fix(consensus/XDPoS): keep only full verifications in the verified-header cache

[gzliudan]

Summary

XDPoS_v1.verifiedHeaders and XDPoS_v2.verifiedHeaders remember a header as verified, keyed by its hash, and every caller answers nil from the cache before running a single check. The fullVerify flag that selects how much of the verification runs is not part of that key, so a header admitted by a reduced check was remembered as if it had passed all of them - and the next caller asking for full verification was answered from the cache without ever looking at the validator signature, the double validation of v1, or the masternode set a v2 epoch-switch block claims for itself.

Both engines now only remember a header when the full verification passed. Consulting the cache under a reduced check stays correct, because passing the full set covers the reduced one.

 	err := x.verifyHeader(chain, header, parents, fullVerify)
-	if err == nil {
+	if err == nil && fullVerify {
 		x.verifiedHeaders.Add(header.Hash(), struct{}{})
 	}

Behaviour and risk

• The change only narrows what may be stored; cache hits under a reduced check are untouched, so it can only cause more verification work, never less - the fail-safe direction.
• The cost is that a header previously admitted by a reduced check is verified again when a full verification is requested for it. The set is small (a pruned sidechain segment, a reorg replay, the testnet shouldDisableFullVerify path), and the cache is an LRU bounded by utils.InMemorySnapshots.

Tests

consensus/XDPoS/engines/engine_v1/verify_header_test.go: TestVerifiedHeadersOnlyRemembersFullVerification turns on SkipV1Validation so that both levels succeed without a chain to read, which makes the verification level the only difference between the two calls - the reduced call must leave no entry behind, the full one must.

Removing the && fullVerify guard makes it fail with a reduced verification must not be remembered as a full one.

The v2 side is the symmetric change and has no dedicated test: building a header that passes the whole v2 verification is too expensive for a test of the cache.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 283664fc-6a3d-4abb-a2c0-e78c4914f087

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}

[Copilot]

🟡 Changes recommended

The v1 testnet override still allows an effectively reduced verification result to enter the cache.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

This PR prevents reduced XDPoS header verification results from being cached as fully verified.

Changes:

• Restricts v1 and v2 cache insertion to full verification.
• Adds a v1 regression test for reduced and full verification caching.

File summaries

File	Description
engine_v1/engine.go	Guards verified-header cache insertion.
engine_v1/verify_header_test.go	Tests cache behavior by verification level.
engine_v2/verifyHeader.go	Applies the equivalent v2 cache guard.

Review details

• Files reviewed: 3/3 changed files
• Comments generated: 1
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.