meshcore-client

Commit Graph

Author	SHA1	Message	Date
diffwireauto	762b60dc96	fix(#710 ): init flserial mutexes before the reader thread can use them fl_open started SerialThread and only then called mtx_init. Upstream's thread locked a mutex only after its first read returned data, so the gap rarely mattered; #710's room check locks inFifo_mutex on the thread's first pass, which made the race reachable on every connect. The mutexes are now initialised right after the FlSerial is allocated, before open() can throw, so _fl_close_slot also never destroys an uninitialised mutex after a failed open. Both copies stay identical. Found by the pre-PR Gemini review. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	1 month ago
Strycher	03266e709e	fix(#710 ): flserial never overwrites incoming serial bytes flserial's reader thread copied every byte into an 8 KB ring that overwrote the oldest bytes when full, and told Dart about data only once the port went idle. The RK4 sends its 350-contact list (about 52 KB) in about 0.17 s, so most of it was overwritten before the app read it: the app saw 106-154 contacts and #660's contacts were lost. The thread now reads only as much as the ring has room for. When the ring is full it leaves the bytes in the OS driver, and USB flow control holds the radio, so nothing is dropped. The ring is 256 KB (the size the header already declared as MAX_PORT_FIFO_LEN, and a power of two as the ring's mask needs); the per-read chunk stays 8 KB on the thread stack. Dart is called back during a burst (4 KB waiting or 10 ms) and whenever the ring fills, not only when the port goes idle. Same change in the macOS copy of the source (macos/Classes), which was identical. (std::min) is parenthesised because the Windows headers define a min macro. Verified on the owner's RK4 (read-only harness driving the DLL): new DLL 350/350 in every mode, including a reader 2 s late, zero bytes skipped; the old DLL, as the control, got 345 app-like and 54 (8191 bytes, the old ring's capacity) when 300 ms or 2 s late. Epic #702. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	1 month ago
Strycher	3f4bd23f08	chore(#709 ): dart format the vendored flserial Dart files for CI Whitespace only, from dart format, so CI's repo-wide format check passes. Same step #696 took for the vendored Android BLE plugin. Epic #702. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	1 month ago
Strycher	c1ffa3302b	chore(#709 ): vendor pristine flserial 4821631 into packages/flserial Pristine import of MeshEnvy/flserial at 48216310061efc8d5d217cc18014fc2cb501646e (MIT, Grzegorz Lesniak), the exact commit the app already pinned, into packages/flserial, with the dependency switched from git to a path. Mirrors the #695/#686 vendor pattern. Vendored with git itself (fetch of that commit, read-tree --prefix), so every file is the upstream blob: all 70 files verified blob-for-blob against the commit. Left out, as #695 did: example/, .github/ and .vscode/. No behaviour change. The Windows release build compiles the plugin from the vendored path and produces a flserial.dll of the same size as the installed one. pubspec.lock also picks up flutter_blue_plus_android 8.2.2+offband.1, the version dev's own vendored package already declares. The fix for the 8 KB ring that overwrote contact-sync bytes (#704) follows in #710. Epic #702. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	1 month ago
diffwireauto	a04db40b8a	fix(#717 ): apply pre-PR review, rebuild MTU subscriptions once services resolve Gemini gate found a real defect that would have shipped: the Linux MTU stream would have attached to nothing and reported an MTU for a device never. `devicesChanged` fires only on deviceAdded / deviceRemoved. At the moment a device is added BlueZ has not finished service discovery, so `gattServices` is still empty; this plugin waits on `while (!device.servicesResolved)` elsewhere for exactly that reason. The original patch built its per-characteristic subscriptions inside that one event, found no characteristics, and had nothing to re-trigger it when discovery later completed. Verified in the vendored source, not taken on the reviewer's word. Now nested: for each device, rebuild the characteristic subscriptions on every `device.propertiesChanged`, which includes ServicesResolved flipping true. Same shape the plugin's own `_initFlutterBluePlus` already uses for its logging, so this is the file's existing idiom rather than a new one. Also emits once per rebuild (`startWith(device)`), so an MTU already negotiated before we were listening is reported rather than waiting on a change that may never arrive. Safe when unknown: 0 is filtered downstream. Two other findings reviewed and accepted as clean, recorded so they are not re-litigated: "max across characteristics" is right because the ATT MTU is per-connection, not per-characteristic; and the web exemption is sound because the browser owns fragmentation and 172 is far inside the spec's 512-byte writeValue limit. analyze clean, format clean, 981 tests pass. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	1 month ago
diffwireauto	ce11da8295	fix(#717 ): report the negotiated MTU on Linux so the composer is usable Upstream flutter_blue_plus_linux returns `Stream.empty()` from onMtuChanged, so no MTU ever reaches Dart on Linux. FlutterBluePlus._mtuValues stays empty, mtuNow falls back to its 23-byte ATT default forever, effectiveMaxFrameSize returns the 20-byte floor, and maxChannelMessageBytes clamps to 0. Every keystroke is then rejected by the length limiter while the field still focuses and blinks: the #592 symptom, reported on Linux as locked entry fields. BlueZ does expose the negotiated MTU, as a property on org.bluez.GattCharacteristic1, surfaced by the bluez package this plugin already depends on (BlueZGattCharacteristic.mtu). The value was one layer down the whole time and simply never read. The patch watches characteristic property changes for 'MTU' and emits a BmMtuChangedResponse carrying the largest value across the device's characteristics. They describe one ATT link so they should agree; the max stops a characteristic that has not been exercised yet, and so reports nothing, from dragging the figure down. Zero is filtered out, because FlutterBluePlus would cache it and hand callers something worse than its own default, and identical consecutive values are suppressed. Nothing downstream needed changing: the device.mtu subscription added for Windows in #686 is platform-agnostic, so Linux now gets the same live budget recovery through the same path. Also points dependency_overrides at the vendored copy. Third vendored BLE backend, alongside android (#694) and winrt (#686). Verified: analyze clean, format clean, 962 tests pass. NOT verified on Linux hardware; nobody here runs Linux BLE, which is why the gap survived this long. #717's acceptance requires a real Linux BLE connect showing a non-default `MTU now:` in the app log. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	1 month ago
diffwireauto	dede4d6a6a	chore(#717 ): dart format the vendored Linux plugin Dart file for CI CI runs `dart format --output=none --set-exit-if-changed .` across the whole repo, vendored packages included, so the upstream file's own formatting fails the analyze check. Third time: #694 hit it on the Android plugin (`b33e8485`, chore(#696)) and #686 on the WinRT one. Formatting only, kept separate so the MTU patch that follows is reviewable instead of buried in reflowed lines. Re-vendoring upstream later means re-running dart format on this file. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	1 month ago
diffwireauto	754dca86f6	chore(#717 ): vendor pristine flutter_blue_plus_linux 8.2.2 Pristine import of pub.dev flutter_blue_plus_linux-8.2.2 into packages/, mirroring the WinRT vendor from #686 and the Android vendor from #695. The patch rides on top as its own commit so the upstream baseline stays diffable. Verified against the pub cache with `diff -r --strip-trailing-cr`: all 8 files identical. FlutterBluePlus License v1.4 and the BSD NOTICE retained unchanged. No wiring yet: dependency_overrides is not pointed at this copy until the patch lands, so this commit is inert. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	1 month ago
diffwireauto	307777f411	chore(#686 ): dart format the vendored WinRT plugin Dart file for CI CI runs `dart format --output=none --set-exit-if-changed .` across the whole repo, vendored packages included, so the upstream file's own formatting fails the analyze check. Same wall #694 hit on the vendored Android plugin, resolved the same way there (`b33e8485`, chore(#696)). Formatting only: no logic change, and the C++ that carries the actual fix is untouched. Re-vendoring upstream later means re-running dart format on this file, which the two-commit structure (pristine import, then patch) keeps obvious. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	1 month ago
diffwireauto	9a472be85f	fix(#686 ): create GattSession on connect so Windows reports negotiated MTU Upstream flutter_blue_plus_winrt never populates gatt_sessions_ or mtu_tokens_, so OnMtuChanged never fires and the Dart MTU cache stays at the 23-byte default for the whole session (#683 diagnosis). Add CreateGattSessionAsync, fired after the connect result is reported: GattSession::FromDeviceIdAsync, store under gatt_cache_mutex_ (never held across a co_await), subscribe MaxPduSizeChanged, emit the initial value, and undo via ClearDeviceResources if the device drops mid-creation. MaintainConnection is never set (#688). The destructor now revokes MTU tokens and closes sessions, same use-after-free class upstream fixed for connection tokens in 7b4504a2. Existing ClearDeviceResources and the requestMtu read path work unchanged. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	1 month ago
diffwireauto	cf0332612f	fix(#686 ): wire the app to the vendored Windows BLE plugin + live MTU updates App-side half of the #592 fix; the plugin-side GattSession patch is a separate commit on packages/flutter_blue_plus_winrt. - dependency_overrides points flutter_blue_plus_winrt at the vendored copy (version 0.0.20+offband.1; interface constraint relaxed to >=8.2.2 since the Dart layer uses nothing 9.x-only - the 9.x/2.3.x upgrade was deliberately skipped, owner decision on #685). - Connector: requestMtu is now attempted on Android only. flutter_blue_plus throws "android-only" on every other platform before any platform code runs (#683), so the old call could never succeed off-Android and its failure log was noise. - Connector: subscribe to device.mtu and notifyListeners on each value. The composers read effectiveMaxFrameSize on rebuild, so this is what lets the message byte budget recover from the 20-byte floor the moment the platform reports the real MTU, instead of waiting for an unrelated rebuild. Subscription is cancelled on reconnect and disconnect. - test/connector/mtu_composer_budget_test.dart: characterizes the collapsed budgets at the unknown-MTU floor (DM 4 bytes, channel 1 or 0) - the numbers both field reports reduced to - and the healthy-link budgets. Rescued from the throwaway diagnosis branch (#683). - Regenerated plugin registrants picked up plugins missing from the committed copies (file_selector, screen_retriever, window_manager); stale on dev, refreshed by pub get after the override. analyze clean, 958 tests pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	1 month ago
diffwireauto	01db7eb621	chore(#686 ): vendor flutter_blue_plus_winrt from upstream master 7b4504a2 Pristine, unmodified import of chan150/flutter_blue_plus_winrt at 7b4504a2bf8409e7f01118c870ead435c1b1d681 (MIT, Copyright 2025 Himchan Park; LICENSE retained). All 14 files hash-verified against the fetched upstream tarball. Vendored rather than depended on because no published version reports the BLE MTU on Windows: gatt_sessions_ is declared but never populated, so the MTU stays at the 23-byte default and the composer is capped at 0-4 bytes (#592, diagnosed in #683). Our fix lands as a separate commit on top of this one so it can be reviewed and re-applied against a future upstream. Based on master rather than 0.0.19 to pick up two unreleased upstream fixes: 86ef4763 (data race, unbounded caches, teardown use-after-free) and 7b4504a2 (revoke per-device ConnectionStatusChanged handlers on teardown). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	1 month ago
Strycher	b33e8485f3	chore(#696 ): dart format the vendored plugin Dart file for CI CI's format gate reflows upstream's line style; formatting-only deviation from pristine, no content change (same class as the documented EOL whitespace normalization). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2 months ago
Strycher	74e461ba48	fix(#699 ): apply pre-PR review, never block the main thread in the backstop Gemini review (gemini-2.5-pro, docs/llm-consultations) flagged that the 5 s backstop acquired mMethodCallMutex blockingly on the main looper, an ANR class. It now tryAcquires and re-posts itself after 250 ms when a callback holds the mutex. The other two findings were verified and rejected with reasons recorded in the PR body: the disconnectAllDevices race is pre-existing upstream (and naive mutex acquisition there would deadlock the flutterRestart path, which already holds the semaphore), and the untranslated new strings follow this repo's standing l10n process. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2 months ago
Strycher	8a7946972b	fix(#696 ): close leaked GATT handles in the vendored Android plugin Three changes to FlutterBluePlusPlugin.java, per the #689 diagnosis: 1. onConnectionStateChange(CONNECTED): close the handle a map overwrite replaces. A replaced handle is unreachable forever and holds the link and a GATT client slot until process death, which is exactly what the owner's force-stop evidence showed. 2. disconnect: 5 s backstop. If Android never delivers DISCONNECTED for the handle we asked to disconnect, force gatt.close(), untrack it, and synthesize the disconnect event (mirrors the existing cancel-while-connecting path, reason code 23789259). Also caps the Dart layer's 35 s wait at ~5 s. 3. disconnect with no tracked handle now logs at WARN so the orphan fingerprint is visible in exported logs. Vendored version marked 8.2.2+offband.1. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2 months ago
Strycher	3944895550	chore(#695 ): vendor pristine flutter_blue_plus_android 8.2.2 Pristine import of pub.dev flutter_blue_plus_android-8.2.2 into packages/, wired via dependency_overrides, mirroring the WinRT vendor pattern from #686. FBP License v1.4 and BSD NOTICE retained. Verified against the pub cache file-by-file: 9 of 12 files byte-identical; AndroidManifest.xml, LICENSE and FlutterBluePlusPlugin.java differ ONLY in end-of-line whitespace (1, 4 and 11 lines respectively; the byte-perfect license text also ships as LICENSE.md, which is byte-identical). No content changes; verified with git diff --ignore-space-at-eol. Release APK builds clean against the vendored copy before any patch. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2 months ago

16 Commits (3f1997e8d29a2e7e87d7732afef424e17c100e2d)