Map chaos/diod HEAD and open issues against harness XFAIL

[ericvh]

Human Goal

Record the already-landed diod-currency mapping as a GitHub work slice. PR #3 classified every open chaos/diod issue/PR against v9fs/test diod/xfail.txt and confirmed diod-regression already builds chaos/diod master HEAD. This issue existed only as a draft PR because the previous agent could not POST issues.

Product follow-up (comments only, no new FAIL rows): docs/patches/v9fs-test-diod-upstream/
Mapping PR (merged): #3
Plan: docs/plans/diod-upstream-known-fail.md

This is not the t0011/t0013 implementation work. Those wait until issue #1 is closed after mapping closeout.

Evidence Anchor

• Source of truth: chaos/diod + v9fs/test diod/xfail.txt + Harness CI
• Pins:
  • chaos/diod de51d1ee1bd5ccf1d8c16b96227c8bb03ec50106 (still HEAD as of 2026-09-01)
  • harness e896a9b2db1e87dfeaaa1bb0208568469c8172c3
  • Actions 33511542559 unexpected: 0 (16/16)
• Ledger: docs/ledgers/diod-upstream.md (E0007, E0008, P0003); checker scripts/check-diod-upstream-inventory.sh
• Parallel lane: disjoint from issue Formulate first slice: map t0011/t0013 tip residuals (v9fs/test#28) #1 / PR Map t0011/t0013 tip residuals (M0 evidence) #2 files

Target Change

• Component: this forge's diod-upstream inventory (no diod/kernel/harness code in this slice)
• Behavioral contract: none of the 14 open chaos/diod items currently require a new machine-parsed XFAIL FAIL line; t0011/t0013 remain v9fs/test#28
• Invariants: comments in xfail.txt are ignored by eval; do not add synthetic FAIL rows

Proof Promotion

• Current level: Mapped + Static (PR merged; independent review of this slice still outstanding)
• Target level: Mapped + Static
• Exact commands: scripts/check-diod-upstream-inventory.sh; scripts/check-scaffold.sh

Complete this sentence:

The proof would fail if GitHub issue search were treated as a complete open-item list, because numbered GET found 14 open items (search showed only 3 PRs). Adding #163/#164/#166 as XFAIL FAIL lines would be wrong: the pinned run's FAIL strings are only t0011/t0013, and t0010 flock steps 21–24 PASS.

Review Focus

• Highest-risk semantics: inventing FAIL rows for upstream issues that the harness subset does not fail
• What could pass broad tests while still being wrong: trusting GitHub search completeness
• Rollback: leave machine-parsed diod/xfail.txt FAIL lines unchanged

Boundaries

• In scope: inventory, offline checker, comment-only product patch for later v9fs/test landing
• Explicitly not claimed: diod/kernel/harness code; t0011/t0013 fixes; Debian apt diod; floating DIOD_REF=master as a SHA pin
• Deferred: product comment PR on v9fs/test when that write scope is free; copy E0007/P0003 into main evidence/proof tables; independent review of PR Map chaos/diod HEAD and open issues against harness XFAIL #3

Backlog Position

• Rank: parallel to Formulate first slice: map t0011/t0013 tip residuals (v9fs/test#28) #1 (already merged)
• Predecessor: none for this inventory; does not unblock t0011/t0013 implementation
• Unblocks: optional v9fs/test comment PR
• Dependency boundary: disjoint write scope from PR Map t0011/t0013 tip residuals (M0 evidence) #2
• Proof boundary: Unmapped → Mapped + Static
• Parallel lane: diod-upstream vs tip-residual map
• Integration order: already on main via PR Map chaos/diod HEAD and open issues against harness XFAIL #3
• Next issue: independent review of PR Map chaos/diod HEAD and open issues against harness XFAIL #3; do not open t0011/t0013 implementation issues until Formulate first slice: map t0011/t0013 tip residuals (v9fs/test#28) #1 is closed

Ownership and Authority

• Authority level/exceptions: A2; no merge to v9fs/linux; no .github on linux
• Orchestrator: ericvh
• Evidence mapper: PR Map chaos/diod HEAD and open issues against harness XFAIL #3 mapping agent
• Implementer/write scope: this forge (docs/ledgers/diod-upstream*, checker, comment patch)
• Independent reviewer: still required (PR Map chaos/diod HEAD and open issues against harness XFAIL #3 merged without a review-note)
• CI/proof owner: v9fs/test workflows (run 33511542559)
• State closer: after independent review of this slice

Durable State

• Evidence-map rows: E0007/E0008 in docs/ledgers/diod-upstream.md (not yet copied into evidence-map.md)
• Proof-ledger rows: P0003 in the same file
• Decision/experiment/risk rows: X0001, R0003
• Methodology trigger: none

Done

• Evidence source is pinned.
• Target component and behavioral contract are named.
• Positive and negative proof plan is explicit.
• Mapping PR merged (Map chaos/diod HEAD and open issues against harness XFAIL #3).
• Independent review of PR Map chaos/diod HEAD and open issues against harness XFAIL #3
• Follow-up product comment PR on v9fs/test (when write scope is free)

[ericvh]

Independent review of PR #3 is posted: #3 (comment)

Decision: approved. Labels are review:approved + status:landed. PR #3 was already merged.

Remaining follow-up (not claimed by this slice): comment-only v9fs/test diod/xfail.txt patch in docs/patches/v9fs-test-diod-upstream/ is still deferred until that write scope is free. Do not add FAIL rows. Optional DIOD_REF SHA pin waits on the same product PR.

Forge ledger copy of E0007/E0008/P0003: draft #7 (human merge only; this agent does not merge).

This inventory does not unblock t0011/t0013. Next implementer: #5