Files
Michał Pierzchała c1c3d30022 fix(replay): surface warnings on failed runs and keep them honest
A failing `replay` still printed no trace of the run's accumulated warnings:
the human error renderer read message, hint, and divergence but never the
error-level warnings channel, so the pass/fail asymmetry survived on that
surface. It now prints each warning after the error.

Round-two corrections: warnings on the divergence response are scrubbed
against replay vars like message and hint already are; a failed attempt
copies warnings only when non-empty, matching the passing branch; the
optional-skip hint append is removed because no producer of a maestro test
failure emits a hint — capture-error lineage stays with #2450 — along with
the CHANGELOG claim it rested on; the failed-result copy gains its own
module test, and JUnit covers a failed test's warning lines.
2026-09-14 13:17:48 +02:00
..
2026-02-05 15:25:52 +01:00
2026-02-05 15:25:52 +01:00