https://sonarly.com/issue/38314?type=bug
Fresh self-host installs can enter a restart loop after creating the first workspace because startup migrations fail a workspace-version safety check that expects a row that is never written for new workspaces.
Fix: I updated the workspace safety check in `run-instance-commands` to align with cursor semantics instead of exact row existence:
1. Replaced `areAllWorkspacesAtCommand({ commandName })` usage with per-workspace cursor validation:
- Fetch each active/suspended workspace cursor via `getWorkspaceLastAttemptedCommandNameOrThrow`.
- Validate each cursor has reached or passed the required previous-version tail command using sequence order.
2. Added `UpgradeSequenceReaderService.isStepCompletedOrPassed(...)`:
- Returns true if cursor is after required step.
- Returns true on same step only when status is `completed`.
- Returns false when cursor is before required step.
3. Updated safety-check error wording from “not completed” to “not reached” to match cursor/order-based semantics.
4. Added unit tests in `upgrade-sequence-reader.service.spec.ts` covering:
- same-step completed (true),
- same-step failed (false),
- cursor after step (true),
- cursor before step (false).
This preserves the intent of the safety gate (prevent unsafe upgrades) while removing false failures for fresh workspaces initialized with a valid later cursor.
Authored by Sonarly by autonomous analysis (run 43722).
sonarly-bot
bff10a04a1