chore(submodule): bump aw-server-rust to master (c535d77)

[TimeToBuildBob]

Refresh (2026-10-10)

Re-pointed to aw-server-rust master c535d77 and merged current aw-android master. Since the previous pin 9008520, that adds 14 merged server commits, including:

• fix(datastore): make v3→v4 key_value migration idempotent aw-server-rust#792: v3→v4 key_value migration is idempotent
• fix(datastore): commit stopwatch changes before acknowledging aw-server-rust#743: stopwatch changes are committed before acknowledging
• fix(aw-server): re-importing an updated export replaces events in place aw-server-rust#766: re-importing an updated export replaces events in place
• fix(aw-server): return JSON errors with a reason for 400/404/422 under /api aw-server-rust#772: JSON errors with a reason under /api
• aw-webui bump f4b4a1a6 (2026-10-06)

Android master still pins 70ba50d, so none of these reach the app until this merges. Not included yet: ActivityWatch/aw-webui#1061 (merged 2026-10-06 21:05, after the server's last aw-webui bump), which aw-android#321 items 3/7/8 need. That requires an aw-webui bump in aw-server-rust first, then a re-bump here.

---

Summary

Bumps aw-server-rust submodule to 9008520 (master as of 2026-10-02), which includes:

• #721 / #722: stream headers before body on JSON and CSV export endpoints (fixes large export hangs)
• #770: bump bundled aw-webui to include #997 (server-side CSV generation replacing in-browser materialization)
• Various aw-client-rust, aw-sync, and transform fixes

Together these fix #228: Export button silently fails for large datasets.

Testing

• Large single-bucket CSV export should now stream instead of blocking until full response body is ready
• JSON export path already fixed by #304

Fixes #228

[TimeToBuildBob]

Review guide (auto-posted)

Structured warm-up for reviewers — what changed, what to run, where to look.

Key files

• aw-server-rust (+1/-1, Δ2)

Suggested verification

• Run the repo's usual CI-equivalent checks locally

Known risks / watch points

• No automatic high-risk tags; use file list + diff for judgment.

Suggested review focus

• Confirm behavior matches the PR description acceptance criteria.
• Skim the largest diffs first (listed above).
• If CI is green, spot-check the highest-risk paths called out here.

Generated by scripts/github/pr-warmup-review-guide.py for #305.

[greptile-apps]

Confidence Score: 5/5

[High risk] Bumps the Rust backend submodule with test changes.

No actionable new issue or outstanding previous finding is established by the available review evidence.

Summary

The PR advances the bundled Rust server pointer to 9008520 and adjusts the large-history Android upgrade test to avoid competing database access during startup.

• The test marks the already-sanitized fixture hostname as migrated.
• It closes the prepared statement and database before removing SQLite sidecar files.

_{Reviews (2) · Last reviewed commit: "chore(submodule): bump aw-server-rust to..."}

[TimeToBuildBob]

E2E Test E2E (29) failed on UpgradeWithHistoryTest.serverAnswersWhileLegacyHistoryMigrates (legacy events were not merged within 60000 ms). Not the export-header bump — 70ba50d..05c7b38 does not touch aw-datastore.

Logcat from run 35826089151: rust worker panics at journal_mode=WAL with database is locked, then events/count returns 500 (datastore worker is gone) for the rest of the wait. /api/0/info still answers, so the 20s ready wait passed.

Cause: the fixture set hasMigratedHostname but startup uses the distinct sanitizedHostnameMigratedTo flag. Java opened the 17MB seeded sqlite.db to rewrite hostnames while the worker switched WAL.

Pushed b908cfd so the fixture also sets sanitizedHostnameMigratedTo. CI queued on the new head.

[TimeToBuildBob]

CI-green and mergeable (Greptile 5/5) — waiting only on a maintainer click.

This PR is ready to merge, but the bot has pull-only access to this repo and can't self-merge — surfacing it here so it isn't lost. The monitoring loop will stop re-flagging it now that this note is posted.

[TimeToBuildBob]

E2E Test E2E (29) on run 35828479752 failed on NativeWindowInsetsTest.rotatingMainActivityKeepsTheSameWebView — "Instrumentation run failed due to Process crashed" — a different test from the previous run's failure. Two runs, two distinct E2E tests, both crashing on the software-rendered emulator (no KVM): classic environment flakiness.

This PR only bumps aw-server-rust (Rust server code). Neither rotation behaviour nor WebView identity is touched. Pushed an empty commit to re-trigger CI.

[TimeToBuildBob]

E2E Test E2E (29) on run 35831060120 failed on UpgradeWithHistoryTest again (legacy events were not merged within 60000 ms). The hostname-skip in b908cfd did stop the WAL journal_mode panic; the worker then died later on the v6 exclusive index migration:

Failed to run v6 migration transaction: database is locked (aw-datastore/src/datastore.rs:223).

Cause: the 17MB seed left its compiled INSERT SQLiteStatement open. That statement holds a SQLiteClosable ref, so db.close() did not drop the last reference and the Java connection stayed alive across process startup. rust's BEGIN EXCLUSIVE then hit SQLITE_BUSY (5s busy_timeout), panicked, and events/count stayed 500.

Pushed f4cce4c — close the statement before db.close(), and delete leftover -wal/-shm. CI queued on the new head. Not merge-ready until this E2E run is green.

[TimeToBuildBob]

E2E Test E2E (29) on run 35833630532 failed on NativeWindowInsetsTest.syncToggleReceivesRealTap — "A screen tap must change the persisted setting expected: but was:". This is an emulator UI-tap flake: the test already has retry logic with readiness polling (tapUntilPrefChanges, up to 3 attempts, bounds-stable wait), added specifically because the software-rendered CI emulator (no KVM) is slow to settle between activities. The failure is unrelated to the aw-server-rust bump. Pushed 27ad04b to re-trigger CI.

[TimeToBuildBob]

E2E Test E2E (29) is green on 27ad04b (run 35836334678). Unit Test also green. MERGEABLE.

The previous failure on f4cce4c (run 35833630532) was NativeWindowInsetsTest.syncToggleReceivesRealTap — emulator tap flake, not the rust bump. Empty retrigger was the measurement; it passed.

Waiting on a maintainer merge (pull-only on this repo). Does not close #228 until a device confirms the large JSON export.

[TimeToBuildBob]

🤖 AI code review

Safe to merge — no P0/P1 findings on latest review

_{Updated after inline dispositions on finding threads — this is the current state; the verdict below is frozen at review time and is kept as the historical record of that pass.}

Finding disposition

Finding	Severity	State
mobile/src/androidTest/java/net/activitywatch/android/UpgradeWithHistoryTest.kt:220	P2	superseded by latest review (not reproduced)
mobile/src/androidTest/java/net/activitywatch/android/UpgradeWithHistoryTest.kt:95	P2	rejected

Bumps the aw-server-rust submodule from 70ba50dd to c535d77a and adjusts the UpgradeWithHistoryTest instrumentation: sets the sanitizedHostnameMigratedTo preference, wraps the seed inserts in a try/finally that closes the compiled SQLiteStatement, and deletes the -wal/-shm files after closing the seeded database.

Needs a look — P2 only

Confidence 4/5

1 finding · ⚠️ 1 P2

⚠️ P2 medium — mobile/src/androidTest/java/net/activitywatch/android/UpgradeWithHistoryTest.kt:95

The test is supposed to simulate an upgrading install with both buckets present and six-figure history, but this line pre-seeds sanitizedHostnameMigratedTo to the current hostname. The comment on lines 91-94 explains that without that preference, startup opens sqlite.db from Java to rewrite bucket hostnames while the rust worker is switching journal_mode to WAL, racing SQLITE_BUSY, panicking the worker, and leaving events/count at 500. A real upgrade from a release before this preference existed would not have it set, so it would still hit that race. The test therefore skips exactly the startup path that the regression was about and passes while the production failure remains for users upgrading large histories; CI no longer detects the worker panic.

If the production code sanitizes hostnames, seed the sanitized value, e.g. .putString("sanitizedHostnameMigratedTo", hostname.lowercase()) or whatever the production sanitization does.

_{How this was verified: Checked the test file: hostname comes from deviceHostname(context) at line 83, and the preference is set to that raw value. The production code that reads sanitizedHostnameMigratedTo is in the Rust server or Java migration code, not in this diff, so I cannot confirm the exact sanitization, but the key name implies sanitization is applied. If the production code sanitizes before comparing, the raw hostname will not match.}

_{Consensus: 2/3 passes agreed — pass 3 looked and disagreed}
Distinct keys: 1 (general)

Files changed (2) — the diff as I read it

• aw-server-rust — Updates the submodule pointer from 70ba50ddc3852acadfb5e90ab20d54e04d8a60b2 to c535d77af495681e4b3960c4e4f5991e1c04a3c7.
• mobile/src/androidTest/java/net/activitywatch/android/UpgradeWithHistoryTest.kt — Sets sanitizedHostnameMigratedTo, closes the event insert statement in a finally block, and removes -wal/-shm files after seeding.

Previous review passes

commit	score	findings	engine	when
27ad04b43e67	5/5	0	llm	2026-09-25 10:27 UTC
173e6335edec	5/5	0	llm	2026-10-02 15:03 UTC
173e6335edec	5/5	0	llm	2026-10-09 03:05 UTC
5cea4169cd2f	4/5	2	llm	2026-10-10 03:22 UTC

_{Reviewed 5cea4169cd2f · openrouter/deepseek/deepseek-v4-flash-0731 · llm engine · 1099s · about this reviewer}

_{Maintainer commands}

_{@TimeToBuildBob review (own line) — fresh review · @TimeToBuildBob fix — a worker acts on the findings. Once per comment; 👀 = received.}