Board manifest for esp32-c6-devkitm-1 defaults build.flash_mode to qio,
which this module's flash chip does not support -- causes a boot
crash-loop (repeated USB-Serial-JTAG reconnects) on real hardware.
Override with board_build.flash_mode = dio in the common M5Stack_Unit_C6L
section so it applies to all envs (ble/usb/repeater/room_server).
CustomLR1110::startReceive() passed RADIOLIB_LR11X0_IRQ_PREAMBLE_DETECTED
(1<<4 = 16) as RadioLib's first argument, which is the RX *timeout*, not an
IRQ mask. At the LR11x0's 30.52us tick that armed the receiver for ~488us, so
it dropped out of RX before any packet could arrive and the node received
nothing at all -- while transmitting normally.
Symptoms on a SenseCAP T1000-E: tx_air_secs rising, rx_air_secs stuck at 0,
recv_errors 0, and the noise floor pinned at the -120 clamp because
getCurrentRSSI() never sampled a live receiver.
Pass RADIOLIB_LR11X0_RX_TIMEOUT_INF (continuous RX), which is what
LR11x0::startReceive() itself uses, keeping the PREAMBLE_DETECTED flag in the
reported IRQ flags as intended.
Introduced in ea5d7c8b ("LR1110: add PREAMBLE_DETECTED to reported irq flags").
Verified on two T1000-E units: with only the repeater fixed it began receiving
(last_rssi -29, SNR 17.0) while the unfixed companion stayed deaf; fixing both
brought up the link in each direction.
The misc-fixed font (full Latin/Greek/Cyrillic) had only ever reached
SH1106Display and the e-ink driver. GAT562 30S -- like 24 other variants
-- uses SSD1306Display, which stayed on the built-in 5x7 font, so every
keyboard alphabet beyond ASCII and every accented contact name was
transliterated (Łódź -> Lodz) or drawn as a filled block.
Extract the renderer into MiscFixedRenderer.h rather than copying it:
Adafruit_SH110X and Adafruit_SSD1306 both derive from Adafruit_GFX and
glyph drawing is pure pixel plotting, so one implementation serves both
drivers (-79 lines from SH1106Display.cpp, same flash usage as before).
The header is .cpp-only by contract -- the font tables are static const,
so including it from a driver header would land a copy in every
translation unit.
On SSD1306 the path is opt-in behind OLED_MISC_FIXED_FONT, set only in
the two GAT562 30S solo envs: the font costs ~14 KB of flash and the
other 24 variants' repeater/companion builds have no keyboard to type
those alphabets on. Without the flag that driver is unchanged, verified
by building GAT562_30S_Mesh_Kit_repeater.
Drive-by consistency fix from the extraction: a newline in print() now
advances the cursor by yAdvance * text_size instead of a single yAdvance,
which was half a row short at setTextSize(2). The e-ink driver always
scaled it. No effect at size 1, which is everywhere the UI prints '\n'.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The screen faded as the device was used: text went grey a few updates
after it was drawn, while whatever had just changed stayed crisp. Only a
full refresh brought it back, and the next few updates ate it again.
A partial update is differential -- the controller drives only the pixels
that differ between its "current" (0x24) and "previous" (0x26) RAM and
leaves the rest to hold their own charge, which this panel doesn't do
well. Priming the previous-image RAM with the inverse of the incoming
frame makes every pixel a difference, so all of them get driven to their
target, in both directions.
The inverse matters: priming with flat white makes only white->black a
difference, so ink is re-driven but never erased and every screen ever
shown accumulates as a ghost. Confirmed on hardware, both ways round.
Costs one extra full-screen RAM write (a few ms of SPI). The refresh
itself is unchanged -- the waveform clocks the whole panel regardless of
how many pixels it drives -- so nothing got slower. Clearing ghosts is
still the periodic full refresh's job (Settings > Full rfsh), which can
now stay off.
The helper needs GxEPD2_BW's private _buffer/_page_height, so it goes in
the patched copy of the header, which every e-ink build now includes
unconditionally rather than only screenshot-enabled ones. All e-ink
variants pin GxEPD2 1.6.2, the version that copy tracks.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A message containing a line break drew two words on top of each other in
the fullscreen reader. wrapLines() treated '\n' as an ordinary character:
it measured it via getCodepointWidth() -- which reports a full 6px cell
for it, since 0x0A sits below the font's first glyph -- and copied it into
the wrapped line. Both display drivers' print() then acts on '\n' by
resetting the cursor to x=0 and stepping down one row, so the tail of that
line was drawn straight over the following one.
wrapLines() now ends the line at '\n'/'\r' (CRLF counts as one break) and
consumes the byte rather than emitting it, preserving blank lines the
sender typed while still skipping degenerate empty wrap segments so the
loop can't stall. This covers the fullscreen view and the history list's
portrait bubbles, which share the function.
drawTextEllipsized() folds newlines into spaces for the same reason: it
draws one line clipped to max_width, and the compact one-line message
previews in the landscape list feed it raw message bodies. A space keeps
the words apart and measures the same, so the ellipsis maths is unchanged;
for names and labels it's a no-op.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
If Bluefruit.begin fails, BLE OTA mode won't actually start and the board might require a reboot to reattempt.
Fixes:
- Bluefruit.begin returns false in NRF52Board.startOTAUpdate if OTA mode fails to start. User is notified of the fault through existing error message in CommonCLI and can reattempt "start ota" command.
The text preview was the last part of the keyboard still working in bytes
rather than codepoints. cpl is how many characters physically fit on a
line, so dividing byte offsets by it counted every 2-byte Cyrillic/Greek/
accented character as two: lines held half the text they had room for, and
a break could land inside a codepoint. Both display drivers are
permanently single-font, so translateUTF8ToBlocks() passes UTF-8 straight
through -- the truncated sequence reached print() and drew as garbage on
both sides of the break. Line boundaries now walk the buffer with the same
kbUtf8*() helpers insertion/backspace/T9 already use, and the per-line
buffers are sized for a full line of 2-byte characters.
Caps-lock also gets an underline on the shift key: it sets caps too, so
the highlight alone made a one-shot Shift and a held lock indistinguishable
despite capitalising one letter vs. every following one.
Drops UITask::applyFont() -- setSingleFont() is a no-op on both drivers
since they were pinned to misc-fixed, so it did nothing, and use_lemon_font
has had no Settings row for a while. The pref itself stays: it's part of
the on-disk layout. Retires the matching stale rationale on scriptHint().
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Compact mode (Settings > Keyboard's "Ext. KB" row) is meant to guarantee
operation with no joystick at all, but it was still half-tied to the
on-screen grid it hides:
- arrows now move the text cursor directly instead of a grid selection
nobody can see, and Tab opens the placeholder picker directly instead of
the row/col-dependent Hold-Enter dispatch
- plain Enter submits the field (there's no grid cell to have deliberately
landed on), same as Fn+Enter
- Fn+letter's accent popup no longer gates on the grid's script/T9
settings -- CardKB always types plain Latin regardless of them, so the
gate only made the gesture silently stop working
- the whole status line is gone: nothing it showed (script, T9-vs-ABC,
caps) is actionable from an external keyboard. The freed height goes to
message-preview lines, floored at the smallest grid's footprint so
cursor mode's own hint block still fits
- the accent popup gets a fixed slot instead of anchoring on a `row` that
is never deliberately navigated to in this mode
Also fixes a text-corrupting invariant break: moveCursorDirect() and
openPlaceholders() move the cursor without finalizing a pending T9
multi-tap cycle, so a later tap on the same cell within the timeout
overwrote an unrelated character. Every other cursor-moving path already
cleared it.
Fn+Tab is dropped as a separate shortcut -- plain Tab already covered
every case it did. Fn+Enter no longer reads as a dead key in cursor mode.
Direct typing moves into insertTyped(), one translation point documenting
what a future relabelled-keycap layout (Cyrillic/Greek) would need.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
- DM/room unread badges could claim messages the ring no longer held
(same class as the channel fix in 6470afaf, not covered by it):
getDMUnread()/getDMUnreadTotal() now clamp to dmHistCountForContact(),
and a new reconcileDMUnread() (called once per loop()) frees any
_dm_unread_table slot whose ring occupancy has dropped to 0, so a
17th sender isn't starved by stale entries. onContactRemoved() now
also clears _dm_unread_table -- the one per-contact table it was
missing.
- Shift didn't capitalise ł/ń/ź/ż (+ĺ/ľ/ň/ž): the Latin Extended-A
case-pairing rule assumed a single parity for the whole block, but it
flips around the unpaired codepoints ĸ/ʼn/Ÿ. Fixed with four
sub-ranges, verified exhaustively over U+0100-U+017F.
- Triple-click could still toggle the buzzer while locked on
PIN_USER_BTN/PIN_USER_BTN_ANA boards (joystick path already guarded
this).
- millis() wraparound: 4 absolute comparisons in UITask.cpp (battery
poll, auto-off, lock-wake, backlight) converted to the existing
(int32_t)(millis()-deadline)>=0 idiom; MyMeshBot.h's DM-throttle
eviction now picks the oldest slot by elapsed time instead of raw
t_ms, which picked the wrong slot right after a rollover.
- Long-press bypassed checkDisplayOn() on all 5 call sites -- neither
woke the display nor extended auto-off, and could deliver
KEY_CONTEXT_MENU to the invisible screen. Moved the gate inside
handleLongPress() itself instead of patching each site.
- CardKB's backspace/printable-insert branches didn't reset t9_cell,
so typing right after a T9 cycle tap could get silently overwritten
by a same-cell re-tap within the T9 timeout.
- buildContactList()'s counts[MAX_CONTACTS] was a 1400 B int array on
the 4 KB loop() stack; values are bounded by DM_HIST_MAX (32), so
now uint8_t.
- ACK table treated ack==0 as a wildcard: isAckPending(0) matched any
free slot, and processAck() with an all-zero ACK matched the first
free slot and returned its stale contact pointer. Both now skip/reject
ack==0, and the matched slot's contact pointer is cleared alongside
its ack hash.
- ensurePageOrderInit() could write one byte past page_order[13] when
migrating a saved order with all 13 slots full and CLOCK last --
guarded on insert_at < PAGE_ORDER_LEN.
Two findings from the same review were resolved as no-op decisions,
not code changes: !buzz over DM ignoring quiet hours is intentional
(the pull exemption is meant to cover the buzzer), and the offline
queue's full-queue drop-newest behaviour is upstream code, left alone.
Build-verified green on WioTrackerL1_companion_solo_dual (RAM 71.1%,
Flash 66.6%) and WioTrackerL1Eink_companion_solo_dual (RAM 73.0%,
Flash 67.9%).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The per-channel unread counter was independent of the ring's actual
contents, so the two drifted apart:
- Opening a channel whose entries had been evicted left the badge
claiming messages the list could no longer show. The viewing-session
bookkeeping computes the count from an _unread_at_entry snapshot, and
with an empty list _hist_visible is 0, so entering only knocked the
count down by one instead of clearing it (badge "7", empty list, then
"6").
- Eviction from a full ring decremented the counter for any dropped
entry, including already-read ones, undercounting the newer unread
messages the counter actually refers to.
chUnread()/getTotalChannelUnread() now clamp to the channel's ring
occupancy, so the badge can never promise more than the history holds
whatever the raw counter says, and eviction only decrements when the
entry being dropped was itself unread.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
mkdocs will only consider the first H1 (if any) and subheaders under it for the table of contents
this increases the header levels of everything below "important concepts" by 1 so that the table of contents correctly resolves them
Post-review cleanups, no behaviour change:
- botScanCommands() parsed the command name and its two args with three
near-identical read-token loops; extracted a single readToken() lambda.
- Fn+Esc lock branch turned the display on twice (the unlock arm repeated
what the branch head already did); dropped the redundant call.
- setGpioMode()'s comment said "Cycle" (cycling lives in GpioScreen); now
describes what it actually does — set a specific mode + persist.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Bot Actions (!buzz/!gps/!advert/!gpio1-4) ran their side effect
immediately during botScanCommands(), before quiet-hours/cooldown/
per-contact throttle were checked -- those gates only suppressed the
reply text, not the actual buzz/GPS toggle/advert/pin write. botCommandReply()
now only records what was requested; applyPendingBotActions() runs the
deferred effects once a wrapper's throttle checks pass and the ack sent,
mirroring the existing _locfix_requested pattern. resetPendingBotActions()
clears everything on every throttled/aborted path.
- CardKB's Fn+<letter> accent-popup shortcut bypassed the locked-input gate
(it called into KeyboardWidget directly instead of through the
enqueueKey()/dequeue path every other key uses, so it wasn't discarded
while _locked). Now checks _locked itself.
- Since a locked device now correctly ignores CardKB entirely, Fn+Esc
(single press) is added as CardKB's own lock/unlock gesture -- otherwise
a CardKB-only setup had no way to unlock. Esc rather than the adjacent
Fn+Backspace, to avoid an accidental press.
- botScanCommands() now parses up to two arguments per command instead of
one. Used by "!gps fix [seconds]" to override the default 90s timeout
(clamped 15-300s) for a poor sky view where 90s isn't always enough to
reach isLocFixReady()'s HDOP/satellite bar.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Satellite count alone is a poor proxy for fix quality -- few satellites
in good geometry can beat many in poor geometry. LocationProvider now
exposes getHDOP() (default -1 = unsupported); MicroNMEA implements it.
isLocFixReady() prefers HDOP <= 2.0 when available, falling back to the
old >=8 satellite threshold for providers that don't report it (e.g.
RAK12500/u-blox).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Same fix as hotfix/admin-login-timeout (96b44460). AdminScreen's only
guard was "_phase == LOGIN" (true for any node sat at the login
screen), not that the reply actually named _target. Now also checks
pub_key against _target.id.pub_key.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
AdminScreen::onRoomLoginResult()'s only guard was "_phase == LOGIN" --
true for *any* node currently sat at the login screen, not specifically
_target. Combined with UITask::onRoomLoginResult()'s current-screen
dispatch (not requester-based), a slow reply for an earlier login
attempt (this screen's own previous target, or even MessagesScreen's)
arriving while the user has since opened Admin on a different,
password-less node -- still parked at the blank LOGIN keyboard, so
_phase == LOGIN here too -- was accepted as that new node's own login
result, flipping _admin_ok/_phase to COMMAND without ever actually
authenticating with it.
Root-caused by cancelUiPendingLogin() (previous commit): that fix
covers the "gave up, then it resolved late" path, but not "a reply for
a genuinely different pubkey arrives while merely _phase == LOGIN".
Checking pub_key against _target.id.pub_key closes that regardless of
which path let the reply through.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Scott Powell
Huw DuddyandGitHub
Liam CottleandGitHub
me
Mike Damon
JakubandClaude Opus 5
taco
entr0p1
Che177
DG1TAL
Talia