• v1.98.0 68dce78d04

    v1.98.0 — Pinky OOOO: the fixture-id hardening + the timeline fold
    All checks were successful
    Checks / Functional checks (push) Successful in 22s
    Stable

    ClaudiuSchuster released this 2026-10-16 22:05:46 +00:00 | 13 commits to main since this release

    Mission O1 — the fixture-id hardening

    The suite's fixture builders derived message/part row ids from (sid-slice, timestamp) pairs; sibling inserts sharing a millisecond AND a sid suffix collided on the id primary key (UNIQUE constraint failed: message.id — 5 errors on vanilla main in the fast CI container, NNNN's forge run). The cure is shared: tests/__init__.py::fixture_row_id appends a per-process monotonic segment, derivation shape preserved verbatim; nine builders routed through it. Four regression legs in tests/test_round_oooo.py reproduce the exact CI shape deterministically (same-ms triples + a frozen-clock burst) — tripwire-proven against the vanilla derivation. Flake dead: 15/15 tight loops, full suite stable at 1692 legs (+4).

    Mission O2 — the timeline fold (NNNN-V1, MINOR)

    The README round timeline promised every shipped round but ended at v1.96.0/#131. Added the v1.97.0/#132 (NNNN) and v1.98.0/#133 rows; completed the three empty PR cells found by the sweep (#61 → v1.34.0, #73 → v1.44.0, #94 → v1.64.0 — the merge-wedge read-false row whose squash is on main).

    Downloads