dennisthiessenandClaude Opus 5
247a7889b9
active_only tested delisted_on IS NULL, so a symbol dropped out of signals the moment a Form 25 was detected — ten days before Rule 12d2-2 makes the removal effective, while it was demonstrably still trading. A manual future-dated mark behaved the same way. It now compares against the database's own date, so a pending delisting stays live until the day it takes effect. That exposes a second problem the fix would otherwise create. Trading typically stops before the ten-day delay expires, so across that window the symbol is correctly active yet produces no bars — and confirm_delisting returned None for an already-marked row, which would have fired the staleness warning daily for ten days, the exact noise this flow exists to remove. It now reports the known effective date on every path where the delisting is established, so the caller warns only about gaps that are still unexplained. CURRENT_DATE renders identically on postgres and sqlite, and the OR is parenthesized when callers chain further where clauses. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>