Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request removes the project's linter configuration and updates the Renovate settings. Specifically, it introduces a vulnerabilityAlerts configuration to automate security updates while disabling general automerge for patches and digests. Feedback indicates that the removal of the automerge property from the packageRules section has left a redundant configuration block that should be cleaned up to maintain a concise configuration file.
| ] | ||
| } | ||
| ] | ||
| ], |
There was a problem hiding this comment.
The packageRules entry (lines 11-20) is now redundant because it contains no configuration properties (such as automerge or groupName) after the removal of the automerge setting. Additionally, note that disabling automerge for patch and digest updates will require manual intervention for all dependency updates, which may increase maintenance overhead. Consider either removing the redundant block entirely or restoring the automerge setting for non-major updates if automatic updates are still desired.
Three issues, all valid. #237 Codex P1 -- preserve item generation when grouping rows. Item keys carry !ddb|item|<table>|<generation>|... but the prior parser dropped the generation and Finalize merged all generations under the active schema. In real clusters, in-flight delete/recreate cleanup leaves stale-gen rows visible in the snapshot for a window; those rows would silently restore as items under the new schema (potentially resurrecting deleted data, or failing Finalize when PK names changed across generations). Fix: - parseDDBItemKey now returns (encodedTable, generation). - ddbTableState.items -> itemsByGen map[uint64][]*pb.DynamoItem. - flushTable filters by st.schema.GetGeneration() and emits a ddb_stale_generation_items warning carrying the count and active gen so the operator can correlate the orphan window. #182 Gemini -- fail-fast on flush errors. Finalize previously deferred the first error and continued through remaining tables. Real flush errors (out of disk, bad permissions) should surface immediately, not be misattributed to a later table whose flush also fails. Orphan-table warnings still continue (those are informational, not errors). #442 Gemini -- empty set serializes as [] not null. setAttributeValueToPublic now uses make+append so nil/empty SS/NS/BS become "{\"SS\":[]}" rather than "{\"SS\":null}". Downstream tools that distinguish "present-but-empty" from "missing" no longer see a misleading null. Tests: TestDDB_StaleGenerationItemsExcludedAndWarned, TestDDB_EmptyStringSetSerializesAsEmptyArrayNotNull, TestDDB_ParseItemKeyExtractsGeneration. Existing fixtures updated to set Generation=1 on schemas (matching the item key gen).
1 participant
No description provided.