- A restart between commitTempFile()'s remove and rename left only the .tmp;
loads now recover it (prefs, contacts, channels, scopes).
- CustomLR1121Wrapper lacked readLiveRSSI/readLiveSNR: LR1121 boards didn't
compile.
- The repeater and room server sims gain DynamicConfigSerializer.cpp and
CommonRadioPrefs.cpp (broken since the v1.18 merge).
- Comments still describing the removed DualSerialInterface.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- DisplayDriver takes ColorVal; each driver's .cpp defines the UIColor roles
(window_bkg, primary_txt, ...) in its own values. ui-new draws in two:
window_bkg and primary_txt. OLED and e-ink look as before (sim screens
pixel-identical).
- E213, E290, LGFX, Null, ST7735, ST7789LCD, U8g2 taken whole from upstream
(with them the rewritten ST7735 driver and the Heltec V4 R8 TFT display).
- SH1106, SSD1306, ST7789, EinkGfxDisplay and the sim drivers converted.
- ui-orig, ui-tiny, simple_* get upstream's colour roles; the repeater's UI
taken from upstream (its power-off screen).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Bluetooth + USB (+ WiFi/Ethernet/UART per build) in one interface_manager,
as upstream; frames go to every enabled link, USB is read while BT is
connected. DUAL_SERIAL now means ENABLE_USB_INTERFACE.
- WiFi interface with upstream's wifi_* prefs and set/get wifi.* CLI.
- isClientConnected moved into ArduinoSerialInterface (DTR on nRF52 and
ESP32 native CDC).
- UI uses isBluetoothEnabled/enableBluetooth/disableBluetooth; turning BT
off leaves USB and the other links on.
- DualSerialInterface removed.
- Sim build gains DynamicConfigSerializer.cpp (broken since bd64c203).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
NodePrefs is now NodePrefsFields (the fields) plus how they're saved: one
JSON key per field, upstream's keys under "radio", "gps", "repeat", "comp"
and "custom" (so either firmware reads what the other wrote of them),
ours under "solo". Saved atomically (temp file, then renamed), through a
buffer. 2.0-rc's binary /prefs is removed, not converted: 2.0 starts from
the defaults; /new_prefs from before 2.0 is still imported.
From upstream with it: txdelay / direct.txdelay / agc.reset.interval
(prefs and CLI), and the board's own settings through getCustom()
(Board::attachDynamicPrefs) in place of the merge's stopgap adapter.
ConfigSerializer reads values up to 191 characters (was 127): 140-byte
texts, an 80-byte table in hex.
Test: every field round-trips through the JSON byte for byte; the alarm
label tests expect "Once" (the label changed).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fixes and features from upstream: MACThenDecrypt / command buffer
overflow fixes, ACK forwarding delay, transport code 0 as unscoped,
hidden nodes don't answer DISCOVER, LR2021/LR1121 fixes, GPS cold-start
detection and waitFor()/drain(), NMEA fix-age counters, ESP32 RTC after a
crash, remote CLI by text message (TXT_TYPE_CLI_COMMAND),
PUSH_CODE_BINARY_RESPONSE, board.loop(), new boards (Wio Tracker L1 1W,
ThinkNode M4/M8, Muzi Base, Waveshare LR1121, Pico W).
Kept ours: NodePrefs, main.cpp, every UI, GxEPDDisplay. The companion CLI
stays MyMeshCli.h's handleCliCommand(); upstream's handleCommand() calls
it. The board's FEM gains go through attachDynamicPrefs() with an adapter
over our NodePrefs fields. Upstream's JSON prefs and MultiSerialInterface
come as their own stages.
Also: the display defaults were inside RADIO_FEM_TXGAIN's #else (a
missing #endif).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Core: PathEditor (UiCore::route) loads, edits, traces and saves the route to
a contact; MyMesh::sendTrace over an explicit route with per-hop SNRs,
setContactPath / setContactPinned. A route can be kept
(CONTACT_FLAG_ROUTE_PINNED, ContactInfo::flags bit 4): a path heard from the
node doesn't replace it and the last DM resend doesn't drop it.
ui-lvgl: Route screen (node detail), repeater picker by distance, picking
hops by tapping the nodes map (numbered badges, line, Undo / Done). The
map's markers are one drawn layer with hit-testing and name decluttering
instead of three objects per node.
ui-new: Path / trace view in the Nodes action menu; Add reuses the list
filtered to repeaters.
First cut: built for L2 LVGL, L1 OLED and L1 e-ink, tried on L2 only.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
MicroNMEA's valid flag survives a power cycle, so the duty cycle saw the
old fix right after waking the receiver and stopped it again a tick
later: with GPS power saving on the position never updated. begin() now
clears the parser.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With the phone not connected, every frame (each packet heard, each advert)
went to the USB port. With nobody holding it open the core either drops the
bytes or lets its buffer fill, and streamWriteUntil() then spun out its whole
timeout. Skip the write when the native USB Serial has no host.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Clip requested power to the active PA's range instead of failing. A
stored 22 dBm was rejected at 2.4 GHz (HF PA max 13 dBm), and since
setTxPower ignores errors the sub-GHz PA stayed configured.
- Re-apply the requested power after every successful retune: PA choice
depends on the band, and set radio / tempradio only call setParams.
ESP32-S3FH4R2 (4MB flash, 2MB PSRAM, native USB-CDC) with an LR1121.
Pins, the DIO5/DIO6 RF switch table and the 1.8 V TCXO follow the HF
schematic and Waveshare's examples. Envs: repeater, room server,
companion USB/BLE and KISS modem, all on min_spiffs so roles share one
partition layout.
Only RFO_HP_LF is wired (R4 on RFO_LP_LF is not fitted), so the new
opt-in LR1121_FORCE_HP_PA flag keeps RadioLib on the HP PA at <= 14 dBm,
where it would otherwise select the unconnected LP PA.
Review follow-up:
- The rows exist only where the build asks for them: Text size with
EINK_FONT_SCALE, Icon size with STATUS_ICON_SCALE. The stock 2.13", the
portrait builds and the L2 don't get them -- their screens are laid out
for one cell size. The text override also applies in landscape only.
- Stored in NodePrefs (ui_text_scale, ui_icon_scale) instead of a
/ui_scale file, as rows in the schema's Display section: the usual row,
save path and factory reset, and console keys display.text.size /
display.icon.size. The hand-written Settings rows and the load/save code
are gone; UITask::applyUiScale() applies the prefs at boot and on change
(applyDisplayPrefs). On opt-in builds ui_text_scale starts at the
build's EINK_FONT_SCALE (DataStore sanitize), so the row shows the scale
in use.
- EINK_FONT_SCALE, when set, now scales whichever landscape font the build
picks (the 6x9, or the EINK_LARGE_FONT one); unset, it is stock as before.
- setUiScale() skips the re-layout and full redraw when nothing changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
setRamArea() and fastWrite() follow GxEPD2 1.6.2's private
_setPartialRamArea() and _writeImage() for the GDEQ0426T82; say so, so
whoever bumps the library knows to re-check them.
On an 800x480 panel the doubled 6x9 font and the status bar's icons are
tiny. Two opt-in build flags, each also a setting under Settings > Display:
- EINK_FONT_SCALE=<n>: the landscape font's pixel multiplier (stock 2).
Settings > Display > Text size (1x-5x) overrides it at run time
(DisplayDriver::setUiScale -> EinkGfxDisplay re-lays out).
- STATUS_ICON_SCALE=<n>: Home's status icons, battery and page-icon row
drawn at n x their pixel grid, the bar growing to fit. 0 = Auto, 1.5x
the text's pixel scale. Settings > Display > Icon size (Auto, 1x-8x).
Both live in their own /ui_scale file, so NodePrefs is untouched. Icon size
appears only where STATUS_ICON_SCALE is set; Text size appears on every
e-ink build with the full-refresh setting, and builds that set neither
flag draw as before until it is changed. The GDEQ0426T82 env defaults to
4x text with Auto icons.
Tested in the native sim at 800x480.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KxAZVAbhhYr18Dd4guTQM1
GxEPD2 sends the panel RAM a byte at a time; on the nRF52's SPIM the
per-call overhead dominates, and an 800x480 partial refresh writes the
whole 48 KB several times. GxEPD2_426_GDEQ0426T82_Fast overrides the three
image writes to send each row in one transfer() call, falling back to the
stock path for the cases it doesn't cover (first write, mirror_y, PROGMEM).
Byte-for-byte the same data as the stock driver.
Opt-in per env: -D EINK_FAST_WRITES=1, and EINK_SPI_HZ (default 4 MHz,
8 MHz here, SPIM2's ceiling).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KxAZVAbhhYr18Dd4guTQM1
- ui-lvgl: navTextEntry()/TextEntry replace eight copies of popup + field +
keyboard wiring; one byte-capped insert callback instead of two.
- ui-new: drawHeadedBubble()/BubbleMark replace the duplicated DM and channel
bubble drawing in MessagesScreen.
- GpsSky::halt(): stopping the receiver ends an unfinished search but keeps
the first-fix time; keyboard cursor blink restarts after a key.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VeBrxKyHHxWsmttVkUdBrj
- GPS: stop() resets the sky state so the 'Searching' timer does not run
while the receiver is off; the status page shows it only when GPS is on.
- L2 keyboard: put the FOCUSED state back after typing a key or picking a
variant, so the cursor is drawn again after characters like a-ogonek.
- Quick panel: drop the full-screen dim (a blend per frame), so the screen
below stays interactive; tiles are one sideways-scrolling, snapping row;
the brightness slider no longer animates and only updates the backlight
when the value changes.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VeBrxKyHHxWsmttVkUdBrj
The L1 e-ink solo build splits in two, one per screen shape, so each
screen can be laid out for a single shape. Settings > Rotation now only
turns the screen upside down, and the joystick turns with it.
Landscape writes in misc-fixed 8x13 at the panel's own resolution
(EINK_LARGE_FONT); portrait keeps the 6x9. All drawing moves into
EinkGfxDisplay, which the panel driver and the simulator share; the
FreeSans fonts and the CRC32 library go.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- ui-core: Diagnostics::systemLines, NearbyModel::typeShort,
rptctl::currentPresetName, RoomSessions::pendingKey, settings::isSwitch,
UiCore::reconcileDMUnread.
- ui-new: the retired Recent page, fmtAlt, the battMvToPercent wrapper,
setFavouriteChannelSlot, AccordionList onHeader/visibleCount, drawBoxedIcon
and seven unused icons; info::row folded into info::valueRow.
- ui-lvgl: onOpenNav, TileDownloader::withTrails; the trail header size is
one constant (TRAIL_HDR) for live and saved trails.
- Home quick panels save prefs only when a value changed.
- Comments brought in line with the code (status bar reserve, settings rows,
page order, SH1106 frame skip, UiCore header).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DisplayDriver gains slideBegin()/slideCompose(): the driver keeps the frame
on screen and, in the next one, moves it out while the new rows follow.
SH1106 does it on its 1 KB buffer (+1 KB RAM for the copy), the sim on its
canvas; drivers that can't read their frame back keep cutting over (e-ink).
Home slides the content below the page-icon row in 220 ms with an ease-out
(the whole screen to and from the header-less clock), and the current-page
pill glides between icons.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Splash: the L2 wordmark and lettering, shared through ui-core/Lettering.h,
rise in with loading dots under the version line (static frame on e-ink).
- Home: the header shows the time once the clock is set; icon pages carry
only their title. One 32 px icon family (3 px strokes) bobs 2 px over a
dithered blob shadow that shrinks as it rises; the unread badge rides on
the Messages bubble; Bluetooth off is the icon at half density.
- Soft corners on every selection, the page pill, tabs and popups;
dotted scrollbar track with a rounded thumb; the tab bar lays out whole
tabs and shows edge arrows only in free space.
- DM bubbles drop the sender row and wrap to a second line, with the
delivery mark and age closing it.
- Keyboards name what is typed (channel, contact, Password, Scope...)
and show characters left near the limit; PIN entry has its own preview.
- Loading dots for waits (login, scan, GPS) and signal bars in place of
raw RSSI in Nearby.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DisplayDriver gets drawSoftRect / fillSoftRect / drawPanel (the unread badge's
corner knockout; a panel also clears a 1 px moat so lines below stop short of
its border). Used by alerts, popups, the accent picker, the lock hint, chat
bubbles and the send button.
Fixes: no "Msg:" toast for a message landing in the open thread; Nodes scan
text fits the screen; the PIN pad skips the blanks beside 0; Auto-Advert is a
plain value row; the Admin picker says what it needs when empty; the GPS page
shows "-" instead of 999 before a first fix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each frame is hashed per 32-column block of each 8-px page and only changed
blocks go over I2C. A full frame was 1 KB / 31 ms on every change; a moved
selection, a scrolling name or a clock tick is now 1-8 ms. 128 B of hashes,
no frame copy; full sends stay for wake/clear and SPI panels.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- CPU at 80 MHz with the screen on after 2 s without touch or animation;
a touch restores the full clock before its frame is drawn
- setText(): periodic refreshes no longer redraw unchanged labels
(status bar, Home, lock, Diag, GPS, compass, clock, map, nav, nearby)
- map markers kept across the 3 s rebuild when unchanged
- app connected: the loop blocks until a BLE frame arrives
(serialFrameArrived / hasPendingFrames) instead of polling every 2 ms
- PSRAM: boot frame sprite and the HTTP response buffer released
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Working plans (docs/development, docs/design redesign notes) removed; the
two developer guides moved to docs/developer/. README cut to devices,
flashing, links and building (Wio Tracker L2 added); its feature list and
doc tables moved to docs/solo_features/README.md. RELEASE.md describes the
Solo release flow. Upstream's GitHub Pages workflow (docs.meshcore.io)
dropped, CLAUDE.md untracked, Buy Me a Coffee back in FUNDING.yml.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two views, swapped with left / right: a polar sky plot (filled = used in the
fix, hollow = tracked, a pixel = no signal) beside fix, used / in view, HDOP,
PDOP and time to first fix; and a signal bar per satellite (used first, a
dotted 30 dB-Hz line) under per-constellation counts. The GPS stays awake
while it's open. Data from GpsSky (-D GPS_SKYVIEW, now on the L1 solo build).
The simulator's made-up sky moves from the L2's GpsScreen into
helpers/sensors/GpsSkySim.h, so both sims show it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
helpers/esp32/InternalFS.h picks the filesystem: -D ESP32_LITTLEFS mounts
LittleFS on the same "spiffs" partition, otherwise SPIFFS as before. Only the
L2 LVGL envs set it; every other ESP32 board is unchanged. A settings save
drops from ~0.5 s to ~80 ms and the files survive power loss mid-write.
No migration: the first boot formats the partition (fresh identity).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
While nothing is connected, every 5 s: if advertising has stopped (a host
reset ends it with no disconnect to restart it), start it again. A link up
for 60 s without pairing is dropped, so advertising comes back.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- BLE (ESP32): fast advertising (20-30 ms) for 30 s, then 152.5 ms, as the
nRF52 interface; the library default kept 30-60 ms forever
- Live map tiles: the WiFi drops after a minute with nothing missing
- ES8311 to standby while the I2S clocks are off, set up again before the
amp comes on
- Buzzer: no channel (DMA out of memory) waits out the samples instead of
spinning above the UI loop
- OTA: the image header check waits for its 16 bytes
- Comments: P10 is just the LED; the LCD SPI clock is 40 MHz on the wire
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The speaker's I2S channel was only disabled when idle, which leaves MCLK
running. Its fractional divider (160 MHz / 39 1/16 = 4.096 MHz) makes a
comb of lines every 256 kHz: one at 869.632 MHz, in the EU narrow mesh
channel, lifted the LoRa noise floor from ~-102 to ~-85 dBm, and others
land near GPS L1 and BeiDou B1. The channel is now deleted and its pins
held low after each sound, and made again for the next one. Measured:
noise floor -102 dBm, outdoor first fix 87 s (was 620 s and more).
Also: TCXO at 3.0 V (the L2's part, as Meshtastic has it); GLONASS
enabled on the L76K after each power-up; begin() on a running receiver
no longer resets it (a reset threw away the acquisition in progress).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
JakubandClaude Opus 5.5
ripplebizandGitHub
agessaman
pmowandClaude Opus 5.5
taco
Huw DuddyandGitHub
Liam CottleandGitHub