fix(rust): log datastore write errors instead of discarding them

[ThisIsLeoS]

Problem

heartbeat() reports a failed write by returning an {"error": …} object; it never throws. RustInterface.heartbeatHelper() discarded that response (its Log.w was commented out), and insertEvent() logged it only at debug level, mixed in with every successful response. A watcher whose writes fail therefore loses every event with nothing in logcat to show it.

One way to get there: deleting the media or web bucket from the web UI while the app runs. Those watchers create their bucket only when their service starts, so every later heartbeat fails with NoSuchBucket until the service is recreated. A datastore worker that has died fails all writes the same way.

Change

• A new WriteFailureLog logs a warning for the first failed write of a run and, once a write to the same bucket succeeds again, how many failed in between. A line per failure would flood logcat, since watchers write every few seconds.
• Runs are counted per bucket, in one log for the whole process: the batch watchers build a new RustInterface on every read, and a run of failures has to outlive them.
• heartbeatHelper() and insertEvent() pass every response to it. insertEvent() no longer logs each response at debug level.
• heartbeat() is declared String?: the native side returns null when it cannot build even its error object, and that counts as a failed write.
• Corrects the comment in MediaWatcher.onCreate: its init calls never go through the HTTP server, so whether the server has started doesn't matter.

#359 reads the same {"error": …} object in getBucketsJSON(). Once one of the two PRs is merged, the other can reuse its check instead of adding a second copy.

Testing

• New WriteFailureLogTest: no log on success, one line per run of failures, the recovery count, buckets counted separately, a missing response, and parsing of the error object.
• ./gradlew :mobile:testStandardDebugUnitTest passes.

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable issues were found.

Summary

Adds warning logs for failed datastore writes without logging every response.

• Datastore write failures now show up in logcat by bucket.
• The media watcher comment describes its local startup work without calling it HTTP.

No actionable issues were found. Tests were inspected but not run.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[heartbeatHelper or insertEvent] --> B[Native heartbeat response]
    B --> C[Shared WriteFailureLog]
    C --> D{Error or missing response?}
    D -->|Yes| E[Increase this bucket's failure count]
    E --> F[Warn only on first failure]
    D -->|No| G{Previous failures for this bucket?}
    G -->|Yes| H[Log recovery count and clear it]
    G -->|No| I[No log]

Loading

_{Reviews (1) · Last reviewed commit: "docs(media-watcher): drop the wrong HTTP..." · Reviewed by Greptile}