task_021026_19 — P3 Deploy Blocker + Open Security/dincli Findings: EIP-170 Size Gate (#201 A), onlyCurrentGI (#206), Sender-Bound Audit Commit (#192), Auditor Commit Retry (#202), Test-Data Dispute Drain (#205), Wiki DIN-Rep (#207) (Umer Majeed)

[umermjd11]

cc @umeradl. This is the tracking discussion for task_021026_19. It's opened late: all six Parts have already landed. It records the status, PR links and the follow-ups in one place, following the convention of the earlier task-tracking discussions.

Full spec: Developer/tasks/task_021026_19.md
Source plan: task-plan-021026-1.md (reviewed in PR #209, with the reviewer decisions; forwarded in PR #210)

Timeline

Planned: Oct 2 – Oct 12, 2026. Actual: Oct 2 – Oct 5, 2026.

Status

Part	Issue	Delivered	State
A: EIP-170 size gate, DINTaskCoordinator back under budget	#201 Part A	PR #211, commit 1 (merged 2026-10-02, follow-up 420d48d)	✅ Landed. #201 stays open for its Part B
B: onlyCurrentGI on registerDINaggregator	#206	PR #211, commit 2	✅ Closed
C: sender- and slot-bound audit commit hash	#192	PR #212 (merged 2026-10-02, follow-up 7785acf)	✅ Closed
D: auditor commit salt kept across retries; aggregate-t2 batch id	#202	PR #214 (merged 2026-10-05, follow-up e491f0b)	✅ Closed
E: test-data dispute reward-pool drain	#205	PR #215 (merged 2026-10-05, follow-up 740a613: no dispute penalty after settleRewards)	✅ Closed
F: wiki DIN-Representative page	#207	Wiki rewritten (diff, #207 comment)	✅ Done. #207 is open for the maintainer to close

Decisions applied (PR #209)

1. Size gate: CI fails below 1,024 B of EIP-170 headroom and warns below 2,048 B. At merge, DINTaskCoordinator was 23,497 B (1,079 B margin) and DINTaskAuditor 22,718 B (1,858 B, in the warning band).
2. Parts A + B: one PR, two commits.
3. security: anyone can drain a GI reward pool through test-data disputes (unauthenticated resolveTestDataDispute, zero default bond) #205: options 1+2. Only a batch auditor can open a dispute, only the owner resolves it, an expired dispute is upheld, and the default bond is 100 * 1e18 DIN.
4. View-ABI change accepted: the per-aggregator getters were folded into getAggregatorSubmission, and the dincli callers were updated. A note was posted on PR feat(indexer): implement DIN Protocol subgraph — platform + task-level contracts #29 for the subgraph ABI.

Part F scope note

The wiki update was widened from the DIN-Representative page to the whole wiki. Every page now opens with a banner: "This wiki describes DevNet 2.0, soon to be launched", linking #102. The pages were checked against develop @ 740a613.

Follow-ups (not in this task)

• Contract size headroom for all contracts (DINTaskCoordinator −9 B over EIP-170 with PR No. 197) + slash reason for committed-but-unrevealed across commit-reveal flows #201 Part B (slash reason for committed-but-unrevealed), security: S5 recidivism is per-model, so a validator can spread missed votes across models and never escalate #193, P3 dispute: upheld S4 dispute slashes colluders but never replaces the wrong finalCID or refunds the honest dissenter #194, security: dual-role registration allows one operator to control both the auditor filter and T1 aggregator in the same GI #180, P3 dispute: decentralize adjudication — owner-adjudicator trust assumption must be removed before mainnet #181, BL-11 follow-up: mainnet-grade randomness (VRF / multi-party commit-reveal) for dispute subgroup seed #178: listed as out of scope in the spec.
• dincli gaps before launch:
  • model-owner deploy still passes the old constructor arguments, without modelId.
  • No commands yet for depositing or claiming rewards, registerEncryptionKey, or disputes.
• Repo docs drift found while writing the wiki:
  • din-workflow.md still lists 4 contracts and withdraw.
  • setup.md has the broken @main#subdirectory=dist install line.
  • Old command names remain in roles/clients.md, roles/auditors.md, roles/model-owner.md and model-workflow.md.
  • ROADMAP.md:19 says "DevNet 3.0".
• Spec nits: the "9.5 days" estimate, and Part F's mintEmission wording.

Questions or comments on any Part go here. A small docs PR will replace the spec's "Discussion: TBD" with a link to this thread.

🤖 Generated with Claude Code

[umermjd11]

Spec-link PR: #217. It links this discussion in task_021026_19.md, ticks the Part A–F deliverables, and fixes the two review nits.

[umeradl]

Closing: task_021026_19 complete

Every deliverable in task_021026_19.md was checked against origin/develop and the wiki itself, not against this thread. All of it has landed. PR No. 217 (this discussion's link, plus the ticked boxes) merged as a1fcce2; its develop push run is green.

Part	Delivered by	Evidence on develop @ a1fcce2
A: issue No. 201 Part A, EIP-170 gate	PR No. 211: 84d0eef + 420d48d	contract_size_gate.py runs in CI (ci.yml:69). The 1,024 / 2,048 B thresholds are in CONTRIBUTING.md, and anvil.sh documents --code-size-limit. getAggregatorSubmission (DINTaskCoordinator.sol:800) replaces the per-aggregator getters; dincli aggregator.py uses it. The ABI note is posted on PR No. 29. Issue No. 201 stays open for its Part B (out of scope).
B: issue No. 206	PR No. 211	registerDINaggregator(uint _GI) public onlyCurrentGI(_GI) (DINTaskCoordinator.sol:401). Issue No. 206 is closed.
C: issue No. 192	PR No. 212: 9fb6507 + 7785acf	abi.encode(score, vote, salt, msg.sender, gi, batchId, modelIndex) (DINTaskAuditor.sol:1172), _audit_commit_hash in dincli/cli/auditor.py, plus foundry/test/utils/AuditCommitHash.sol and tests/test_auditor_commit_hash.py. Issue No. 192 is closed.
D: issue No. 202	PR No. 214: d2d2213 + e491f0b	The hasCommittedLM skip and cache-before-send in evaluate_lms, plus T2-named aggregate-t2 work, with 6 tests. Issue No. 202 is closed.
E: issue No. 205	PR No. 215: ac6f8f9 + 740a613	Owner-only resolve, an unanswered dispute upheld on expiry, batch-auditor-only open, and a 100 * 1e18 default bond, with all the tests the spec lists. Issue No. 205 is closed.
F: issue No. 207, wiki	Wiki 4ea2744 + 6a20a3b	All 8 rows checked on a fresh wiki clone. Issue No. 207 is closed.

Deliberate deviation: at the PR No. 215 merge, 740a613 added a fix on top. A dispute can't be opened on a GI whose rewards are settled, and one upheld after settlement takes no pool penalty. Without it, an upheld dispute after settleRewards left the GI's claims underfunded. It goes beyond Part E's spec, but its own tests pin it.

Final numbers on develop: forge test 491 passed. pytest -m "not integration" 395 passed. Size gate green: DINTaskCoordinator 23,497 B (1,079 B margin) and DINTaskAuditor 22,718 B (1,858 B), both in the warning band.

The follow-ups listed in the opening post (issue No. 201 Part B, the dincli launch gaps, the repo-docs drift) are outside this task; the dincli gaps and docs drift aren't tracked in issues yet. Closing as resolved.