[pasrom]

What changed and why

Follow-up to #960 (recording retention, issue #723). Unit tests pin the default (off) and the policy, but nothing checked the production chain: the retention pass that runs when a job reaches done, against the app's real settings and output folder.

• Every lane that goes through run_one_meeting (default, --two-meetings, --title-source, --mic-device-change) now checks that the job's relocated recordings/<stem>_mix.wav is there right after done and still there 5 s later. The second look is needed because the pass runs on its own just after the state change, and with the switch off it logs nothing the lane could wait for.
• Under the default period of 30 days a recording minutes old is never due, so that check alone would stay green with the switch ignored. In CI the app is therefore launched with -recordingRetentionDays <integer>0</integer> as a launch argument, so nothing is stored or has to be restored. The switch is then all that keeps the recording.
• The 0-day period is applied only in CI ($GITHUB_ACTIONS) and only when the host has not switched retention on. With 0 days a broken switch would delete every listed recording in the host's output folder, and on a developer's Mac those are their own recordings.
• A host with retention switched on skips the check and logs that it did. "On" follows UserDefaults.flag: yes or true in any case, or any integer but 0.
• The verdict is retention_default_verdict in scripts/lib/e2e-helpers.sh, pinned by scripts/tests/test_retention_default_verdict.sh (runs in about a second in the existing script-test step).

Tests

• Script test, 18 checks, run locally and in the existing CI step. Removing the second look (always reporting kept) makes deleted_after_done_is_caught fail.
• bash -n and shellcheck -S warning are clean on the changed files, and every scripts/tests/test_*.sh passes locally.
• The lane itself runs on this PR through the run-e2e label. What to look for in the e2e-app log: CI: recording retention period 0 days for this run and recording still in ... 5s after done.
• Not checked: that the lane goes red with the switch broken. That would need a deliberately broken build on the runner. The unit test for the switch covers that direction.

Known limits, left as they are:

• A pass that starts more than 5 s after done goes unseen.
• The mix path comes from the transcript's stem. The lanes keep the default file names, so this holds for them.
• The --mic-stall lane does not go through run_one_meeting and is not covered.

AI assistance

• Written with: Claude (Opus) in Claude Code.
• Reviewed with: Claude (Opus), one cleanup pass and one correctness pass. The correctness pass found that the first version could not fail under the 30-day default, which led to the 0-day period in CI.

Checklist

• Rebased on main, no merge commits
• Script tests pass, shellcheck clean on the changed files
• Conventional Commits, one logical change per commit

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.1%. Comparing base (f5ba602) to head (ed934d2).

Additional details and impacted files

@@          Coverage Diff           @@
##            main    #977    +/-   ##
======================================
  Coverage   92.1%   92.1%            
======================================
  Files        322     322            
  Lines      24795   25374   +579     
  Branches       6       6            
======================================
+ Hits       22838   23385   +547     
- Misses      1953    1985    +32     
  Partials       4       4            

Flag	Coverage Δ
appstore	92.3% <ø> (-0.2%)	⬇️
audiotap	92.5% <ø> (+0.4%)	⬆️
homebrew	92.1% <ø> (+<0.1%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 56 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.