browserstack.yml

# browserstack-node-sdk config for the BS Automate variant of the e2e
# suite. Wired by `npm run test:bs` (=
# `browserstack-node-sdk playwright test --config tests/e2e/playwright.config.js`).
# The SDK monkey-patches Playwright at runtime to accept the real-device
# platforms below, then appends one project per platform to the config
# (see playwright.config.js header for the version pin caveat).
#
# Conventions adopted from feature/form_v2's browserstack.yml — they
# survived a round of debugging on real devices we don't want to repeat:
#   - `browserName: safari` (NOT `playwright-webkit`). The SDK accepts
#     'safari' and translates it; plain Playwright rejects with
#     "expected one of (chromium|firefox|webkit)".
#   - Empty `userName`/`accessKey` literals — that signals the SDK to
#     read BROWSERSTACK_USERNAME / BROWSERSTACK_ACCESS_KEY from env.
#     Inline secrets here would commit them to the repo. The
#     `npm run test:bs` script sources ../.env.dev before exec.
#   - `testReporting: false`. The SDK's testHub observability layer
#     mis-serializes Playwright errors on long-running iOS Safari tests
#     ("Serialized error must have either an error or a value"); turning
#     it off still leaves Automate dashboard sessions + videos.

userName: ''
accessKey: ''

projectName: formr.org PWA
buildName: pwa-reload-fix
buildIdentifier: '#${DATE_TIME}'

# Single iOS target — the platform where the bug manifested. Adding more
# platforms multiplies BS minute usage; expand only when there's a
# regression worth catching on a second device.
platforms:
  - browserName: safari
    osVersion: 17
    deviceName: iPhone 15 Pro Max

parallelsPerPlatform: 1

# study.researchmixtape.com is publicly reachable so we don't need a
# BrowserStack Local tunnel.
browserstackLocal: false

debug: false
networkLogs: false
consoleLogs: errors

# See above — testHub serialization bug mangles iOS Safari errors.
testReporting: false