navidrome/db
Deluan d3a1395896 fix(pglite): flush output left after ReadyForQuery before the next request
PGlite's interactive_one discards an incoming request when output from
the previous frame is still pending ('flush after frame'), which showed
up as replies without ReadyForQuery, first after handshakes and then in
bursts mid-scan, each one breaking a pooled connection and the scanner
transaction on it. After a complete reply the bridge now runs one empty
tick and appends anything the backend still had, and logs the message
tags of any reply that is still incomplete.
2026-09-05 01:37:39 -04:00
..
migrations perf(pglite): artist stats as hash joins, one statement for a full refresh; bigint library size 2026-09-04 23:23:50 -04:00
pglite fix(pglite): flush output left after ReadyForQuery before the next request 2026-09-05 01:37:39 -04:00
backup.go feat(pglite): run Navidrome on the embedded PostgreSQL end to end 2026-09-04 20:41:52 -04:00
backup_test.go chore(deps): upgrade to Go 1.27 (#5990) 2026-08-30 12:43:15 -04:00
db.go perf(pglite): keep pooled connections idle instead of churning them 2026-09-05 00:08:43 -04:00
db_test.go feat(bfr): Big Refactor: new scanner, lots of new fields and tags, improvements and DB schema changes (#2709) 2025-02-19 20:35:17 -05:00
export_test.go perf(db): keep query planner statistics trustworthy with full ANALYZE (#5740) 2026-07-13 12:04:29 -04:00
optimize.go feat(pglite): checkpoint after ANALYZE at the end of a scan 2026-09-05 01:14:48 -04:00
optimize_test.go perf(db): keep query planner statistics trustworthy with full ANALYZE (#5740) 2026-07-13 12:04:29 -04:00