Gabriel39 commented on PR #68780: URL: https://github.com/apache/doris/pull/68780#issuecomment-6072639960
Restored lazy indexed page skipping and narrowed the validation scope to avoid sparse-scan overhead. Unselected pages no longer require a header read or page-cache lookup, and advancing an indexed nested V1 page no longer loads/decompresses/drains its remaining levels solely for validation. The reader records an unverified prefix and rejects a subsequent unsafe sequential fallback. Selected-page range reconciliation, physical-value accounting, and the metadata-only index coverage check remain in place. The scope is explicit: structurally accepted indexes remain authoritative for unread pages. This update removes the prior exhaustive skipped-prefix checks and their corresponding forged-span guarantees; it does not claim comprehensive detection of coordinated corrupt index/header metadata. The PR description now reflects that boundary and the remaining conservative fallback for index gaps/padding. Both added work-count regressions failed before the change. All 747 related BE ASAN tests now pass. The 64-page sparse seek parses two headers and makes three stream reads; partial V1 page advances neither read more bytes nor drain remaining levels. All eight affected C++ files pass clang-format 16. No legacy `be/src/format` code is changed. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
