fix(crisp): account for remainder in multi-option segment boundaries

[cedoor]

Closes #1344

---

• Add remainder offset to start_idx when D is not evenly divisible by num_options in check_coefficient_values_with_balance and check_coefficient_zero
• TypeScript puts remainder zeros at end; after k1.reverse() they appear at start — use start_idx = remainder + (circuit_segment * segment_size) to align segments
• Fixes total_votes <= balance constraint failure for 3+ options with D=512 (remainder=2)
• Add generateTestLeaves helper and refactor tests to use dynamic leaves from (address, balance) pairs
• Replace hardcoded LEAVES with generateTestLeaves in vote and utils tests

Summary by CodeRabbit

• Bug Fixes

  • Corrected index computation and remainder handling in circuit utilities to support non-divisible cases.

• Contracts

  • Updated verifier verification key and related constant for proof verification.

• Tests

  • Replaced static test leaves with a generateTestLeaves helper and SLOT_ADDRESS constant; updated tests to use dynamic leaves and adjusted vote inputs and proofs.

• Chores

  • Bumped package and client SDK versions across example packages.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

2 Skipped Deployments

Project	Deployment	Actions	Updated (UTC)
crisp	Skipped		Feb 18, 2026 10:50am
enclave-docs	Skipped		Feb 18, 2026 10:50am

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

---

📝 Walkthrough

Walkthrough

Adds remainder-based offset to circuit segment indexing to handle non-divisible polynomial degrees; introduces test helper and SLOT_ADDRESS constant replacing static LEAVES; updates tests to use generated leaves; updates verifier verification key and bumps several package versions.

Changes

Cohort / File(s)	Summary
Circuit Index Correction examples/CRISP/circuits/src/utils.nr	Compute remainder = D - (segment_size * num_options) and use start_idx = remainder + (circuit_segment * segment_size) in check_coefficient_values_with_balance and check_coefficient_zero; add tests for non-divisible D.
Test Constants & Helpers examples/CRISP/packages/crisp-sdk/tests/constants.ts, examples/CRISP/packages/crisp-sdk/tests/helpers.ts	Remove static LEAVES; add SLOT_ADDRESS constant and new generateTestLeaves(entries, minSize) helper that builds/pads Merkle leaves using hashLeaf.
Test Suite Updates examples/CRISP/packages/crisp-sdk/tests/utils.test.ts, examples/CRISP/packages/crisp-sdk/tests/vote.test.ts	Replace hard-coded leaves/slot with generateTestLeaves and SLOT_ADDRESS; adjust tests to use generated leaves and updated vote tuple/inputs.
Verifier & Packaging examples/CRISP/packages/crisp-contracts/contracts/CRISPVerifier.sol, examples/CRISP/**/package.json	Replace verifier VK coordinates and update VK_HASH; bump package versions (crisp-contracts, crisp-sdk, crisp-zk-inputs, client) to 0.5.11.

Sequence Diagram(s)

(omitted)

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Possibly related PRs

• chore: check vote is zero for masking #938: Modifies the same Noir utils (examples/CRISP/circuits/src/utils.nr) affecting coefficient checks.
• fix: crisp circuit validation of encoded vote #973: Related adjustments to coefficient segment/indexing and layout handling in the circuit utils.
• docs: update crisp circuit documentation #1132: Another PR updating check_coefficient_values_with_balance / check_coefficient_zero in the same file.

Suggested reviewers

• ctrlc03
• 0xjei

Poem

🐇 I hopped through bits and found the gap,
Remainders tucked back into their map,
Leaves now sprout when tests request,
Slot addresses dressed in their best,
Circuits sing — I nibble a nap.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix(crisp): account for remainder in multi-option segment boundaries' directly and clearly describes the main change: adding remainder offset handling for segment boundary calculations when D is not divisible by num_options.
Linked Issues check	✅ Passed	All coding requirements from issue #1344 are met: remainder offset added to start_idx calculation in check_coefficient_values_with_balance and check_coefficient_zero, with supporting test cases validating the fix for non-divisible D scenarios.
Out of Scope Changes check	✅ Passed	Changes include in-scope fixes (remainder offset in Noir code), necessary test refactoring (generateTestLeaves helper, dynamic leaves), and version bumps aligned with the release cycle; no extraneous modifications detected.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix/multi-option-crisp

---

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 and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@examples/CRISP/circuits/src/utils.nr`:
- Line 32: Add a targeted regression test that exercises the non-divisible D
case so the new remainder logic is actually used: create a unit test that sets D
and num_options such that D % num_options != 0 (e.g., D=100, num_options=3) and
supplies a non-zero vote/coefficient so start_idx (computed from remainder,
segment_size, circuit_segment) must be correct; verify the circuit output or
index access that previously could go out-of-bounds. Update or add this test
alongside the existing tests (e.g., augment test_zero_vote_passes /
test_check_coefficient_zero_passes or add a new test) and ensure it runs for the
code paths using remainder, D, segment_size, num_options, start_idx and any
k1.reverse() logic; replicate the same regression test for the other occurrences
noted (lines referenced around 44, 105, 110) so all similar fixes are exercised.

[ctrlc03]

<3

[coderabbitai]

🧹 Nitpick comments (1)

examples/CRISP/packages/crisp-sdk/tests/vote.test.ts (1)

37-37: zeroVote uses 2 options while vote uses 3 — intentional but potentially fragile.

getZeroVote(2) on line 37 is used for previousCiphertext (line 69) and the mask vote tests (numOptions: 2 on lines 265, 288), while the real vote [10n, 0n, 0n] has 3 options. These are independent test scenarios so it's currently fine, but if someone later unifies numOptions across tests, the zeroVote(2) will silently produce the wrong-sized ciphertext.

Consider deriving zeroVote from a shared constant or adding a brief comment clarifying the intentional mismatch.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@examples/CRISP/packages/crisp-sdk/tests/vote.test.ts` at line 37, zeroVote is
created with getZeroVote(2) while the real vote array (vote) has 3 options,
which may lead to silent size mismatches later; update tests to derive zeroVote
from a shared constant or the same source used for vote (e.g., replace
getZeroVote(2) with getZeroVote(NUM_OPTIONS) or construct NUM_OPTIONS and use it
for both zeroVote and vote), and/or add a short comment next to getZeroVote(2)
explaining the intentional mismatch referencing zeroVote, getZeroVote, vote,
previousCiphertext, and numOptions so future maintainers won’t accidentally
unify sizes.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@examples/CRISP/packages/crisp-sdk/tests/vote.test.ts`:
- Line 37: zeroVote is created with getZeroVote(2) while the real vote array
(vote) has 3 options, which may lead to silent size mismatches later; update
tests to derive zeroVote from a shared constant or the same source used for vote
(e.g., replace getZeroVote(2) with getZeroVote(NUM_OPTIONS) or construct
NUM_OPTIONS and use it for both zeroVote and vote), and/or add a short comment
next to getZeroVote(2) explaining the intentional mismatch referencing zeroVote,
getZeroVote, vote, previousCiphertext, and numOptions so future maintainers
won’t accidentally unify sizes.