Krzysztof kuhy Rudnicki
6258b35f88
Some checks are pending
file-length / file length <= 250 lines (push) Waiting to run
pre-commit / pre-commit (push) Waiting to run
Tests / test (3.10) (push) Waiting to run
Tests / test (3.11) (push) Waiting to run
Tests / test (3.12) (push) Waiting to run
release-apk / release (APK) (push) Waiting to run
Two independent costs in the background tick, both pure waste.
`list_directory` was called four times per tick -- once by the log pull and
once by each of the three bank syncs. It is the single most expensive call
in a tick at ~445ms mirrored (the GitHub half alone is ~360ms), so that was
~1.3s spent asking the same question four times. It is now listed once in
`run_sync` and passed down.
The curated-bank loop was worse: 17s and 62 requests to fetch *zero* useful
bytes. Firebase returns None for every peer's `food_bank_manual.json`, so
`MirrorSyncClient.get_file_text` treated each miss as "this device has not
migrated yet" and fell through to GitHub, which answers `{}` after ~320ms.
`_primary_only_text` skips that per-peer fallthrough for this one artifact.
Checked before narrowing rather than assumed, because CLAUDE.md requires
both halves of the food bank to sync: reading every peer's curated bank
straight off the mirror found 23 absent, 2 empty, and zero non-empty. No
curated food exists on the mirror alone. The probe is written into the
docstring so it can be re-run before anyone widens this again.
Measured end to end: manual-bank loop 17.0s -> 3.4s, full tick ~27s -> 15.2s,
log/bank/peer state unchanged (441 entries, 117 foods, 22 peers).
`_pull_remote_logs` now returns the revisions it saw instead of mutating a
caller-supplied dict, which keeps it under the argument limit and makes the
data flow visible at the call site.
Deliberately NOT done: revision-skipping the derived food bank. `run_sync`
rebuilds that bank from the merged log immediately before syncing it, so
unlike the log the peer records are not durably held locally -- skipping a
peer would drop its counts from the pushed bank and oscillate under
max-count-wins. Verified empirically: the on-disk bank and a pure log-derived
rebuild are identical (117 foods, 0 differing counts), and the 2 foods peers
know that the local bank lacks are discarded by the existing `if name in
local` filter anyway. The loop is off every interactive path now, so the
remaining cost is background-only.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LJ5YnA3CERitQ57HjMu2Qz
|
||
|---|---|---|
| .. | ||
| .agents/skills | ||
| android | ||
| assets/icon | ||
| bin | ||
| desktop | ||
| lib | ||
| test | ||
| web | ||
| .gitignore | ||
| .metadata | ||
| analysis_options.yaml | ||
| install_arch.sh | ||
| pubspec.lock | ||
| pubspec.yaml | ||
| README.md | ||
| run.sh | ||
| skills-lock.json | ||
diet_guard_app
Companion app for diet_guard: log meals on the phone, and on the Linux
desktop, sharing one food log through GitHub-backed sync.
Two targets, and only two
| Target | What it is |
|---|---|
| Android | a normal Flutter app (flutter build apk) |
| Linux desktop | the Flutter web build, served by a local wrapper and shown in a Chrome --app window |
There is no linux/ directory. Flutter's GTK embedder manages ~20fps at
3840x2160 -- measured on a stock flutter create counter app, so it is the
toolkit rather than this code -- where the same Dart rendered by Chrome
sustains ~144fps (~/todo/docs/desktop-performance-findings.md). Running
flutter create --platforms linux here would silently restore the slow path.
Commands
flutter test # the app's suite
flutter analyze
bash run.sh # build the web bundle and launch the desktop app
bash install_arch.sh # build + install the pacman package and .desktop entry
flutter build apk --release # the phone build
adb install -r build/app/outputs/flutter-apk/app-release.apk
Never adb uninstall or pm clear this package: both wipe the on-device log.
The desktop wrapper
bin/diet_guard_desktop.dart is the other half of the desktop app, because a
browser cannot touch the filesystem. It:
- serves the built web assets from
build/web(or/opt/diet-guard-app/webonce packaged); - keeps an on-disk copy of every document and photo under
~/.local/share/diet-guard-desktop/, which is what makes a wiped Chrome profile recoverable -- for photos it is the only copy, since they never sync; - holds the GitHub token (
~/.config/diet_guard_app/sync_token, falling back to the PC gate's~/.config/diet_guard/sync_token) and proxiesapi.github.complus the device flow, whose endpoints send no CORS headers and so cannot be called from a page at all.
The port (8732) and the Chrome --user-data-dir must stay fixed.
IndexedDB is keyed by origin and lives inside that profile, so changing either
looks to the app like a device with no history.
Platform seams
dart:io does not fail a web compile; it becomes a stub that throws at
runtime, so the symptom of a missed seam is a blank white window rather than a
build error. Every platform edge is a conditional export -- document_store*,
blob_store*, background_tasks*, notification_backend*, token_vault*,
github_client_factory*, attached_image*. Branch on kIsWeb before any
Platform.is…, which itself throws on web.
What the desktop genuinely loses, by design rather than omission:
- Background reminders. A browser runs nothing once its window is closed,
so the due-slot check runs in-page every 5 minutes while the window is open.
On the PC the real backstop is
diet-guard-gate.timer, which locks the screen. - Camera capture. The browser picker is a file input.
- Photo sharing between devices -- unchanged:
imagePathwas always stripped before push.
Sync identity
Three devices push under diet-guard-sync/devices/: pc (the Python gate),
phone, and desktop. Each must keep its own id -- two devices sharing one
would overwrite each other's pushed log on every tick.