fix(#375): make the Parity Gate check all four targets, and correct what it finds

[dcccrypto]

Part 3 of 3 for #375. The SDK half: the workflow fix, the corrected specs, and the two SDK gaps the gate exposed once it could actually run.

The state this fixes

The gate has been red on main every day since at least 2026-08-23, with three of four targets never running:

[percolator-prog]  cargo run failed:  error: failed to get `percolator` as a
                                      dependency of package `percolator-prog`
[percolator-stake] cargo run failed
[percolator-nft]   cargo run failed        <- fixed earlier
[percolator-match] parity OK

A permanently red gate is an ignored gate. ABI drift in prog and stake had no detection at all.

Workflow

Checks out the percolator engine, which percolator-prog compiles through path = "../percolator". Without it the prog fixture cannot build regardless of whether it exists — that was the second of prog's two failure causes.

Deliberately unpinned, for the reason percolator-prog's own ci.yml gives for leaving the engine unpinned while pinning the sibling programs: the engine is compiled into the wrapper, so its main is what the wrapper ships.

What the gate found

specs/wrapper-tags.json — missing twelve live instructions (tags 81–92), seven of which move money: WithdrawProtocolFee(84), WithdrawInsuranceReserveToStake(87), WithdrawCreatorFee(90), RebalanceLpVaultBacking(91), plus the fee-split and protocol-fee-authority setters. Gaps, names and every other tag already agreed.

specs/stake-parity.json — understated percolator-stake by fifteen. Tags 5–9 were in removed_tags, asserting they were gone; tags 19–28 were simply absent. All fifteen are live at 474079f, d0c6ecb and main alike. The layout half was already correct — all ten offsets and reserved_start return byte-identical, independent evidence the new fixtures are faithful.

SDK changes beyond the specs

• IX_TAG gains CancelRedemption(81) and UnwrapEscrowedPortfolio(82) — the only two v17 wrapper instructions with no entry. Both tags were represented solely by deprecated v12 names annotated "Not in v17", which was false; those two annotations now use the COLLIDES with v17 … wording the file already applies at 73/74/75/78/79/80/85/90.
• test/parity-fixtures.test.ts had hand-curated maps that agreed with the stale specs and disagreed with both programs — including a literal expect(fixture.removed_tags).toEqual([5, 6, 7, 8, 9, 11, 17]). STAKE_IX already carried all fifteen stake tags; only the spec and this subset were wrong, and they corroborated each other. Now [11, 17].

This is the third instance today of an assurance nobody could verify: a check, a spec and a test agreeing with one another while the program disagreed with all three.

Verified

check	result
pnpm test	33 files, 1034 passed, 0 failed
npx tsc --noEmit	clean
negative control	spec WithdrawProtocolFee 84 → 85 fails with WithdrawProtocolFee tag mismatch: expected 84 to be 85; restoring goes green — the gate is not vacuous
byte-equality	both new fixtures reproduce their committed spec exactly, which is what check-parity-fixtures.mjs asserts

Ordering

Land dcccrypto/percolator-stake#285 and dcccrypto/percolator-prog#460 first. In between, the gate reports DRIFT DETECTED rather than cargo run failed — already a better failure.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NgoNgagkvw7i5SSRC3FJ8D

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Team

Run ID: d5da3b2b-29b8-40b8-97cd-650f19d04697

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}