Map chaos/diod HEAD and open issues against harness XFAIL

[ericvh]

Human Summary

• Behavior/outcome changed: harness diod-regression is confirmed on current chaos/diod master HEAD, and every open chaos/diod issue/PR is classified against v9fs/test known-fail. None of those upstream items currently need a new machine-parsed XFAIL FAIL line.
• Mechanism or implementation: pinned inventory (docs/ledgers/diod-upstream.md + JSON snapshot), offline checker, comment-only product patch for later landing on v9fs/test.
• Review focus: GitHub search is not a complete issue list; do not add synthetic FAIL rows for #163/#164/#166.
• Proof result: Mapped + Static. scripts/check-diod-upstream-inventory.sh passed; dropping #164 from the ledger fails the checker; scripts/check-scaffold.sh passed.
• Known gaps not claimed: no diod/kernel/harness code change; t0011/t0013 remain v9fs/test#28 (PR Map t0011/t0013 tip residuals (M0 evidence) #2); Debian apt diod / diod-9p2000.L is not in CI; floating DIOD_REF=master is not a SHA pin.

Evidence and Target

• Related issue: not created (this token cannot POST issues). Product tracker for current XFAIL rows: Tip XFAIL residuals: t0011 allsquash and t0013 ACL test#28
• Parallel to: Map t0011/t0013 tip residuals (M0 evidence) #2 (does not edit that PR's files)
• Milestone: M0 - Bootstrap
• Evidence version and paths/scenarios:
  • chaos/diod de51d1ee1bd5ccf1d8c16b96227c8bb03ec50106 (2026-06-23; still HEAD 2026-09-01)
  • harness e896a9b2db1e87dfeaaa1bb0208568469c8172c3
  • Harness CI 33511542559 unexpected: 0 (16/16)
• Target components/APIs/artifacts: this forge's diod-upstream ledger and checker; proposed diod/xfail.txt comments only

Proof Movement

• Before: Unmapped (diod treated as floating master; outstanding chaos/diod issues not compared to XFAIL)
• After: Mapped + Static
• Discriminating observation: trusting GitHub issue search would miss 11 open issues (including #163 and #164). Adding those as XFAIL FAIL lines would be wrong because the pinned run's FAIL strings are only t0011/t0013, and t0010 flock steps 21–24 PASS.
• Exact commands/artifacts and results:
  • scripts/check-diod-upstream-inventory.sh → passed (14 open items, 16 XFAIL FAIL lines, diod de51d1ee1bd5)
  • negative: remove #164 from the ledger → checker fails
  • scripts/check-scaffold.sh → passed
  • git diff --check → clean

Independent Review

• Creator: this mapping/implementer
• Reviewer: distinct from creator (required)
• Review note: pending
• Decision: pending
• Resolved blockers/questions: none
• Unresolved objections: none yet

Follow-Ups

• Land docs/patches/v9fs-test-diod-upstream/xfail.txt as comments on v9fs/test diod/xfail.txt when that write scope is free. Trigger: test-repo push access. Do not add FAIL rows.
• Copy E0007/P0003 into main evidence-map.md / proof-ledger.md after PR Map t0011/t0013 tip residuals (M0 evidence) #2 lands (that PR owns E0003–E0006).
• Optional: pin DIOD_REF to de51d1ee1bd5ccf1d8c16b96227c8bb03ec50106 instead of floating master. Trigger: chaos/diod HEAD moves, or a desire for reproducible CI.
• Re-run the inventory when chaos/diod HEAD or open_issues_count changes.

Durable State Updated

• Evidence map (docs/ledgers/diod-upstream.md, disjoint from PR Map t0011/t0013 tip residuals (M0 evidence) #2)
• Proof ledger (P0003 in the same file)
• Decision/experiment/risk records as applicable (X0001, R0003)
• Methodology trigger (none; not a repeated blocker; first-slice count unchanged)
• Compact audit-report link: docs/reports/diod-upstream-known-fail.md

Closeout

• Current GitHub status/proof/review labels: intended kind:feature, area:test, proof:mapped, status:review, review:needs-independent (apply if API allows)
• Next dependency-ordered issue: independent review of this PR; PR Map t0011/t0013 tip residuals (M0 evidence) #2 remains the t0011/t0013 map; product comment PR on v9fs/test after review
• State closer: after independent review

  

[ericvh]

Work-slice issue for this already-merged mapping: #4

Labels applied as intended (kind:feature, area:test, proof:mapped, status:review, review:needs-independent, milestone M0). Independent review of this PR is still outstanding. This is not the t0011/t0013 implementation work.

[ericvh]

Review: approved
Focus: whether open chaos/diod items were classified against the machine-parsed XFAIL surface without inventing FAIL rows, and whether the inventory would still fail if GitHub search/list were treated as complete.

Comments:

• suggestion: Copy E0007/E0008 and P0003 from docs/ledgers/diod-upstream.md into docs/ledgers/evidence-map.md and docs/ledgers/proof-ledger.md. PR Map t0011/t0013 tip residuals (M0 evidence) #2 has landed, so that deferred merge is now unblocked. Reconsideration trigger: this closeout (draft PR on the forge; do not merge from this review).
• suggestion: Land docs/patches/v9fs-test-diod-upstream/xfail.txt as comments only on v9fs/test diod/xfail.txt when that write scope is free. Do not add FAIL rows. Reconsideration trigger: an assigned implementer with v9fs/test push, after t0011/t0013 are not also writing diod/xfail.txt.
• nit: With a PAT that can read issues, GET /repos/chaos/diod/issues?state=open currently returns the same 14 items. Search is:pr is:open still returns only the three PRs (#31, #91, #165), and a combined is:open search 422s unless is:issue or is:pull-request is set. Numbered GET remains the robust enumerator; X0001 is still right about not trusting PR-only search. Not blocking.
• praise: The checker fails if diod server: infinite loop chaos/diod#164 is dropped from the ledger, and the product patch is comments-only. That is the right discriminator for “do not hide #163/#164/#166 behind synthetic XFAIL FAIL lines.”

Tests/proof requested:

• Reconstructed from run 33511542559 artifacts, not from the mapping text: empty diod-xfail/unexpected.txt; qemu.log diod-regression-eval: unexpected: 0 (16 FAIL lines, 16 XFAIL baseline); FAIL/ERROR strings are only t0011/t0013; t0010 flock steps 21–24 PASS; t0012 PASS (xattr SKIP). Live chaos/diod master is still de51d1ee1bd5ccf1d8c16b96227c8bb03ec50106. All 14 pinned items are still state: open. scripts/check-diod-upstream-inventory.sh passed; dropping #164 from a ledger copy failed the checker.

Evidence appendix:

• Reviewer role: independent Cloud Agent session, distinct from the mapping commit (7804f1e, Cursor Agent). GitHub will attribute this comment to the PAT identity (ericvh); that is token identity, not self-approval of the PR Map chaos/diod HEAD and open issues against harness XFAIL #3 creator.
• Scope inspected: issue Map chaos/diod HEAD and open issues against harness XFAIL #4; PR Map chaos/diod HEAD and open issues against harness XFAIL #3 body/diff/files (diod-upstream ledger, JSON snapshot, xfail snapshot, checker, comment-only product patch, plan, report; no product code in v9fs/linux or v9fs/test); live chaos/diod HEAD and numbered GET 31–166; run 33511542559 artifacts (diod-regression-logs).
• Commands/source refs/report links:
  • gh api repos/chaos/diod/commits/master → de51d1ee1bd5ccf1d8c16b96227c8bb03ec50106
  • numbered GET of 14 open items; search is:pr is:open → 3 PRs only
  • run https://github.com/v9fs/test/actions/runs/33511542559 head e896a9b2db1e87dfeaaa1bb0208568469c8172c3
  • bash scripts/check-diod-upstream-inventory.sh → passed (14 open items, 16 XFAIL FAIL lines, diod de51d1ee1bd5)
  • negative: remove diod server: infinite loop chaos/diod#164 from a ledger copy → checker fails
  • bash scripts/check-scaffold.sh → passed (25 required files, 4 skills)
  • PR Map chaos/diod HEAD and open issues against harness XFAIL #3 already merged by ericvh at 2026-09-01T15:51:00Z (8f9ad84). This note does not re-merge and does not start t0011/t0013 product implementation.