test(updater-probe): add two probe casks to drive the per-cask updater

[edbfi]

What changed

Two temporary casks, updater-probe and updater-probe-b, for the production test of the per-cask updater (#49, #50). Owner decision, 2026-10-12. A follow-up PR removes all of it.

• Casks/test/updater-probe{,-b}.rb, pipelines/updater-probe{,-b}/: minimal universal (arm64 + x86_64) apps, Updater Probe.app and Updater Probe B.app. Their upstreams are updater-probe-vX.Y.Z and updater-probe-b-vX.Y.Z releases in edbfi/edbfi-ci-playground.
• Resolver: the newest published <token>-vX.Y.Z release, not /releases/latest. test(updater-probe): add a temporary cask to drive one updater push #47's resolver used /releases/latest, but with two probes in one repo the latest release can belong to the other one. The exact asset and URL checks are unchanged.
• inspect-macos.py gets both APPS entries, and test_discovery_separates_kinds lists them.
• update-casks.yml is untouched. This tap now has eight casks. test(updater-probe): add a temporary cask to drive one updater push #47 had to raise the timeouts for a seventh.

Evidence

• inspect-macos.py passes on all four DMGs (0.0.2 and 0.0.3 for each probe): the right bundle IDs, arm64 + x86_64, minimum macOS 12.0.
• discover.sh lists eight casks, each at 20 minutes. Without upstream releases, resolve.sh updater-probe-b reports skip=true.
• prek run --all-files --hook-stage manual: every hook passes.

Plan after merge

1. Run A: publish updater-probe-v0.0.2 and updater-probe-b-v0.0.2 in the playground, then dispatch. Expected: publish (updater-probe) and publish (updater-probe-b) run in parallel; push (…) runs one cask after the other; two chore(…): update to 0.0.2 commits from the deploy key, each logged as bypass and each starting push CI.
2. Run B: dispatch again. Expected: every check says no update and nothing after plan runs.
3. Run C (failure isolation): publish 0.0.3 of both, and first put a different UpdaterProbeB-0.0.3.dmg on updater-probe-b-latest. Expected: publish (updater-probe-b) fails with "differs from upstream" while updater-probe goes through to its push.
4. Teardown PR: after it merges, delete both -latest releases and the playground's probe releases and tags.

Owner actions

None.

[pullfrog]

✅ No new issues found.

Reviewed changes: the full PR diff (one commit) against main, cross-checked against scripts/lib/common.sh, scripts/update.sh, the qview resolver and update-casks.yml.

• Probe casks — Casks/test/updater-probe{,-b}.rb follow the standard pattern (<token>-latest release, <ASSET_PREFIX>-#{version}.dmg, placeholder version/sha256 that the updater rewrites).
• Pipelines — config.env and resolve.sh for each probe. The resolver picks the newest published <token>-vX.Y.Z tag. The anchored regex keeps updater-probe from matching updater-probe-b-v… tags, and sort -V ordering, the skip=true path and the exact asset and URL checks all behave as described.
• Registration — inspect-macos.py APPS and test_discovery_separates_kinds list both tokens, so the APPS == discovery assertion holds.

  ^{｜ Fix it ➔ ｜ View workflow run ｜ Using claude-sonnet-5-5 ｜ 𝕏}