (issue779_arm_headline_summaries2.py, issue810_analyze.py, ...) violate the detector without being grandfathered. why_workflow_gap: a workflow-invariant test that is red on main FAILs every task's Step 9c test-verdict regardless of the diff under review, training sessions to dismiss it as noise. proposed_change: delete the fixed entry from GRANDFATHERED_895 and either fix or explicitly grandfather (with issue refs) the current main-side violators so the suite is green on main. confidence: high related_task: #922

Activity