#261 Background catalogue text refresh can fail with "Broken timeline index" while timelines are deleted
–
- github
- github/openchronology/timeline-app
- github/openchronology/timeline-app#261
- issue
- state:closed
Seen while running `test/shared-history-postgres.mjs` repeatedly (2 of 6 runs): a caught background error is logged before the test passes: ``` Catalogue text refresh failed: error: Broken timeline index ``` It comes from the catalogue text refresh scheduled after a copy (`store.refreshText`, server/store.mjs around line 1015) reading through `oc_event_text` → `oc_events_lineage`, apparently racing with timelines (or the index rows they read through to) being deleted at the end of the test. The test doesn't fail, and it isn't known yet whether it happens outside tests or predates #259. A refresh for a timeline that is gone, or whose read-through source went, should end quietly rather than log a broken index. https://github.com/openchronology/timeline-app/issues/261
Related
- #259 Owners can fork their own timelines repeatedly, and see Propose changes on them
- #262 Owners duplicate, others fork, once; propose only without write access (#259)
- #284 The catalogue text rebuild reads one snapshot
- #285 Rebuilding a large fork's catalogue text after many upstream saves is slow
- The catalogue text rebuild reads one snapshot