Some checks failed
CI / Unit & Component Tests (pull_request) Failing after 3m56s
CI / OCR Service Tests (pull_request) Successful in 48s
CI / Backend Unit Tests (pull_request) Failing after 3m13s
CI / Unit & Component Tests (push) Failing after 4m40s
CI / OCR Service Tests (push) Successful in 56s
CI / Backend Unit Tests (push) Failing after 3m20s
The multi-person filter e2e previously typed 'a' then 'b' into the
typeahead and trusted the dev seed to contain matching names.
If the seed ever changes, the test would silently degrade — both
calls might resolve to the same row, or the listbox might never
populate.
Refactor to use a single broadly-occurring probe vowel ('e') and
extract person ids straight from the listbox option DOM (the option
id encodes the person id as `${listboxId}-option-${personId}`).
For the second pick, iterate options and select the first whose
id differs from the first selection. The test now only depends on
the seed having ≥2 distinct persons whose name contains 'e' — a
much weaker, more durable assumption — and asserts on the URL
params with full equality instead of toHaveLength + first-element
spot checks.
Addresses Sara's iteration-3 concern #4 on PR #382.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>