Distillation

Date: Sep 10, 2026

• 🔴 (04:40) User specified upgrade checks “never uses the cached target after %s”; after the relevant responses, the upgrade flow must report errors rather than reuse a cached target.
• 🔴 (04:40) User stated: “Switch to nightly and use npm method → triggers migration.”
• 🟡 (04:40) Upgrade tests added resolves a pinned check target from its exact source, ensuring a pinned version is resolved against that version’s source rather than the generic latest release.
• 🟡 (04:40) For sentry cli upgrade --check --method curl 2.0.0, the fetch mock returns { tag_name: "cli@2.0.0" } for /releases/tags/cli%402.0.0 and [{ tag_name: "cli@99.0.0" }] otherwise; expected warning remains Run 'sentry cli upgrade 2.0.0' to update.
• 🟡 (04:40) For unpinned curl check mode, the mocked latest-release response is [{ tag_name: "cli@99.0.0" }]; expected result has action="checked", targetVersion="99.0.0", and warning Run 'sentry cli upgrade' to update.
• 🟡 (04:40) Upgrade-test fetch mocking includes an npm-registry fallback: requests whose hostname is registry.npmjs.org receive { version } with HTTP 200 and content-type: application/json.
• 🟡 (04:40) Nightly check output is expected to include Channel: nightly, version 0.99.0-dev.9999999999, and Run 'sentry cli upgrade' to update.
• 🟡 (04:40) Forced-upgrade test expectations were made deterministic: output must not contain Already up to date, must contain Downloading 1.0.0 and Upgraded to, and must contain 1.0.0; previous assertions using CLI_VERSION were replaced.
• 🟡 (04:40) Nightly download tests use a 307 redirect to https://blob.example.com/nightly.gz.
• 🟡 (04:40) Upgrade tests added request tracking via const requests: string[] = []; and a mock branch for the exact request https://api.github.com/repos/getsentry/toolkit.