A discover response often arrives twice — the zero-hop direct copy and a
re-flooded copy relayed by another repeater carry different packet hashes,
so the mesh duplicate filter passes both. The standalone scan appended the
same node twice and app-driven discover forwarded both copies, so one
repeater showed up as two.
Track (tag, responder pubkey) for a short window in isDupDiscoverResp() and
drop the second copy in both the standalone and app-forward paths.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- locked device: a ringing alarm/timer was invisible — wakeForAlarm() now
holds the lock wake window for the whole ring, the lock screen draws the
alert overlay, and a key-dismissed ring falls back to a 5 s glance
- status bar: A / live-share / trail / repeater icons no longer vanish when
Bluetooth is off (moved outside the isSerialEnabled() gate)
- alert overlay: long text wraps to up to 3 lines inside the box instead of
overflowing the border (new UITask::renderAlertOverlay, shared with lock)
- MyMesh: force NUL on contact.name copied from an app frame (unterminated
name could overrun the AdvertPath strcpy)
- clock tools: ring/timer deadline compares now wrap-safe (signed diff),
matching the trail/loc-share timers
- messages: channel context menu freezes its target channel at open; fav
toggle no longer retargets the menu when the fav-only filter drops the row
All three solo envs build green (WioTrackerL1 OLED/Eink, GAT562-30S).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Surfaces the Alarm/Timer/Stopwatch screen (previously reachable only
from the home Clock page) as a System tool, via the existing
UITask::gotoClockTools() and ICON_ALARM.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The shared on-screen keyboard seeds {loc}/{time} placeholders on begin()
for message composition, but they make no sense in a radio-preset name.
Clear them right after opening the keyboard from the "+ Save current..."
flow in both Settings > Radio and Tools > Repeater; the message-slot
editor still keeps them.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
GPS support was implemented incorrectly for the ThinkNode M3. The reset
pin was being driven when it should be left floating for this unit. The
enable pin also wasn't being picked up by `MicroNMEALocationProvider`
because of a mismatch in constant naming conventions. I did a general
cleanup of the GPS and ThinkNode M3 bring-up code so that constant names
line up and "*_ACTIVE" constants are used consistently vs hardcoding
`HIGH`/`LOW`. After making these changes, serial data immediately starts
streaming in from the NMEA on boot and GPS detection just works.
LED handling was also not quite right for the ThinkNode M3. The LoRa TX
LED was being driven high to turn it on when it should actually be
driven low. I changed the code to use the `LED_STATE_ON` constant and
also added a little code to the shutdown path to properly make sure that
all LEDs are turned off.
Tested and confirmed working on real hardware.
Resolves both #1864 and #2879.
The per-read VBAT_ENABLE gating from 4d46abb2 (energy optimizations)
powered the battery voltage divider HIGH for only 10ms before each ADC
read, then LOW. 10ms is too short for the high-impedance divider node
to settle before the nRF52 SAADC samples it, so readings came out high
and inconsistent (e.g. 4.6-5.0V on a LiPo that maxes ~4.2V).
Hold VBAT_ENABLE HIGH continuously again (as before 4d46abb2) so the
node is always settled at sample time, and drop the toggle + delay(10)
from getBattMilliVolts(). Standby cost is ~2uA on the 2x1M divider,
negligible next to the device's mA-level draw.
The CPU-sleep and LED-off energy optimizations from 4d46abb2 are left
untouched — those are the real savings; only the VBAT gating is reverted.
The on-screen keyboard grid has been a-b-c...z alphabetical order since
it was first introduced, never an actual QWERTY row layout. The
"QWERTY" name only appeared with the T9 option (needed something to
contrast it against) and was inaccurate from the start.
Merges PR #21 (T9 on-screen keyboard option, marczykm) plus a follow-up
fix bumping SCHEMA_SENTINEL to 0xC0DE001A — the PR added a new persisted
field (keyboard_type) without bumping the schema version, which would
have silently misaligned every NodePrefs field after alarm_min for
devices already on the previous schema.
keyboard_type was inserted into the append-only NodePrefs layout without
bumping SCHEMA_SENTINEL, leaving it at 0xC0DE0019 — the same value plain
main already uses without this field. A device already on that schema
would silently misalign every field read after alarm_min once updated
to a build containing this change. Bump to 0xC0DE001A and fix the
migration comment's stale version label (it said 0xC0DE0018, which was
already taken by the MAP home-page migration).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Growing page_order 11 → 13 bumped sizeof(NodePrefs) 2488 → 2496 (the 2 added
bytes plus alignment padding to the next 8-byte boundary). Save/load/clamp and
SCHEMA_SENTINEL were already updated; this syncs the static_assert. Verified
with a host compile (fixed-width members → layout matches the target).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Map page had no Settings entry and Shutdown was pinned to the end of the
carousel because page_order held only 11 slots — full in the common
GPS+SENSORS config, so both pages fell back to being appended at nav time and
couldn't be moved. Grow page_order to 13 (== HPB_COUNT) so every page has a
reorderable slot, and add Map + Shutdown to the default order and the
required-append list.
To keep the persisted format backward-compatible (page_order sits mid-record,
not at the tail), the on-disk head stays the original 11 bytes at its existing
offset and the 2 new slots are appended at the file tail, matching the
append-only schema design. Bump SCHEMA_SENTINEL to 0xC0DE0019; on a pre-0x19
save the tail bytes are the old sentinel/EOF, so they're clamped to 0 and
ensurePageOrderInit re-appends Map/Shutdown into the freed slots on first use.
Drop the now-dead Shutdown-eviction path (13 slots fit all pages).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The prefs schema-migration block runs on every SCHEMA_SENTINEL mismatch, and
the "turn Favourites on" step was ungated. Since each firmware release bumps
the sentinel, every update re-applied `home_pages_mask |= HP_FAVOURITES`,
clobbering a user who had hidden it. Gate it to the transition that added
Favourites (sentinel < 0xC0DE0002), matching the trail_units_idx migration.
The Map/Trail carousel page shipped with no visibility toggle: pageBit(MAP)
returned -1 so it was always visible with no Settings row. Add a proper "Map"
toggle (new HPB_MAP/HP_MAP bit, pageBit/bitToPage cases, HOME_MAP settings
row) plus a gated one-time migration (sentinel < 0xC0DE0018) that keeps it
visible for upgraders. Bump SCHEMA_SENTINEL to 0xC0DE0018; page_order[] stays
a literal 11 so the persisted layout is unchanged.
Also fix the fresh-install seed, which used a stale 0x01FF literal that left
Favourites hidden on new installs despite its "all pages visible" comment;
seed NodePrefs::HP_ALL so fresh installs match upgraders (Favourites + Map).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The `recv_pkt_region` is set when processing a flood packet in `filterRecvFloodPacket`
but direct/non-flood packets would never pass through that function, so the pointer was
not cleared for them.
`sendFloodReply` would then later use it blindly, which meant that the response would
either inherit the region from the last flood packet, or refer to a non-initialised pointer
if no region floods had been received yet.
hasSeen() was simultaneously a predicate and a mutator — it inserted the
packet hash on every miss, making five call sites that only wanted to mark
a packet as sent call it with the return value discarded.
Split into:
- wasSeen() — pure predicate, no side effects
- markSeen() — explicit insert
All query sites now call markSeen() immediately after wasSeen() returns
false, preserving identical runtime behaviour. The five mark-only send
sites (sendFlood, sendDirect, sendZeroHop x2) now call markSeen directly.
Also fixes three bridge sites (BridgeBase, ESPNowBridge, RS232Bridge)
that had the same query+implicit-insert pattern.
Tests: add test/test_mesh_tables/ covering wasSeen purity, markSeen,
dup stats, and clear. Update SHA256 mock to produce deterministic output
(previously finalize() was a no-op). Add Packet.cpp to native build filter.
- release-notes v1.21: add Clock tools (alarm/timer/stopwatch) to What's new,
e-ink button responsiveness + double-tap debounce to Fixes, and the input
edge-capture/key-queue and hardware-timer ringtone notes to Under the hood.
- README: note clock tools on the Clock Screen entries.
- solo_ui_framework: document the hardware->handleInput path (mandatory
begin() per button, IRQ edge capture, key queue + coalesced redraw).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A single button press could surface as two CLICKs on the e-ink build, most
visibly as start+stop on the stopwatch. Two contact-bounce paths fed the
IRQ edge-capture machinery a phantom press/release pair:
- a bounce edge accepted just after a clean release was replayed as a real
press — ISR_DEBOUNCE_MS (5 ms) was too short for the joystick switch; raised
to 25 ms so the settling burst is swallowed.
- the live-pin self-heal reconciled prev against a single raw digitalRead,
which can sample a bouncing contact mid-flap and synthesise a transition.
It now acts only once the divergence has been stable for ISR_DEBOUNCE_MS, so
a momentary read can't inject a click; a genuinely lost edge still heals
(~25 ms later) so the button can't stick.
Both thresholds stay far below any human tap cadence (>100 ms), so rapid
multi-tap navigation still registers every tap.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The joystick directions and Back were never begin()'d, so on the e-ink build
they never claimed a GPIOTE channel — edges landing during a blocking panel
refresh were lost and the IRQ-capture work didn't reach them. begin() them.
Even with edges captured, rapid taps of one direction replayed into a single
check() and collapsed into a double/triple-click event the navigation handler
ignores, and the loop only ever dispatched one key per render. So:
- MomentaryButton: for multiclick=false buttons emit one CLICK per completed
release (one per check() call) instead of collapsing — each tap stays a
discrete key. multiclick=true buttons (double/triple) are unchanged.
- UITask: add a key FIFO; drain each direction fully into it, apply the whole
queued burst, then redraw once. N taps captured during a refresh become N
navigation steps at the cost of a single refresh. Also fixes losing a key
when two buttons fire in the same loop iteration.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Clock Tools: one-shot alarm, countdown timer and stopwatch on the Clock
page, with the alarm/timer/ring engine in UITask (sync-robust absolute
alarm scheduling; millis-based timer/stopwatch).
Add a Clock Tools screen (Enter on the home Clock page) with three time
utilities, plus the engine that drives them from UITask::loop() so they
work regardless of the current screen / display state:
- Alarm: a single one-shot wake alarm. Persisted in NodePrefs as a local
time-of-day; UITask schedules it as an ABSOLUTE fire instant recomputed
from that time, so it survives RTC re-syncs (mesh/app/GPS/CLI all jump
the clock) — small corrections still fire on time, a jump over the
target still fires (late, up to 6 h). Disarms after firing. A bell
icon signals an armed alarm on the clock face and the top status bar.
- Timer: a millis-based countdown (sync-immune), big HH:MM:SS readout,
rings even when off-screen.
- Stopwatch: millis-based, keeps running in the background.
Numeric fields use the shared framework DigitEditor (digit-by-digit,
min/max enforced). The alarm/timer/ring engine lives in UITask alongside
the locator/live-share engines (no screen-cast for time-critical logic).
E-ink: live readouts refresh coarsely (and on any key); timing is exact
regardless, and the countdown's buzzer fires on time. Rings override mute
and are dismissed by any key (auto-stop after 1 min). Cannot wake from a
full Shutdown (CPU/RAM powered down).
NodePrefs: +alarm_on/alarm_hour/alarm_min (fit existing tail padding,
sizeof unchanged 2488), SCHEMA_SENTINEL 0xC0DE0016 -> 0xC0DE0017,
DataStore read/clamp/write in lockstep.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Added build flag "SX126X_USE_REGULATOR_LDO" (set to 1 to enable) to use LDO for the SX1262 radio instead of DC-DC when the DCDC pin isn't wired up (e.g. t-echo lite)
- Added additional debug mode diagnostic messages during SX1262 init to better highlight faults
Extract the DM + channel message-history rings from QuickMsgScreen into a
self-contained MessageHistory store, dedupe the notif/melody override
accessors, share the history row-frame / [+ send] button between the two
history lists, and fix the [+ send] row vanishing in room/DM history.
Hardware-validated before merge.
The DM/room and channel history lists drew each message box's selection frame
and the bottom-left [+ send] button with byte-identical inline code. Extract
two small static helpers (drawHistRowFrame, drawComposeButton) and route both
lists through them.
Only the parts that were already identical are unified — the divergent
per-type logic (sender parsing, ACK semantics, body handling) stays inline, so
this is behaviour-preserving. The DM compose button now also gets the channel
list's border (and its always-reset-to-LIGHT), folding in the earlier
[+ send] visibility fix.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
In the DM/room message history, the message loop leaves the ink DARK when the
last drawn box is the selected one, so the unselected [+ send] row printed
dark-on-dark and vanished. Most visible in room servers, where there's always
a selected message. Reset to LIGHT before drawing the row (matching the
channel-history version). Pre-existing bug, independent of the history-store
refactor.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The eight per-contact/channel notification + melody accessors were four
near-identical copies of two storage shapes. Collapse them onto two
primitives:
- maskPairGet/Set — a 3-state packed in a (presence, variant) channel-index
bitmask pair. The notif and melody uses disagree on which state the variant
bit means, so the caller passes the state value for variant-set (v_set).
- prefTableGet/Set<Entry> — the {prefix[4], value} DM table find/update/clear/
insert/overwrite-slot-0 logic, templated over the entry struct + a member
pointer to its value field (DmNotifEntry::state / DmMelodyEntry::slot).
The eight accessors are now thin wrappers; behaviour identical (verified
state↔bit mappings). QuickMsgScreen 1738 -> 1715 lines.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Move the two RAM history rings (channel + DM) out of the 1988-line
QuickMsgScreen into a self-contained MessageHistory.h component, mirroring
the WaypointsView extraction from TrailScreen. The store owns the ring
buffers, per-entry delivery state (channel relay echo, DM end-to-end ACK +
auto-resend) and the per-channel unread counters; the screen keeps all
view state (selection, scroll, fullscreen readers, the unread "viewing
session" bookkeeping) and reaches entries through accessors.
Behaviour-preserving:
- The shared types (AckState, ChHistEntry, DmHistEntry, MSG_TEXT_BUF) are
file-scope in MessageHistory.h, so the phase machine still refers to them
unqualified — only data + storage logic moved.
- addChannelMsg keeps a thin screen forwarder that computes the "viewing"
flag (a phase fact the store can't see) and returns the ring pos;
afterSend uses armChannelRelay(pos, seq) instead of poking the ring.
- The public API called via UITask/MyMesh/the bot (addChannelMsg, addDMMsg,
markDmDelivered, markChannelRelayed, tickDmResends, clearAllChannelUnread,
getTotalChannelUnread, getRecentDMContacts) stays on QuickMsgScreen as
forwarders, so no caller changes.
QuickMsgScreen 1988 -> 1738 lines; MessageHistory.h 324. Both solo boards
build. Single-TU fragment: included by UITask.cpp before QuickMsgScreen.h.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The M6's L76K GPS streams NMEA at 9600 baud on its own, but the firmware
reported no GPS. Root cause: pin 29 (PIN_GPS_RESET / the module's REINIT
line) was driven HIGH, which holds the L76K silent so it never emits any
sentences and detection fails.
variant.cpp drove pin 29 HIGH at boot, and because PIN_GPS_RESET was
defined, MicroNMEALocationProvider also drove it HIGH in begin(). Bench
testing on a sealed M6 (passive NMEA capture, no logic analyzer) confirmed:
pin 29 driven HIGH = 0 bytes; pin 29 floating = full NMEA stream.
Define GPS_RESET (-1) so the location provider never touches pin 29, and
stop driving it in initVariant. This matches the Meshtastic M6 variant,
which leaves the same pin as a floating input.
Add Tools › Trail → Hold Enter → Track back (shown when the trail has ≥2
points). It reuses NavView but, instead of a single fixed target, snaps onto
the route at the nearest recorded breadcrumb and walks the points in reverse,
auto-advancing the target as each is reached (within 20 m). The header shows
points remaining ("Back: 12 pt" → "Trail start"); reaching the start toasts
"Back at start" and exits. Cancel leaves at any time.
Lives in WaypointsView as a new sub-mode; advancement runs in poll()
(forwarded from TrailScreen) so it tracks GPS independently of the render
cadence. An EtaTracker drives the approach/ETA line and is reset on each
breadcrumb advance to avoid a bogus closing-speed spike.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add a "Mark avg" trail setting (Off/5/10/30 s). When set, Tools › Trail ›
Mark here samples the GPS fix once a second for the chosen window and stores
the mean position instead of one instantaneous fix — a steadier, more
accurate mark for a precise spot. A progress screen shows time left + sample
count; Cancel aborts; the window then opens the label keyboard as usual.
Off (default) keeps marking instant.
Sampling runs in WaypointsView::poll() (forwarded from TrailScreen::poll())
so it ticks on the main loop, independent of the slow e-ink render cadence;
int64 accumulators avoid overflow over 30 samples.
Persisted as NodePrefs::gps_avg_idx (schema 0xC0DE0016): struct field +
option table, DataStore rd/write/clamp in lockstep. sizeof unchanged (the
uint8_t fits existing tail padding) so the 2488 tripwire still holds.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three spots lagged the recent refactors:
- §8 + §10 still taught the inline `if (_dirty) { savePrefs(); _dirty=false; }`
pattern; now point at `_task->savePrefsIfDirty(_dirty)`.
- §1 wiring contract now notes the nullptr-init / setCurrScreen null-guard
fail-safe (steps 1/3/4 are compile-checked, only a missed `new` slips through).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Every drawList/AccordionList row opened with the same hand-written
display.drawSelectionRow(0, y-1, width-reserve, lineStep-1, sel) line —
14 copies of the same geometry and magic offsets. Add a drawRowSelection(d,
y, sel, reserve) helper in icons.h next to drawList and route the canonical
sites through it (Bot/LiveShare/Locator/Repeater/Settings/Tools/Nearby/
QuickMsg/Waypoints).
Rows that intentionally differ (full-width DashboardConfig/QuickMsg, the
keyboard/ringtone grids) keep their explicit drawSelectionRow() call — the
helper is opt-in rather than baked into drawList, so legitimate variants
aren't forced into one geometry. Framework doc updated to match.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The "_dirty bool, then if (_dirty) the_mesh.savePrefs() on exit" pattern was
duplicated across 9 screens with subtle divergence: some cleared the flag
after saving, some left it set and relied on onShow() to reset. Replace all
12 exit sites with a single _task->savePrefsIfDirty(flag) helper that saves
once only if dirty and always clears the flag, so the "did we touch flash?"
answer lives in one place and the reset is consistent.
Edit sites still mark the flag manually (inherent to change tracking); only
the persist-on-exit boilerplate is centralised.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adding a screen touches 4 sites; 3 (member decl, gotoX decl, gotoX def) are
compile-checked, but a missed `new XScreen()` in begin() left the pointer
uninitialised and crashed at first navigation. Give every screen member an
in-class nullptr initialiser and bail early in setCurrScreen(nullptr) so the
mistake is an inert no-op instead of a null deref. Document the 4-site
registration contract on the member block.
A full registry table was considered and rejected: the named gotoXScreen()
methods are a depended-upon API (~30 call sites, menu dispatch + back-nav),
so a table would add an enum + indirection without removing them.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Kevin Le
JakubandClaude Opus 4.8
liamcottle
Quency-DandGitHub
ripplebizandGitHub
Naomi Smith
Marcin MarczykandGitHub
MarekZegare4andClaude Opus 4.8
Neil Alexander
Christoph Koehler
entr0p1
Jody Bentley