AnimaLogica / kathikon / 47db483ad899d23c55b4c54e00c5fb8620d06db0-PR-60

97%

main: 97%

Pull #60

github

The Elixir 1.19 job dies because the scheduler promoter calls `table_exists?/1` while Mnesia is stopped. `:mnesia.system_info(:tables)` then exits with `{:aborted, {:no_exists, {:schema, :tables}}}` or `{:aborted, {:node_not_running, ...}}`, the promoter crashes on every 50 ms tick, and the supervisor takes `Kathikon.Queue` and `Kathikon.QueueControl` down with it. Later tests fail because those processes are gone.

`table_exists?/1` now returns `false` unless Mnesia is running and the schema can be read. `promote_scheduled/1` already treats a missing jobs table as "promoted 0", so a tick during `:mnesia.stop()` no longer kills the process.

```1232:1240:lib/kathikon/storage/mnesia.ex
  defp table_exists?(table) do
    try do
      :mnesia.system_info(:is_running) == :yes and table in :mnesia.system_info(:tables)
    catch
      :exit, _ -> false
    end
  end
```

`mix test test/kathikon/storage/mnesia_test.exs test/kathikon/backend/storage_mnesia_integration_test.exs` — 42 passed, including a new case that stops Mnesia and expects `promote_scheduled/1` to return 0. The promoter no longer terminates in that run.

193 of 194 new or added lines in 3 files covered. (99.48%)

1450 of 1492 relevant lines covered (97.18%)

95.03 hits per line