umputun / tg-spam / 24966212061

83%

master: 83%

Build Type

Pull #294

github

Committed by web-flow

Commit Message

fix: preserve InstanceID across loadConfigFromDB swap; rename collided subtests (#397)

* fix: preserve InstanceID across loadConfigFromDB swap; rename collided subtests

`loadConfigFromDB` does `*settings = *dbSettings` which clobbers the
CLI/env-supplied `InstanceID` with whatever the persisted blob carries.
External orchestrators that write per-instance config blobs without
embedding `instance_id` (e.g. tg-spam-manager) leave it empty, so after
the swap `settings.InstanceID == ""`. Initial load works because the
short-lived store inside loadConfigFromDB was already keyed by the
still-CLI value, but every subsequent `makeDB` call in `activateServer`
opens with `gid=""` — including the runtime SettingsStore.

Symptoms: `POST /config/reload` returns
`500 "no settings found in database: sql: no rows in result set"` from a
clean state. The first UI Save with `saveToDb=true` then writes a fresh
row under `gid=""`, so subsequent reloads succeed against a
manager-orphaned row, leaving manager-side updates and bot-side updates
on different rows.

Fix: snapshot `settings.InstanceID` before the swap and restore it when
the loaded blob's value is empty. A blob that does carry its own
non-empty `InstanceID` (saved by tg-spam itself) is still trusted, so
existing single-binary deployments are unaffected.

Same patch also gives unique names to the three subtests in
`TestDetector_CheckOpenAI` that collided post-master-merge — Go was
running them under `#01` suffixes which made `-run` filtering ambiguous.
Two are byte-identical to their earlier siblings; the third differs
slightly (uses `spam` instead of `viagra` and asserts the LoadStopWords
result), and the new name reflects that.

* add WARN on InstanceID divergence + regression test for empty-blob branch

Address review:

- log a [WARN] when the persisted blob carries a non-empty InstanceID
  that differs from the CLI/env value. Behaviour is unchanged (blob
  still wins) but the divergence shifts the runtime gi... (continued)

Pull Request Pull Request #294: Implement database configuration support

Coverage Stats

1290 of 1591 new or added lines in 7 files covered. (81.08%)

10 existing lines in 2 files now uncovered.

8174 of 9839 relevant lines covered (83.08%)

232.02 hits per line