Stage regression testing

PCZT Stage Regression: compare disclosure changes across Zcash roles

Zerynex compares two local PCZT snapshots with the same Rust disclosure rules. This is an early developer guard for creator/prover/signer handoffs, not a cryptographic transaction equivalence verifier.

Try it in the browser

Open the PCZT Stage Comparator to compare two local PCZT snapshots with the same Rust policy engine compiled to WebAssembly. The page includes three checked-in synthetic scenarios (new proprietary metadata, transparent output topology, resolved metadata) and accepts your own local binary PCZT files, up to 8 MiB each. Values are not sent to a Zerynex server. You can export the resulting sanitized JSON report.

This is the browser implementation of selected structural checks, not a full transaction equivalence verifier.

CLI example: new sensitive metadata

cargo run --locked -- \
  apps/web/public/examples/clean-v6.bin \
  --compare-with apps/web/public/examples/flagged-v6.bin --json

Expected: a new PCZT-GLOBAL-METADATA finding and exit code 2. The report includes rule identifiers and field paths, never metadata values.

CLI example: unexpected change in component structure

cargo run --locked -- \
  apps/web/public/examples/clean-v6.bin \
  --compare-with apps/web/public/examples/transparent-v6.bin --json

Expected: transparent.outputs component count changes, exit code 2, even when the active policy permits transparent components.

Use --policy-file to apply an explicit disclosure policy. Both source and candidate use the same --input-format (binary or base64). The source may be "-" for stdin, while the candidate must be a path.

Python SDK

from zerynex_py import ZerynexClient, Policy
client = ZerynexClient(cli="./target/debug/zerynex")
result = client.compare_files(
    "apps/web/public/examples/clean-v6.bin",
    "apps/web/public/examples/flagged-v6.bin",
    Policy(deny_transparent=True),
)
print(result.agent_summary())

Reusable CI action

- uses: Oluwa-Laughter/zerynex@main
  with:
    baseline-path: tests/fixtures/before-synthetic.pczt
    pczt-path: tests/fixtures/after-synthetic.pczt
    fail-on-findings: 'true'

Review and pin the action to an immutable commit for real integrations. Use synthetic, nonspendable fixtures only. The action adds regression-count to its outputs (new rule findings and component-count changes). finding-count is the number of remaining candidate policy findings.

Even if no NEW finding appears, a pre-existing violation fails the policy check. No component-count drift is allowed in a passing comparison.

Explicit limitations

This can detect:

• newly present rule/field locations covered by existing disclosure checks;
• removed previously violating fields;
• changed component counts for six shielded/transparent component types.

It cannot detect:

• changed field values with the same structure or classification;
• changed recipient, amount, memo, randomness, commitment or signing keys;
• proof validity, consensus correctness or signing authorization;
• privacy leaks in logging, RPC traffic or transport behavior.

This is a partial acceptance test and should not be treated as an audit or an authorization to broadcast funds.