OperationalError: unable to open database file

[gisce-ti-sentry]

Sentry Issue: GAB-4

OperationalError: unable to open database file
  File "threading.py", line 1024, in run
    self._target(*self._args, **self._kwargs)
  File "github_agent_bridge/executor.py", line 181, in _loop
    did = self.work_one(worker_id)
  File "github_agent_bridge/executor.py", line 58, in work_one
    job = self.queue.claim_next(worker_id)
  File "github_agent_bridge/queue.py", line 148, in claim_next
    con.execute("BEGIN IMMEDIATE")

This issue was automatically created by Sentry via Github Issue

[giscebot]

This is covered by the existing fix in #7.

I rechecked the Sentry stack for this issue against the PR: it is the same SQLite queue path failure class as #6, happening during claim_next() after opening the queue DB. PR #7 expands literal ~ DB paths and recreates the SQLite parent directory before every connection, covering long-running executor workers whose state directory is missing at claim time.

I also updated the PR body to include Fixes #8, so merging #7 should close this issue too.

Validation on the PR worktree: .venv/bin/pytest -q tests/test_queue.py -> 16 passed.

[giscebot]

Revalidated this issue against #7 from the dedicated github-agent-bridge-issue-6 worktree.

Verdict: no separate patch is needed for #8. The Sentry stack reaches JobQueue.claim_next(), and PR #7 now recreates the SQLite DB parent directory inside connect() before that transaction starts, plus expands literal ~ DB paths at queue construction.

Validation just run:

• .venv/bin/pytest -q tests/test_queue.py -> 16 passed
• .venv/bin/pytest -q -> 157 passed, 1 existing Starlette/httpx deprecation warning

PR #7 already includes Fixes #8, so merge of that PR should close this issue.