* macOS: enable CH341 LoRa-hardware path — fix serial truncation, document setup
Verified on Apple Silicon with a CH341A USB-SPI bridge (VID 0x1A86,
PID 0x5512) wired to an SX1262 (Meshstick variant) that the existing
`pine64/libch341-spi-userspace` lib_dep works on macOS as-is — Apple's
bundled CH34x driver only matches the CH340 *UART* variant
(PID 0x7523), so the CH341A's interface 0 is left unclaimed and
libusb opens / configures / claims it directly via IOUSBHostInterface.
End-to-end test: meshtasticd boots, libusb claim succeeds, SX1262 init
returns 0, TCP API serves the meshtastic CLI's --info / --sendtext flow.
Two changes:
1. **`PortduinoGlue.cpp:497`**: pass `sizeof(serial)` (= 9) instead of
the literal `8` to `Ch341Hal::getSerialString()`. The function in
`USBHal.h:61-68` treats `len` as buffer size and reserves one slot
for the null terminator (`bytesCopied = (len - 1) < 8 ? (len - 1) : 8`),
so passing 8 produced a 7-char serial — which then broke the
`strlen(serial) == 8` check at line 502, skipping the auto-MAC
derivation from serial + product string. On Linux this was masked
by the BlueZ HCI MAC fallback in `getMacAddr()` at lines 139-157,
but on macOS that fallback is `__linux__`-guarded so the serial path
is mandatory and the truncation left `mac_address` empty, causing
the daemon to exit with `*** Blank MAC Address not allowed!`.
2. **`variants/native/portduino/platformio.ini`**: expand the
`[env:native-macos]` comment block with a "Real LoRa hardware on
macOS" section. Documents:
- Why no upstream library change is needed (Apple kext targets
CH340/UART, not CH341A/SPI; libusb's `#ifdef __linux__` skip is
correct for macOS in this case).
- How to point `meshtasticd` at an existing platform-agnostic
`bin/config.d/lora-*.yaml` for CH341 hardware.
- The auto-MAC-derivation contract (now working with this fix).
- `ioreg` and `LIBUSB_DEBUG=4` diagnostic recipes for the failure
mode where a third-party WCH `CH34xVCPDriver` *would* claim
interface 0 (`kmutil unload -b <bundleID>` workaround).
No upstream library forks, no PR chain, no additional lib_deps —
the existing `pine64/libch341-spi-userspace` + libusb-1.0 stack does
the right thing on macOS already.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Apply suggestion from @Copilot
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Router::handleReceived stores its allocCopy of the encrypted packet in the
class member p_encrypted. callModules() invokes module replies that re-enter
the router via MeshService::sendToMesh -> Router::sendLocal, which on a
broadcast reply recursively calls handleReceived. The inner call overwrites
the outer's p_encrypted without releasing it; on the way out it nulls the
member, the outer release(p_encrypted) now releases nullptr, and the original
allocation is permanently leaked. ~381 B per recursion.
Promote p_encrypted to a function-local so each invocation owns its own copy
for its full lifetime. The MQTT-publish null check at the call site (added by
PR #9136 as a workaround for this bug) stays in place because allocCopy can
still legitimately return nullptr on packetPool exhaustion.
Copilot's review of PR #8999 (the original introduction) flagged this exact
pattern at merge time:
"Storing p_encrypted as a class member can cause issues with recursive or
concurrent calls to handleReceived() since each call would overwrite the
previous packet pointer."
The historical reason for the member (S&F needing to retain the encrypted
copy across calls) was satisfied differently by PR #9799 (ServerAPI converted
to std::unique_ptr + cleanup on connection close), so the member is no longer
load-bearing.
Reproduces issues #9632 / #10101 / #8729 (heap leak when MeshMonitor
connected; TCP drops on Station G2 / LILYGO ServerAPI dump abort).
Hardware A/B on Station G2 under sustained TCP-API retry storm (open :4403,
request config, disconnect mid-stream, repeat at ~0.6/s) - 9 min run:
| Build | heapFree drift | rebootCount delta |
| this patch | -1.5 KB (noise)| 0 |
| stock 2.7.13 | -73 KB (8.1KB/min) | +1 (OOM crash) |
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
The T-Deck-Pro V1.1 PR (#9378) had a backwards merge (e7b66281) that pulled
236 commits of master INTO a 1-commit fork branch, then a second merge
(8fd0a7f2) duplicated the entire T_DECK/T_DECK_PRO/T_LORA_PAGER/HACKADAY
peripheral init block in setup(). Also removes a stray
digitalWrite(BLE_LED, LED_STATE_OFF) in the HACKADAY section that was an
artifact of the evil merge at e7b66281.
* niche: add InkHUD port for LilyGo T5-E-Paper-S3-Pro (ED047TC1)
Add a NicheGraphics EInk driver adapter for the 4.7" ED047TC1 parallel
e-paper display used on the T5-E-Paper-S3-Pro (H752-01). The driver
wraps FastEPD and handles the polarity difference between InkHUD's
buffer format (0xFF = white) and FastEPD's (0x00 = white).
Rewrite variants/esp32s3/t5s3_epaper/nicheGraphics.h which was an
incomplete copy of the Heltec VM-E290 setup referencing undefined SPI
pin macros and a non-existent BUTTON_PIN_SECONDARY. The board uses a
parallel display, not the small SPI DEPG0290BNS800 that was referenced.
* fix: guard inputBroker null dereference in TouchScreenImpl1::init()
When MESHTASTIC_EXCLUDE_INPUTBROKER is defined (e.g. InkHUD builds),
inputBroker is nullptr. Calling inputBroker->registerSource() in that
state caused a LoadProhibited panic on any board that has both
HAS_TOUCHSCREEN=1 and the InputBroker excluded.
Add a null check before registerSource() to prevent the crash.
* niche: fix display rotation for T5-E-Paper-S3-Pro InkHUD port
Set rotation=3 (270° CW) in nicheGraphics.h to correct for FastEPD
scanning the ED047TC1 panel in portrait orientation, resulting in
correct landscape display output.
* fix: update buffer format descriptions and remove polarity inversion for InkHUD and FastEPD
* fix: update ED047TC1 driver to handle inactive pixel borders and adjust safe-area dimensions
* fix: comment out ruler diagnostic for E-Ink driver
* feat: implement TouchInkHUDBridge for direct touch event handling in InkHUD
* niche: add FreeSans 18pt/24pt Win1253 (Greek) fonts for larger InkHUD displays
Add Win1253-encoded FreeSans 18pt and 24pt font headers to support Greek
script on larger InkHUD screens (e.g., the 4.7" ED047TC1 at ~234 DPI).
Register FREESANS_24PT_WIN1253 and FREESANS_18PT_WIN1253 macros in AppletFont.h.
Set fontLarge=24pt, fontMedium=18pt, fontSmall=12pt in nicheGraphics.h for the
T5-E-Paper-S3-Pro.
* feat(ed047tc1): use true partial update for FAST refresh
Replace fullUpdate(CLEAR_FAST) with partialUpdate() for FAST display
updates. FastEPD's partialUpdate() diffs pCurrent against pPrevious
and only applies the update waveform to rows that have changed, leaving
unchanged rows with a neutral signal.
This reduces visible flicker on routine updates (new messages, position
changes) — only the affected region of the screen refreshes. Full-screen
CLEAR_SLOW updates are preserved for periodic ghosting cleanup, driven
by InkHUD's setDisplayResilience() ratio.
* feat(t5s3-epaper): enable frontlight via LatchingBacklight
Wire up BOARD_BL_EN (GPIO11) to InkHUD's LatchingBacklight driver.
Enable the backlight menu item so users can toggle "Keep Backlight On"
via Settings. The backlight turns on automatically when the menu opens
and off when it closes.
* Fix RTC chip (PCF8563 not PCF85063) and GT911 I2C address collision
- variant.h used PCF85063_RTC but the board has a PCF8563. The difference
is the RAM register: PCF85063 has 1 byte of RAM; PCF8563 does not. The
PCF85063 driver was trying to write this register on init, failing every
time, and setDateTime writes were silently discarded — RTC time was
never persisted across reboots. Switch to PCF8563_RTC/PCF8563_INT.
Before:
[E][SensorPCF85063.hpp:375] initImpl(): Failed to write to RAM memory
register. Maybe this chip is pcf8563.
Read RTC time from PCF85063 getDateTime as 2026-04-05 00:00:23
PCF85063 setDateTime 2026-04-05 18:40:59
Read RTC time from PCF85063 getDateTime as 2026-04-05 00:00:19 ← lost
After:
PCF8563 found at address 0x51
Read RTC time from PCF8563 getDateTime as 2026-04-05 18:58:37 ← persisted
PCF8563 setDateTime 2026-04-05 18:58:44
Read RTC time from PCF8563 getDateTime as 2026-04-05 18:58:44 ← round-trips
- GT911 touch was initialized with GT911_SLAVE_ADDRESS_L (0x5D), which
collides with the SFA30 air quality sensor also at 0x5D on the same
I2C bus. Switch to GT911_SLAVE_ADDRESS_H (0x14): the library drives
INT high during reset to program the GT911 to address 0x14,
eliminating the address conflict.
Before:
SFA30 found at address 0x5d
[I][TouchDrvGT911.hpp:568] initImpl(): Try using 0x5D as the device address
After:
SFA30 found at address 0x5d
[I][TouchDrvGT911.hpp:544] initImpl(): Try using 0x14 as the device address
* t5s3_epaper: fix GT911 ghost-SFA30 via early I2C address latch
Investigation findings
----------------------
Boot logs showed "SFA30 found at address 0x5d" on every cold power-on,
and AirQualityTelemetry was registering an SFA30 sensor. However, every
readMeasuredValues() call returned error 268 (0x010C = Sensirion
WriteError | I2cAddressNack), meaning the I2C write to 0x5D was being
NACK'd — inconsistent with a real SFA30.
Root cause: the GT911 touch controller latches its I2C address from the
INT pin level at reset time (GT911 datasheet §4.3). GPIO3 (INT) defaults
LOW on ESP32-S3 cold boot → GT911 always powers up at 0x5D
(SLAVE_ADDRESS_L). The I2C scanner runs before lateInitVariant() had a
chance to reprogram the chip.
The scanner's SFA30 detection (ScanI2CTwoWire.cpp) sends the 2-byte
command 0xD060 to 0x5D and requests 48 bytes back. GT911 ACKs the
write (treating it as a register address) and returns 48 bytes of
register data, passing the length check — a false-positive SFA30
detection.
Confirmed via second cold-boot log: after the previous commit moved GT911
to 0x14 in lateInitVariant(), address 0x5D *still* appeared in the scan
because the scanner runs first. The board has no physical SFA30 fitted.
Fix
---
Add the GT911 address-latch reset sequence to earlyInitVariant(), before
Wire is initialised and before the I2C scan runs. Per the datasheet:
drive RST LOW, drive INT HIGH (selects address 0x14 / SLAVE_ADDRESS_H),
hold >100 µs, release RST, wait >5 ms startup. GPIO-only, no Wire
dependency. lateInitVariant() then repeats this sequence internally via
touch.begin(); the double-reset is harmless.
Verified in boot log:
Before: "SFA30 found at address 0x5d", 5 I2C devices, NACK errors
After: no SFA30 entry, 4 I2C devices (TCA9535/PCF8563/BQ27220/BQ25896),
GT911 found at 0x14 and touch initialised successfully,
AirQualityTelemetry registers no sensors (correct — no SFA30 present)
* t5s3_epaper: add variant_shutdown() for touch sleep and backlight off
Put GT911 into low-power standby (command 0x05) and drive BOARD_BL_EN
LOW before deep sleep to avoid unnecessary current draw.
* t5s3_epaper: fix touch gesture routing and coordinate mapping
readTouch() now transforms raw GT911 axes to visual-frame coordinates
based on the current display rotation (rotation=3 is the hardware
identity). This ensures TouchScreenBase detects swipe direction
correctly regardless of which rotation the user has selected.
TouchInkHUDBridge dynamically sets joystick.alignment = (4-rotation)%4
on each touch event so that (rotation+alignment)%4==0 always, keeping
nav calls pass-through without remapping.
nicheGraphics.h now calls loadSettings() first so that rotation is
persisted across reboots. rotation=3 and other first-boot defaults are
only applied when tips.firstBoot is set. alignment is recomputed from
the loaded rotation on every boot.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* t5s3_epaper: fix GT911 sleep timing via notifyDeepSleep observer
touch.sleep() was called from variant_shutdown(), which runs inside
cpuDeepSleep() — after Wire.end() had already torn down the I2C bus in
doDeepSleep(). This caused Wire NULL TX buffer errors and left the GT911
awake during deep sleep.
Register a CallbackObserver on notifyDeepSleep, which fires before
Wire.end(), so the I2C command reaches the chip while the bus is live.
Pattern matches LatchingBacklight and other NicheGraphics components.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* t5s3_epaper: fix touch nav and applet defaults in nicheGraphics
Enable joystick mode post-begin so menu scroll and swipe-up/down
gestures are not silently dropped by the joystick.enabled gate in
Events.cpp. Activate DMs and Channel 0/1 applets with correct
autoshow defaults matching the mini-epaper-s3 reference pattern.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Update nicheGraphics.h
* t5s3_epaper: fix ED047TC1 driver docs and remove spurious beginPolling
Addressing PR review comments:
Remove beginPolling(1, 0) after the blocking FastEPD update — it
incorrectly set updateRunning=true for one loop cycle after the
hardware was already done, causing busy() to briefly return true.
Since isUpdateDone() always returns true, no polling is needed.
Also fix stale comments: safe-area buffer size was 944×532, now
944×523; V_OFFSET_ROWS didn't exist, replaced with the actual
V_OFFSET_TOP=9 / V_OFFSET_BOTTOM=8 constant names.
* t5s3_epaper: clean up applet addition formatting in setupNicheGraphics
* t5s3_epaper: guard ED047TC1.cpp against non-T5S3 InkHUD builds
The InkHUD base config pulls in all of src/graphics/niche/ so every
InkHUD device compiled ED047TC1.cpp, triggering the #error on line 48
for boards that define neither T5_S3_EPAPER_PRO_V1 nor V2.
Wrap the file body with #ifdef T5_S3_EPAPER_PRO so it is only compiled
for T5S3 targets. The #error is preserved inside the guard to catch
future hardware revisions that forget to update the driver.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: HarukiToreda <116696711+HarukiToreda@users.noreply.github.com>
* Add authoring guide for native unit tests in README.md
* Enhance documentation for agent tooling and native unit tests in README and related files
---------
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
* Add ESP32 Power Management lessons learned document
Documents our experimentation with ESP-IDF DFS and why it doesn't
work well for Meshtastic (RTOS locks, BLE locks, USB issues).
Proposes simpler alternative: manual setCpuFrequencyMhz() control
with explicit triggers for when to go fast vs slow.
* docs(prompts): fix markdown fence language tags
* docs: remove ESP32 power management notes
Add a meshtasticd config for the Luckfox Pico Max with the Waveshare Pico LoRa SX1262 TCXO HAT.
Tested on hardware with successful SX1262 init, broadcast, and direct messaging.
The CALIBRATE_ALL (0x7F) command inside resetAGC() clears bit 0 of the
undocumented 0x8B5 register. That bit is set once in init() by #9571 and
#9777 to improve SX1262 RX sensitivity, and the AGC-reset path was not
re-applying it. Result: every SX1262 node silently loses the RX
sensitivity patch ~60s after boot and never recovers until reboot.
Empirically confirmed on Heltec Mesh Node T114 (nRF52840 + SX1262):
- Post-calibration read of 0x8B5 = 0x04 (bit 0 cleared)
- After re-apply: 0x05 (bit 0 set)
Reproducible every AGC_RESET_INTERVAL_MS tick.
Fix re-applies the register bit alongside the existing post-calibration
re-applies (setDio2AsRfSwitch, setRxBoostedGainMode).
