Greying every name that two keys ever used was too strict: a reflash with an
erase makes a new key, and 34 names on this mesh have one, JP's own among
them. Smog went grey this morning because of two keys silent since September.
The newest key now gets the name, and the name is grey only when another key
used it within name_overlap_days (a new Display preference, "Shared names",
default 7) of it: two live claimants. A contact that was never heard never
wins. On this machine's real data, LOG.IC 3R goes to its active key 47368a7c,
Smog to fa331c83, and 8 of 883 names stay grey.
The stored names come with their times from the one scan node_names() already
ran (node_names_with_times), so opening a chat costs no second scan.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A contact's page has enough action rows that, on a 24-row terminal, they left
the preview under its five-row floor, and it stepped aside everywhere. It now
keeps its floor and the page scrolls. Only a short frame (the Cardputer's
11-row viewport) leaves it out.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
A name is a label any key can claim, and a channel message carries only its
sender's name. When two nodes use one name, the resolver used to let the last
contact win, which handed one node's hue, map, and page to a message that
nothing tied to it: a second "LOG.IC 3R" with a new key took the first one's
green and location. Such a name now resolves to None, like any unplaced name.
A contact's full key and the stored id of the same node still count as one.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
In a channel, a DM, or a room, ^O on a selected message opens the node page
of its author, the same page the Contacts list opens: our own for a message
of ours, the contact the author's key finds otherwise (ContactPages, the twin
of the map's Locator). It opens to look, without the management verbs, which
belong to the list a contact is in. The hint names ^O only where it acts; the
handheld chip is Author, on Shift+F3 behind Map, since F1-F3 are full.
The room's login moves from ^O to ^G.
The Unreleased changelog now also says where the Map chip went, and covers
the shared lane and the route-end chips.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
F1-F3 hold what the three conversations share, in the hint's order: Paths,
QR, Map. The Shift bank holds each screen's own verbs: ^R on Shift+F1 (Retry
in a DM and a room, Resend in a channel) and the room's Log in on Shift+F2.
F4/F5 stay the pager with Latest/Oldest behind Shift.
Day ↑ and Day ↓ are gone from the lane; the desktop keeps the day jump on
^PgUp/^PgDn.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Messages list's preview draws its author as an origin chip (no colon: the
point says it) and its mentions as destination chips, as the transcript does.
The packet viewer's from rows close on the point and its to row opens on the
notch, around the via chain that already opens and closes on both. The live
feed's addressed subject is an origin chip and a destination chip facing each
other across one cell of page (widgets.route_ends): a seam would claim the two
are neighbours. A pair too wide for the lane cracks instead of eliding. The
courier's recipient is a destination chip, in the hue of its key while it
waits and muted once finished.
Arrow mode reads exactly as before everywhere. A lone hop there carries no
PATH_INK stamp, so a highlighted list row whitens it like any other name.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A message goes from its author toward the nodes it names, so the two chips are
the two ends of one route that is not drawn. name_chip takes a role (ORIGIN or
DESTINATION); arrow mode ignores it, so a degraded terminal draws exactly what it
did before. A chip in a message body never breaks across lines: render_hanging
holds each whole run together, spaces included. _sender_style lost its last
caller and is gone; its tests now read the rendered chips.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
locate moves from ^Y to ^L, and the room's login from ^L to ^O, under the
chord rule of 3675226. On the map and in the mesh walk, ^L still goes to
our node. In a chat and on a room's board, ^L opens the map on the author
of the selected message (our own message: our node); on Contacts, on the
highlighted node. map_screen.Locator holds the markers that
gather_markers returns when the screen opens and answers find() at each
paint, so the "^L map" hint atom and the Shift+F1 Map chip show only
where the key acts. A room keeps Log in on Shift+F1, so ^L has no chip
on a room's board yet.
The chat's hints for a selected message go through _fit_hint, which
drops ↑↓ and then ^P only when the line passes 72 cells; the room now
reuses them. LaneDeck.action_for returns None for a slot that has only
a Shift companion.
Right Ctrl with a letter chord still does nothing on the Canadian
Multilingual layout: JP chose not to repair that.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The viewer now opens on the stop the feed has. On the pin (^), the card
follows the stream and the feed stays pinned; on a selected packet, the
viewer holds that packet as before. Enter no longer drops the pin.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Both tests walk scripts/ with rglob, so they also read the MeshCore
checkout that build-firmware.sh makes in scripts/picocalc-lyra/
xiao-radio/_meshcore-build. Its third-party code and libraries have
contractions and no SPDX lines, and every developer who followed the
PicoCalc guide got three failures. conftest.not_ignored asks git
check-ignore which paths it ignores and removes them; a tracked file
is never ignored, and without git the tests examine everything, as
before.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
scripts/cardputer-zero/build-deb.sh runs on arm64 Debian 13 (the
Cardputer itself) and makes meshterm_<version>-<revision>_arm64.deb.
MeshTerm and all its Python packages, the spi extra included, go in
/usr/lib/meshterm, compiled to bytecode. The package depends only on
python3: the store app installs a missing dependency with apt-get but
never runs apt-get update, so a Debian dependency can fail. For the
same reason the package carries Terminus beside its OFL, because an app
from the launcher has no terminal for --fetch-fonts.
packaging/notices.py gets a site-directory mode, because the walk of
the closure asks for no extras and so misses the spi packages.
scripts/cardputer-zero/store holds the listing that czdev publish reads
and a 256-pixel icon. The screenshots are still to make.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
test_the_panel_is_written_through_its_mapping named its stand-in file fb0
and trusted that no sysfs node would answer for it. A Linux CI runner has
a real /sys/class/graphics/fb0, so Framebuffer took that device's stride,
mapped far more than the 32-byte file held, and mmap refused. It failed on
every Ubuntu job of the 0.10.3 push; Windows skips it and macOS has no
/sys. The test now stubs the sysfs read, as its comment always assumed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
MeshTerm already stores every channel packet it hears, including those
on channels the device doesn't have. A packet names its channel only by
a one-byte hash, but a public channel's key is derived from its name, so
a candidate name can be tested: derive the key and check the packet's
MAC with it, exactly as the firmware does before decoding. The guide
tries the device's channels, Public, every #name mentioned in a stored
message, and every known region. On one desktop history that names 25
public channels.
A guessed name has to match at least two messages before the guide
shows it. The check compares three bytes, and with thousands of names
against thousands of packets roughly one false match is expected. The
candidates are grouped by their hash byte, so each packet only checks
the ~1/256 that could match; the naive loop took 40 s, this takes 1 s
on 31,000 messages, in a worker thread under a busy card, and the result
is kept for the session while its inputs are unchanged.
"Channel guide…" is the first row of "Add a channel" on the Channels
page, and it stays there when every slot is full. Channels not on the
device come first, and Enter adds one to the next free slot. Channels
the device already has follow with a ✓ the highlight doesn't walk, and a
closing line counts the messages no name matched. Each row shows the
message count, the age of the last message and seven days of activity.
`channels guide` lists the same channels on the command line.
Refs: #25
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A new device ships with Public and nothing else, so changing devices
meant adding every channel again by hand: each private key pasted, each
# name typed, each send scope and mute set again. The Channels page now
has an "Export and import" section. Export writes every channel to one
TOML file, in slot order, with its key, its send scope and its mute.
Import reads such a file and adds its channels to the device.
Import never removes a channel. The channels of the file go first, in
the file's order, and whatever else the device holds follows them. A
channel with no free slot is reported, not forced in. Before anything is
written, a dialog shows the plan (what is added into which slot, what
moves) and ends on "Nothing is removed from the device." The slots are
read again in full before the plan is made, because a probe cut short
makes an occupied slot look free.
The file uses the same [[channels]] table as config backup, so a config
backup imports too. A # channel carries no secret, since the device
derives its key from the name. A private channel carries its key, so the
file is written readable by its owner only, and the export says how many
private keys it holds.
On the command line, `channels export FILE` writes the file and prints
its path, and `channels import --file FILE [--dry-run]` prints the
device's layout after the import, one row per channel with add, move,
keep or skip. `channels import INDEX URL` still imports a single link.
Refs: #25
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Trophy case drew each hop of a walk at the first position that it
found, and it looked at the heard adverts before the contacts. The trace
scoring and the name resolver do the opposite: a contact wins. Thus a hop
could get its name and its colour from one node and its place on the
drawing from another.
On real data, the one-byte hop 3d is named YUL-Cartierville (a contact
with a position). The heard history has no position for that repeater,
so the old lookup took the next heard node that starts with 3d, a node
in St-Jean, 38 km away. Seven of 43 records drew that pin in St-Jean.
The lookup is now node_geo_lookup, a module function with tests. The
contacts are first, the same as in the scoring and the resolver.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
tests/test_writing_style.py rejects an em dash, a semicolon, a
contraction, or a Latin abbreviation in a docstring, a comment, or the
prose of a document. It does not read code font, quoted text, code blocks,
URLs, or legal text between <!-- ste: off --> and <!-- ste: on -->.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A final pass over the docstrings and comments: "the own X of Y" is now
"Y's own X" (84 places), a navigation entry is a screen on the stack, our
node is our node, a paint is a paint, the quoted UX labels are in code
font, and a list of punctuation glyphs is in code font.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Each test docstring states the behaviour that the test proves, in the
simple present. No assertion message, test id, or other string literal
changed. Four statements now agree with the code: the Trace highlight
stays on the last row, a menu count, and two others.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
On the desktop and the Cardputer, a chip was the node's hue with dark
ink on it, so a node's colour moved from its name to the ground behind
it the moment a route drew it. Now the label is the hue, the colour the
name has in every list and in arrow mode, and the fill is the same hue
at 0.60 of its OKLab lightness (CHIP_LIGHTNESS): the rule the PicoCalc
draws with its dim twins, at full resolution. A hash label and an
annotation are the label colour throughout, with the lit bytes bold.
A keyless chip is the same pair in grey; our star, a faded hop and the
composer's white slot keep what they had.
The value came from two ladders over twelve hues. A fixed step down in
lightness made the blue fills almost black (blue is the wheel's darkest
hue), one lightness for every fill made the blue label hard to read on
its own ground, and 0.65 / 0.55 were too weak in blue / too dull in
yellow. Darkening at constant chroma leaves sRGB, and the clip merged
neighbouring greens into one #008000, so oklab.fitted lowers the chroma
instead and keeps the hue.
A thin seam between two chips of one colour is now drawn in the
previous chip's label colour, the light half of its own pair, on both
palettes; the shaded-fill rule (SEAM_SHADE) is gone with it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The console font gains the two powerline chevrons, drawn into the next
two donor slots (U+2559 and U+255B, which nothing draws), in both the
6x12 and the 6x8 builds. With them in its inventory, the PicoCalc's
verdict is core, so its routes are ribbons of chips like the
Cardputer's.
A chip there can only be filled from the eight background slots, since
under a 512-glyph font the VT has no bright backgrounds. So a node's
chip is the dim twin of its hue (theme.DIM_TWIN), and its label is the
hue itself: the colour the name has in every list and in arrow mode. A
hash label and an annotation are that colour throughout. A keyless chip
is white on light grey, the same pair one bank down. Our star, a faded
hop and the composer's slot have no dark grey to stand on, so they
stand on the page, the slot as a white + in a break of the ribbon. Two
chips of one colour join on a thin chevron in that colour.
Bold is brightness on the console, and Rich merges a selected row's
bold into every span on it, which turned the dim fills bright. Every
chip style there says "not bold"; the tests read the rendered row back
through the emulator's VT to hold that.
The new font takes effect once the font script runs again on the Lyra
(as root). Until then the PicoCalc's installed font lacks the chevrons.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Cardputer's panel is drawn in MeshTerm's own Terminus, which has the
powerline chevrons, yet its paths came out as arrows: the chips-or-arrows
verdict asked the terminal that started the app about its font, and on
the device that terminal is no one's, so the answer was unknown. In the
emulator it was worse, since the verdict was whatever the host terminal
could draw, a question about the wrong glass.
A handheld names its glyph inventory (Platform.font), and every frame is
folded down to that font anyway, so the inventory now answers: after the
MESHTERM_POWERLINE override, before any terminal probe. Terminus has
U+E0B0 and U+E0B1, so the Cardputer draws chips. The PicoCalc's 512-glyph
console font has neither, so its paths stay arrows, now said outright
rather than reached by not knowing. Diagnostics reads
"core (platform:cardputer-zero)". The verdict is bound on a platform
switch and its memo cleared there.
Checked by rasterizing gallery screens through the emulator's VT and
Terminus at 53x14; not yet on the device.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Hold-to-quit was the Cardputer's alone. It is now MeshTerm's gesture
everywhere a key can be seen coming up, which a terminal never reports:
- The emulator's window, for either handheld, which reads its own keys
as the Cardputer's panel does.
- A terminal's Esc, where the keyboard can be asked whether it is still
down: GetAsyncKeyState on Windows, as the right-Ctrl rescue already
asks, and EVIOCGKEY on Linux wherever the keyboard's event device
opens (the PicoCalc, a uConsole, a user in `input`). It reads the key
state only, no events.
An Esc arriving while its key is still down is held back as a hold:
delivered as it comes up, a box after a second, gone at three. The
keyboard's repeats are part of the hold, a key typed during it goes in
behind the Esc, and on a VT stream the last repeat, which prompt_toolkit
can parse after the key is up, is dropped rather than going back a
screen behind a box the reader just let go of. On macOS and over SSH Esc
is unchanged.
SIGTERM is a quit for the interactive menu on every platform now, not
only under the Cardputer's launcher: a system shutting down gets the
same clean exit, terminal and console font restored.
EVIOCGKEY checked on the Cardputer's arm64 kernel; the Lyra was offline.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Cardputer Zero's launcher ends any app whose Esc is held: SIGTERM to
its process group at three seconds, SIGKILL three seconds after that
(cp0_esc_exit_policy.hpp in CardputerZero/launcher; M5's packaging guide
still describes an older Home-for-five-seconds rule). Its own hint never
reaches the panel while an app owns it, and MeshTerm took 3.4 s to exit,
so a held Esc ended in SIGKILL.
Now a tap of Esc is typed on its release, and a hold of one second
floats a "Hold Esc to quit" box whose bar fills over the two seconds
left. While it is up the app holds the device's transmit lock: nothing
new goes on the air, and a scoped channel send under way closes its
window first. Letting go gives the lock back and types nothing. Held to
the end, MeshTerm leaves the way Quit does, and the launcher's SIGTERM
joins that exit instead of interrupting it wherever the main thread
happened to be.
The exit itself fits: openhop-core's cleanup joined the radio's IRQ
thread, parked in an uninterruptible 30 s poll, for its full 2 s
timeout. The node drops that daemon thread from the books first. On the
device, under the real launcher, the process group now drains 1.3 s
after the SIGTERM, with no SIGKILL.
The emulator's window does the same on a held Esc, debounced for an X
server's release-press repeats.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Naming the AIO rested on the uConsole's panel alone. It now takes every
mark, from two independent sources: ClockworkPi's uConsole overlay
describing the hardware in the device tree -- the cwu50 screen, the
AXP228 power chip, the OCP8178 backlight -- and the hardware answering
for itself on USB, the built-in keyboard (1eaf:0024, made by
ClockworkPI, named uConsole; the ids alone are any LeafLabs Maple). An
overlay on the wrong machine has no uConsole keyboard, and the keyboard
on another machine has no uConsole overlay. Any one missing, and the
radio is listed unnamed on the AIO's wiring, as before.
`BuiltinRadio.compatible` gives way to `board`, a test only its machine
passes (`is_uconsole`).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The radio on spidev1.0 was listed as a bare "SPI radio" and reported
"MeshTerm SPI node", because spidev1.0 alone could be any Pi. A
uConsole now proves itself by its own screen: ClockworkPi's uConsole
overlay declares the panel `cw,cwu50`, which no other machine carries
(the DevTerm's is cw,cwd686; the CM's model string and the AXP228 and
backlight are shared). Seen on JP's CM5 uConsole.
`BuiltinRadio` takes a device-tree `compatible` beside its marker path,
read once a run from /sys/firmware/devicetree/base. The named AIO comes
first in BUILTIN_RADIOS; without the panel, the unnamed entry still
lists spidev1.0 on the AIO's wiring as before.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The SPI node reported every radio as "MeshTerm SPI node", so the
Cardputer's Cap read that way wherever a device's model is shown. The
node is now told the shipped board it runs on (`board_name`, matched by
the SPI node on this machine's board, so a profile that adjusts the
Cap's pins is still the Cap) and reports that; a radio on a board
MeshTerm doesn't know by name keeps "MeshTerm SPI node".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
On the Cardputer Zero, connecting to an ESP32 companion that wants a PIN
stalled for a minute and never asked for one. Its link came up and its
services resolved; the UART subscribe made BlueZ ask MeshTerm's agent
for the PIN; with none given the agent refused, and the companion
disconnected -- it hangs up rather than refusing the subscribe. The
guard that holds back drops while bleak retries HCI 0x3e swallowed that
hang-up too, so meshcore waited out its timeout on a dead link.
The agent already records whether BlueZ asked it for the PIN. A drop
after that is now flagged, and the connect ends at once as a pairing
failure -- the words the PIN handling answers -- so the reader is asked
for the PIN in seconds. Drops with no pairing asked are held as before.
Also: a link that opens and then drops during service discovery
("failed to discover services, device disconnected", an HCI Connection
Timeout 1.7 s in, at -22 dBm as readily as at -86) is retried like a
link that never opened. It was surfacing on the first try, and the next
try usually connects.
Checked on the Cardputer against the companion: three connects with no
PIN each answer "requires a Bluetooth pairing PIN" in 2-7 s, where each
took the 60 s timeout before.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Cardputer Zero's Cap carries a GPS on the header's UART, which the
device screen listed as "~ n/a" (pyserial's name for a port with no USB
identity): a companion to connect to, and with nothing remembered, the
lone port a command-line connect would pick. A wiring now names its
board's GPS (`gps_port`, `gps_baud`; the Cap's is /dev/serial0 at
115200), and wherever that radio is listed, its GPS port is not: the
device screen, `meshterm devices`, and autodetect.
The radio node reads it the way MeshCore companion firmware reads a
board's GPS: `gps` (off until switched on, remembered in prefs.json)
and `gps_interval` (0, every fix, once a second), reported and set as
custom variables, which is where Device config already shows them. A
valid RMC or GGA fix becomes the node's position, the one its
self-info reports and, with Share location on, its adverts carry. The
last fix is saved as the node shuts down. Standard library only, as
the node's contract asks; NMEA checksums are checked.
On MeshTerm's side, our own self-info is now refreshed behind the
answer after a minute, like the contacts, so screens opened after the
node moves show where it is; and the connect-time settings offer no
longer flags a position the running GPS moved.
Checked on the Cardputer: the ttyS0 row is gone from `meshterm
devices`; `config custom gps 1` opens the port, is saved, and reads
back on; the real-terminal test passes there.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
On the device MeshTerm ran, drew every frame, and stayed invisible: the
panel kept the launcher's last frame while a splash nobody could see
waited for a key. The panel is a DRM driver whose /dev/fb0 is fbdev
emulation, and the kernel copies to the glass only what it sees change
through the *mapping*. pwrite() put the pixels in the buffer, where
reading /dev/fb0 back showed them, and never on the panel. The frame
now goes through an mmap, as M5's own LVGL apps write theirs. Checked
on the panel.
Also scripts/cardputer-zero/type-text.py: paste a long string (an API
key, a Wi-Fi password) on another computer and it is typed on the
Cardputer's keyboard over SSH, into whatever app has the screen. Input
is hidden and travels on stdin, never on a command line.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
With the Cap powered, I2C bus 1 doesn't only fail reads: some come
back garbled (a capacity of 56414, clamped to a confident 100), and
the kernel's bq27xxx driver caches whatever it read, failure or
garbage, for five seconds. Retrying 0.1 s apart only re-read the
cache, so every retry missed.
Now an unreadable or out-of-range percent is asked again 5.5 s later,
past the cache; a step of more than 5 points between polls waits for
the next poll to confirm it, which catches garbage that lands inside
0-100; and the sysfs reads leave the event loop, since a failing
transfer on that bus takes its time. Measured on the device with the
Cap on: a steady 96%, where it showed nothing or 100 before.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Tried on a real Cardputer Zero for the first time. It draws on the
panel, takes the keyboard, and opens from M5's app launcher:
scripts/cardputer-zero/launcher-entry.sh puts a source checkout there
(and `remove` takes it off) until there is a store package.
Three things the hardware showed:
- The splash opened with the highlight on /dev/ttyS0, the Cap's GPS
UART, because serial is listed first as found. Below the companions
we've used, anything that looks like one now comes before a port
nothing vouches for, across transports.
- The title bar's battery is the handheld's own BQ27220, as the
PicoCalc's is. The host pack is now found by what it is (a Battery
supply that isn't a peripheral's), not by the PicoCalc's name.
- While the Cap is powered, I2C bus 1 turns noisy: a quarter to three
quarters of the gauge's reads fail with an I/O error (one in ten of
the clock's), none with the header off. A failed read is asked again,
a status that won't read keeps its last verdict, and a good reading
stands through six failed polls instead of blinking out.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
On a Cardputer Zero the device screen lists the Cap as "Cap LoRa-1262",
and --spi reaches it, with no profile: its wiring ships beside the
AIO's, listed when the board's own ext_5v_out switch says this is a
Cardputer Zero (spidev0.1 alone would claim every Pi with SPI on).
The Cap needs two things before the chip answers, and the wiring can
now say both. `leds` sets the switches the device tree hands to
gpio-leds — ext_5v_out powers the header, ext_usb_gpio_fun=0 keeps
the reset and interrupt pins on GPIO rather than USB — and puts them
back when the node ends, so the header is unpowered again after
MeshTerm quits. `pi4io_*` drives the Cap's PI4IOE5V6408 at 0x43, which
enables the RF path; a silent expander is reported as the Cap not
being attached.
Checked on the hardware: reset 26, busy 22, irq 23 on spidev0.1 (the
lines M5's factory test drives), the node comes up from a cold
header, and it hears the mesh at +12 dB.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A full T1000-E read 9 or 10% in the header on Ubuntu, while the radio
reported 4.21 V. The battery was never the fault. A full pack's "100"
lost its last zero, and 99% lost its 9 to the % sign that followed it.
MeshTerm adds one column to the width the terminal reports, because
prompt_toolkit's Windows console output reports one column fewer than
the window has, on purpose. That was done on every desktop, but a Linux
or macOS terminal reports its width exactly. There the extra column fell
off the screen. prompt_toolkit draws with autowrap off, so each row's
last character landed on top of the one before it, and whatever sat
flush right lost its end. The battery gauge sits flush right.
The column is now added only where the terminal hid one, which is the
Windows console. The full_width preference and MESHTERM_FULL_WIDTH still
override either way. Driven into an exact 80-column terminal under
--mock, the header now ends "100" at 4.21 V and "98%" at 4.18 V, where
it ended "10" and "9%". In a real Windows console the reported 119
columns still widen to 120.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Only 24 of 144 direct messages and posts in the history were ever marked
delivered. Three faults, each fixed:
- The wait was the radio's own estimate times 1.2: about 5.7 s for a
flood, while a flooded round trip from a few hops out takes ten (a
room's answer from five hops landed at 10.1 s on hardware). It is now
the login's patience, the estimate stretched a quarter plus a round
trip sized to the route, about 15 s for a flood.
- An ack that landed after the wait was dropped. The radio pushes one
whenever it arrives, for any of its last eight sends, but nothing was
listening, so a delivered message stayed failed for good. The history
shows ACK and PATH packets heard seconds after messages recorded as
failed. The send now returns a Delivery carrying the ack code, and the
chat service keeps the last sixteen messages still waiting. A late ack
flips one to delivered, in history and in the message the chat screen
holds, and every code a retried message went out under counts.
- The Courier trusted the failure and sent the message again: a
duplicate for the recipient whenever the ack was only late. A late ack
now marks the entry delivered.
The ack listener is also armed before the send rather than after it,
so a neighbour's ack quicker than the send's own return is not missed.
The debug log names how a message went out and the radio's own
round-trip time.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A long room name ran the join's caption ("logging in to YJN-ROOM-OBS
St-Jean along its 6-hop route… 37 s") one cell past its box, and the
count's last letter wrapped alone to column zero. A busy card's caption
may now carry line feeds, each later line hanging under the caption's
text rather than under the spinner. The card is as wide as its widest
line, and a line too long for the screen wraps under the text too. The
join puts the room on the first line, and how the login went out with
its running count on the second.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Joining a room moves out of Chat into a Rooms page beside Channels. It
lists every room server the radio knows, joined ones first, with each
room's access, unread count, last post and last-heard time. Each room
has a page: join, log in again, change password, open the board, forget
it, or delete its stored posts. Chat now lists only joined rooms, the way
it lists only channels on a slot, and only a joined room raises the
unread badge. chat join becomes rooms join, alongside rooms list and
rooms forget.
"No reply" told the reader nothing, because a room is just as silent for
a wrong password as for a login that never reached it. Testing on
hardware showed the commonest cause is neither: a login sent along the
route the radio had learned met silence, and the same login by flood was
accepted in eleven seconds. A silent login is now explained from what is
known: how it went out, when the room was last heard, and whether this
password has worked before. It offers the one next step that evidence
points at:
- Try by flood, after a learned route met silence. The radio forgets
the route (local, nothing transmitted), then sends the login once.
- Try again…, after a flood met silence. The prompt reopens with the
password filled in, so Enter resends it and typing replaces it.
- Add it & try again, when the radio doesn't hold the room as a contact.
Every retry is a button press, and none is ever sent on its own. A
password over 15 bytes is refused at the prompt, since MeshCore sends
only the first 15.
On the board, a failed or unfinished login used to leave the compose
line taking posts the room would drop. It now opens only while we are in
and may post, and otherwise says why and what to press. Opening a joined
room that has gone quiet logs in in the background. ^L runs the
explained flow.
Underneath, the login exchange listens for the radio's own confirmation,
matched to the login by its expected ack. It records whether the login
went by flood or along a route, or the radio's reason for not sending it
at all. This also corrects the debug log: "send unacknowledged" read the
library's wait for the *answer* as the send. On hardware the radio
confirms in under 0.1 s. Device.reset_route is new. The simulator learns
routes, can be given a stale one, and refloods.
Refs: #24
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A review of the room work turned up ways the remembered passwords could
go wrong, and a doubled login:
- A room that lets readers in lets any password in, so a login that only
earned read-only proved nothing about the password. It replaced the
room password anyway, so one mistyped ^L left a member read-only for
good, with no prompt to fix it. It now never replaces a password that
is already kept, and ^L asks for one while you are read-only.
- A blank login only asks "do you know me?". Accepted, it overwrote the
remembered room password, and as an admin it emptied the admin
password Repeater admin shares. It now replaces neither.
- A room's login replaces a known member's role with the one the
password earns, so an owner who held both passwords was demoted to
member each time the room password was tried first. The admin password
now goes first.
- Closing a room view and opening it again mid-login sent a second
login. The room service now shares a login already in flight, and a
room with one under way is not due another.
- A refusal forgot the room password even when it refused a different
one. ^R is no longer offered to a read-only member, whose resend the
room would drop. Enter on a picked post opens its paths when there is
nowhere to put a reply, and a paste is refused there too.
Refs: #24
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
chat join logs in to a room server and prints what it allowed: member,
admin or read-only. It uses --password, or the password MeshTerm
remembers for the room. A room says nothing to a wrong password, so a
join that hears nothing fails (exit 4) and names both possible reasons.
send --to a room posts to it, and its receipt says kind "room". A room's
history names each post's author where a direct transcript names the
peer, drawn the way a route draws a hop: "Alice (d4e5f6a7)", or the bare
hash when nothing names the key. The document carries the author on
every row, null outside a room, so history answers in one shape for
every kind of conversation; listen streams it too. chat list calls a
room a room.
The CLI manual, the features guide and the README describe rooms, and
CLAUDE.md adds room and post to the lexicon, 📌 to the icons, and ^L to
the chord list.
Refs: #24
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A room server is a message board on a radio: it keeps its members' last
posts and sends each member the ones they missed. MeshTerm could see rooms
and run your own, but not join one. Posts that reached the radio were filed
as direct messages from the room, with their authors thrown away.
Chat now has a Rooms section between Channels and Direct. Opening a room
shows its board at once, each post under the member who wrote it. A post
carries the first four bytes of its author's key, so it is named by any
contact or overheard advert with that key and coloured by the key itself.
An author nothing names stands as the hash, in the unknown-node grey. The
first visit asks for the room's password over the board. After that,
opening a room logs in again only once it has been quiet for
room_relogin_minutes (a new preference, 30 by default), and ^L (F6 on the
handhelds) logs in on demand. The title says what the room allowed:
member, admin or read-only. A read-only member gets a line saying the room
drops their posts, in place of a compose box. A post goes out once, never
soft-retried, since a duplicate would reach every member.
Underneath: messages carry an author (schema 18), a room's board leaves
out its command replies, and a re-sent post is recorded once. Room posts
raise the unread badge and a room's command replies don't. The device
gains room_login, sharing admin_login's exchange and reading the access
the room granted. Room passwords are remembered in rooms.json; an admin
login is kept in admin.json, shared with Repeater admin. The simulator
has a room, Lakeside BBS, to join with "hello".
Two fixes the research turned up. An admin command took the first direct
message from anyone as its reply, so a room post or a companion's message
landing mid-command was read as the node's answer. It now listens before
sending, and only for a reply from that node that is not a post. And
Repeater admin names a room server's guest password "Room password", the
password members join with.
Refs: #24
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Each hop row drew its link as a whole route, square at both ends, so every
hop read as a walk that began and ended there. A hop is one link of the
walk: only the first hop's start and the last hop's finish are the route's
own ends. Every other end is a node the route runs on through, and now
wears the notch or the point (a bare → in arrow mode), the marks a path
line already uses for "this goes on".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Each hop used to read as "origin (hash) → destination (hash)" in a
four-column table, which ran past the edge on any link between two long
names. The link now draws through the path widget the way the route lane
above it does: names with no hash after them, an unnamed node by its hash
(the only identity it has), and our own end as the bare star.
Where every row fits on one line the hops stay a table, one line each. Where
any row doesn't, every row takes two lines, all alike so the readings never
jump between lines: the hop number and its path on the first, and the
reading with its meter flush right on the second. A path still wider than
its lane breaks off on its cut mark and slides under ←→, with the hop number
pinned and each path stopping at its own tail. The footer offers
"←→ scroll" only while something can slide, and sliding lets go of the
action cursor so the page stays on the hops being read.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Once a trace had run, ↓ on the last action did nothing. The results sat in a
window of their own under the actions, sized to whatever rows the controls
left, so the page came out exactly as tall as the screen: the frame had
nothing to scroll, and edge scroll with it, while "↓ n more" sat under the
last action. Only PgUp/PgDn slid that window.
The results now follow the actions on the page. ↓ off the last action
scrolls on down through them to the oldest trace, the snap-back brings the
actions back, and PgUp/PgDn/Home/End page the whole screen, the way the TX
sweep and Records pages already work.
So that no other screen hides rows this way, ListWindow.fit and fit_blocks
take the highlight with no default: a window is for a list the highlight
walks, since every row it hides must be one an arrow reaches by walking onto
it. Anything nothing highlights belongs on the page, where the frame scrolls
it. The trace results were the only window over rows no highlight walks.
CLAUDE.md states the rule beside edge scroll.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The transport badge now leads each row with no heading over it. A list
whose rows declare `pans` slides its column header and every table row by
one shared shift, kept as the highlight moves and stopping where the
widest line meets the edge, so the splash box keeps its width. Rows
outside the table, like Add a network device and Quit, stay put.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Each lane was squeezed against a budget so the row fitted the box, which
cut names to "Johnp…" beside a long address. Now every lane is sized to
its longest value, the header keeps its full labels, and a row wider than
the box runs off its edge for ←→ to read. The address moves to the end.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Zooming in showed the ghost, then a black screen, then the map drawn in. The
black was a raster started the moment the view changed, before the new tile
zoom's tiles had loaded. It had nothing to draw, and being the view's own
frame it replaced the ghost. On the PicoCalc it stayed up 0.4 to 1.0 s, and
sometimes a half-drawn frame followed it before the real one. The same
happened when the map first opened.
The map now starts no raster while a tile of its view is still on its way.
It keeps standing on what it has (the aligned frame, or the ghost), and the
first raster it draws is the real picture. A tile that comes back as silence
is no longer on its way, so a map with no network still draws what it can.
Measured with a zoom tour on the PicoCalc: in to z11, z12 and z13, then back
out to z10. Before, every zoom across a tile zoom published a raster with
none of its tiles. Now each zoom draws one raster with all of them. The
finished picture arrives as soon as it did before (0.8 to 1.35 s in, 0.4 to
0.7 s out), with 6 rasters drawn instead of 13.
Tests that exercised raster scheduling now let their tiles answer first,
since a captured load used to stay on its way for ever.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A click on one of the Cardputer Zero's keys drawn under the panel, or on a
chip of the lane along the panel's bottom row, types that slot's key. The
PicoCalc has no keys drawn, so its chips are where you click. Shift held
through the click types the slot's Shift companion. The app reads the bank
off the key that arrives, not off the Shift it saw, so the window spells the
shifted key itself, the way the desktop's Shift+F-key already was: F6 to F10
on the PicoCalc, xterm's shifted F4 to F8 on the Cardputer.
The drawn keys take the lane's own fills, resolved from the deck's styles
through the panel's palette, so a key and the chip above it are one colour.
They turn blue with the lane while Shift is down, and fill solid while
pressed. Losing focus lets go of Shift, which a release in another window
never would.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Panning or zooming the map quickly on the PicoCalc could stop the whole app
for many seconds. A watchdog run on the device found two causes behind it,
plus the port walk fixed in the previous commit.
Memory. A decoded tile holds 1 to 4.6 MB of RAM on the device (65 bytes a
point), not the 0.9 MB the code assumed. The source kept up to 24 of them,
counted as tiles, and the map screen kept 8 or more of its own on top, so a
few pans and a zoom out filled a 104 MB device. Swap is on the SD card, so
the process then stopped while the kernel read its code back in: 16,000
major page faults and a 15 s freeze in one run. Now the source holds decoded
tiles to a byte budget, one eighth of the machine's RAM (12.8 MB on the
PicoCalc, at most 128 MB on a desktop), weighed by mvt.resident_bytes. That
estimate was fitted against tracemalloc on the device's 157 cached tiles: 1%
under to 9% over there, and 6% under to 21% over on a 64-bit desktop. The
screen keeps only the tiles in view, and takes any tile already in RAM
straight from the source on the paint (BasemapSource.resident), with no trip
through a thread. The prefetcher warms the disk cache (BasemapSource.warm)
instead of filling the RAM with guesses, and passes over any tile already
written down. Tile loads take turns, two at a time, and a tile the view has
left before its turn comes is dropped, not loaded.
CPU. fill_polygon asked every edge about every scanline, which at z7 means
coastlines with tens of thousands of edges: 6.8 s of a 7.6 s frame. Each
edge is now filed under only the rows it crosses, and draws the same dots
(checked against the old fill on 3,000 random polygons, and in a test).
_draw_tile also tests each ring's bounds before projecting it, and skips a
ring wholly outside the view. That changes no dot: 1,500 frames from the
device's tiles match rendering with no culling at all. The one frame that
differed from HEAD was HEAD's own building cull, which dropped a footprint
whose fill rounds onto the canvas edge.
Measured on the PicoCalc with the profile-picocalc map_burst tour (keys sent
at 10 a second, from a pinned z10 view, real radio and database). The new
code ran first, so the baseline had the warmer tile cache:
pans, catch-up after a burst HEAD 1.0-1.8 s now 0.4-0.6 s
longest silence, whole tour HEAD 11.6 s now 0.6 s
whole-process freezes HEAD 8.0, 3.1, 2.1, 11.6 s
now one, 1.6 s (idle pages swapped back in)
swap at its peak HEAD 25 MB now 7 MB
warm raster, z7 / z8 / z9 HEAD 7.6 / 1.9 / 1.8 s
now 0.57 / 0.29 / 0.81 s
warm raster, z10 / z13 HEAD 0.49 / 0.56 s now 0.26 / 0.26 s
Still open: z9 spends 0.23 s drawing lines; a prefetch decode of a heavy z7
tile is still 1.3 to 1.7 s on a worker thread; and pages swapped out while
the map sits idle can still cost a second or two to come back.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Edge scroll reached only the eight screens that answered cursor_at_edge, one
gate each; the trace screen, the autocomplete prompt and anything built
tomorrow went without. Now no screen says where its edges are. An arrow goes
to the screen like any key, and the next paint reads the answer off the page:
a highlight on the same line, drawn the same, moved nothing, so the page
takes the line instead. Comparing the drawn line, not just its index, keeps a
list that scrolls in its own window from looking stuck. Arrows pressed faster
than the screen paints are judged together. All a screen supplies is its
highlight's line in cursor_line, which keeping it in view asked for already.
The eight gates and row_at_edge are gone.
The snap-back's arrow is now the one pointing at the off-screen highlight,
read from where the paint left it, so it no longer needs to know which edge
the page went past. Home and End take the page to its ends after the screen
has handled them, without assuming the highlight followed.
The remote CLI opts out (edge_scrolls = False): the line it keeps in view is
its prompt, and its arrows recall history. The autocomplete prompt now names
its highlighted suggestion, so a short box follows it; an empty select list's
"no matches" line stands in for the highlight; and the reorder list's Home
and End jump to its first and last rows, as on every list.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
