fix(store): heal a database that version 1 left empty - #11
Conversation
The launch probe could create 'ftw' at version 1 with no object stores, and an app opening the same version got no upgrade callback and created none either — every read and write broken for the life of the install. The probe no longer does that, but prevention cannot reach a phone already in that state. Only a version it has not seen can. Version 2, with an upgrade that creates each store only if it is missing. A broken database gets its schema; a healthy one is untouched, rows and all. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Deploying with
|
| Status | Name | Latest Commit | Updated (UTC) |
|---|---|---|---|
| ✅ Deployment successful! View logs |
ftw-webapp | 2a7bbf6 | Aug 05 2026, 05:43 PM |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2a7bbf6f8c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| // either, so every read and write failed for the life of the install. The | ||
| // probe no longer does that — but installs that already suffered it cannot be | ||
| // healed by prevention, only by a version they have not seen. | ||
| const DB_VERSION = 2 |
There was a problem hiding this comment.
When an existing v1 install launches with ftw.site set, the inline boot probe still opens indexedDB.open('ftw', 1) and, on success, never closes req.result after checking/reading snapshot (index.html:74 and index.html:93-99). Bumping the app open to version 2 makes that still-open v1 connection block the upgrade that is supposed to repair the database, so the exact broken v1 installs this targets can leave db() pending for the session; after a successful v2 upgrade, the same v1 probe will also just VersionError and skip the first-frame cached snapshot path. Please update the probe to the new version and close or onversionchange-close its connection as part of this bump.
Useful? React with 👍 / 👎.
1 participant
Follow-up to #9, which stopped the launch probe from creating an empty database — but could not help an install that already had one. Version 2 with an idempotent upgrade repairs the schema in place; a healthy database is untouched, rows included. Tests build the broken state deliberately and open it the way the app does.