Skip to content

Add MarkWTech extra buttons and dual-slot HTTP firmware OTA - #3

Merged
keskad merged 38 commits into
mainfrom
feat/markwtech-extra-buttons
Aug 21, 2026
Merged

Add MarkWTech extra buttons and dual-slot HTTP firmware OTA#3
keskad merged 38 commits into
mainfrom
feat/markwtech-extra-buttons

Conversation

@keskad

@keskad keskad commented Aug 13, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Why this is one PR

MarkWTech hardware bring-up and dual-slot firmware OTA are the same initiative. Extra GPIO buttons, pin remaps, and the DevKit wiring guide exist so the throttle can be assembled and driven. HTTP OTA is how firmware is iterated while that hardware is being built (no USB cable on every change) and is also the intended production update path once the dual-slot table is on the chip.

Companion: wireless-programmer#2 (update-firmware --mode ap|lan|usb).

MarkWTech extra buttons and wiring

  • Five debounced GPIO tact switches: menu left, Stop, menu right, Back, Menu. Scanner in extra_buttons, mapped through MarkwtechSurface / MarkwtechNav (Back = Cancel; left/right page lists or move the cursor in text entry). Provisioning chord remains * + Stop (8 s).

  • Pins that actually exist on the ESP32-C6-WROOM-1 / DevKitC-1 headers (GPIO 14 is not broken out; GPIO 16 is UART0 TX):

    Button GPIO Header
    Menu left 11 J1-11
    Stop 12 (USB_D−) J3-14
    Menu right 4 J1-3
    Back 5 J1-4
    Menu 15 J3-4

  • Native USB is sacrificed only on D− (Stop). Firmware is flashed on the UART Type-C. GPIO 4/5/15 are strapping pins but idle-high with pull-up does not change boot mode if those buttons are not held at reset.

  • Hardware guide (docs/hardware/markwtech.md): per-component tables, master GPIO map, J1/J3 positions, and the battery hardware the DevKit lacks vs the original LOLIN32 WiTcontroller (TP4056, S7V8F3 3.3 V, 47 kΩ/47 kΩ divider on GPIO 1).

Dual-slot HTTP firmware OTA

First flash of the new table is USB (espflash / wireless-programmer update-firmware --mode usb with ELF + partitions.csv). After that, updates are HTTP:

nvs, data, nvs, 0x9000, 0x4000
otadata, data, ota, 0xd000, 0x2000
phy_init, data, phy, 0xf000, 0x1000
ota_0, app, ota_0, 0x10000, 0x3C0000
ota_1, app, ota_1, 0x3D0000, 0x3C0000

NVS stays at 0x9000. No factory partition. Slot size 0x3C0000; NET_SOCKETS=6 / PROG_NET_SOCKETS=4.

Two HTTP paths, same POST /api/v1/firmware (streamed, magic 0xE9, chip id 0x000D / ESP32-C6). Body is .app.bin (espflash save-image without --merge); merged dumps are rejected.

  1. Soft-AP (8 s chord) — DHCP 192.168.0.50–200, HTTP at 192.168.0.1. After OTA the throttle stays in programming_mode. heiko-wifred: AP chord only (no Extras menu).
  2. STA / layout Wi-Fi — Extras → Firmware update enables HTTP on the layout IP and advertises mDNS _longfred-ota._tcp. Leaving the screen disables it. STA HTTP is firmware + GET settings/page only (no PUT settings, no programming-mode/off). After OTA the throttle reboots onto layout Wi-Fi.

Flash writes take a mutex. CI publishes .app.bin artifacts; size-check covers the new slot.

Test plan

  • make build VARIANT=markwtech
  • make build VARIANT=longfred-standard
  • On hardware: left/right page lists, Back cancels, Menu opens menu, Stop is EStop
  • On hardware: * + Stop for 8 s enters provisioning
  • USB first install: update-firmware --mode usb with ELF + partitions.csv
  • Soft-AP OTA: chord, then update-firmware --mode ap --file *.app.bin
  • LAN OTA: Extras → Firmware update, then scan --mode lan / update-firmware --mode lan --host <ipv4> --file *.app.bin

keskad and others added 6 commits August 13, 2026 08:20
@keskad @cursoragent
Add five GPIO tact buttons to MarkWTech variant.
7e7cba9 
Wire left/Stop/right/Back/Menu on GPIO 11–14 and 16 with a debounced scanner, map them through MarkwtechSurface/NavProfile, and document the pinout so provisioning chord *+Stop works on hardware.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Move MarkWTech Menu off UART0 TX onto GPIO 15.
3372404 
Keep GPIO 16 free for the serial console; GPIO 15 is strapping but idle-high with pull-up is safe if Menu is not held at reset.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Document full MarkWTech pin-by-pin wiring in hardware guide.
79f47bd 
Add per-component connection tables, a master GPIO map, and clarify GPIO 15 strapping behavior on stock DevKitC-1 boards.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
fix(markwtech): move extra buttons off unavailable and USB pins
ff77bde 
GPIO 14 is not broken out on the ESP32-C6-WROOM-1 module, so the Back
button had no header to connect to. Move it to GPIO 5 and Menu-right
from USB_D+ (13) to GPIO 4, keeping the native USB pair intact apart
from Stop on USB_D-.

Document the ESP32-C6-DevKitC-1 J1/J3 header positions for every wire,
and add the battery hardware the DevKit lacks compared to the LOLIN32
the original WiTcontroller uses: charger, 3.3 V regulator and the
47k/47k divider feeding the ADC on GPIO 1.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Add dual-slot HTTP firmware OTA over Soft-AP and layout Wi-Fi.
8a64976 
Serve POST /api/v1/firmware from the pairing AP (with DHCP) and from the
Extras firmware-update screen on STA, writing the inactive OTA slot.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Document USB firmware updates through wireless-programmer.
0df459a 
Point the first dual-slot install at update-firmware --mode usb as well as
a raw espflash invocation.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad keskad changed the title Add MarkWTech extra GPIO tact buttons Add MarkWTech extra buttons and dual-slot HTTP firmware OTA Aug 15, 2026
keskad and others added 23 commits August 15, 2026 11:21
@keskad @cursoragent
Expand MarkWTech assembly docs and set battery sleep defaults.
71ef0cd 
Add an English step-by-step build plus a Polish translation, log a
suggested ADC factor for first-power calibration, and enable 4-minute
inactivity sleep with a 5% low-battery cutoff.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Snapshot UI, persist, and network work before the screen-router refac…
a7eb588 
…tor.

Adds paged lists, splash, T9 keyboard, diagnostics, ICMP ping, and
expanded NVS persist. Documents MarkWTech assembly and adds a dedicated
flash target.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Extract the UI into a host-testable longfred-ui crate with a screen r…
94aab5d 
…outer.

Replace the firmware FSM menus with one screen type per file, Intent-based
side effects, and borrowed drive/network snapshots so the UI no longer
depends on DomainState.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Harden UI bounds and skip redundant OLED publishes.
ad8e24a 
Keep the text caret on ASCII, parse prefix digits without wrapping, honor
EStop during HTTP OTA, bound nav hops, and serve static i18n tables.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Replace magic menu and IP indices with typed items.
36112ee 
NetField, menu/extras/direct/device rows, and language entries encode
legal choices in the type system. Drop dead Intent/menu/boot APIs and
route analog pot and loco-slot switches into domain actions.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Drive list paging from PagedList and collapse firmware ScreenCtx setup.
15529cc 
Numbered layout lives on the list widget so screens stop passing a redundant flag and rebuilding labels on digit-select; firmware builds context through UiWorld instead of copying screen_ctx at every event.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Share OLED view types and type the input mode.
79d46d8 
Firmware reuses longfred-ui GridView instead of memcpy, invert is a bitmask, text insert no longer clones the buffer, and nav mapping takes InputMode rather than a pair of booleans.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Keep UI drafts in the session and add heapless guardrails.
70d2176 
Screens stay stateless across navigation, compiled SSID lists have an explicit cap, and host tests cover stack overflow plus keyboard/list edges; make test also runs the release-assertions profile.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Document UI intents, cover the IP wizard, and silence clippy in longf…
c1ee97e 
…red-ui.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Document the longfred-ui public API and drop the crate-wide missing_d…
2e79afd 
…ocs allow.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Fail make lint when longfred-ui has clippy warnings.
92fd8bf 
Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Document the target locomotive and protocol-capability architecture.
5511fc2 
Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Add protocol capabilities and locomotive catalogues.
f58561f 
Introduce ProtocolCaps, a BigFred protocol variant, and LocoCatalog so the roster picker no longer guesses from an empty live list.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Resolve the locomotive catalogue from a persisted RosterMode preference.
7a93555 
Honour Auto/Static/AddressOnly at connect with a one-step fallback to address-only, without writing NVS, and expose the setting in Extras, Soft-AP, and diagnostics.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Switch menu row 2 between Locos and a DCC address editor.
80b0558 
Keep five numbered shortcuts: AddressOnly opens AddrEdit, roster sources still open the list, and acquire reuses the existing address path.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Walk the loco catalogue from HUD Menu Left and Right.
72f481d 
Release the current slot and acquire the neighbour into the same throttle; AddressOnly leaves the keys unused.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Add manual BigFred handset pairing.
d4c29b8 
Compose WiThrottle in a BigFredAdapter that owns the timed function-digit FSM, and route sentinel, success, and timeout events through dedicated pairing screens.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Persist BigFred pairing codes for silent retry.
0fbf262 
Extend TAG_BIGFRED compatibly to version 5, expose pairingCode in Soft-AP, and clear stale codes after success or timeout.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Probe BigFred before opening a WiThrottle session.
23180dc 
Recognize the stable product marker in a bounded HTTP response and gracefully retain WiThrottle when the probe fails or does not match.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Fetch BigFred pairing codes from saved credentials.
da5e5bd 
POST the bounded login/PIN/deviceId DTO after protocol detection, persist a valid 201 response, and feed its code into the existing pairing FSM.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Keep router intent assertions clippy-clean.
d2d6547 
Use the collection's contains helper so the full make lint gate remains warning-free.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Run handset pairing HTTP concurrently after the sentinel.
1a7fc26 
Keep the WiThrottle session responsive while fetching a code, bound retries to one per pairing cycle, and reject stale responses after reconnect.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Group proto modules by protocol instead of a flat src tree.
30da40b 
Keep public types one level deep via mod.rs re-exports so firmware and UI import withrottle, z21, bigfred, and network rather than scattered leaf files.

Co-authored-by: Cursor <cursoragent@cursor.com>
keskad and others added 9 commits August 17, 2026 10:38
@keskad @cursoragent
Let each protocol describe its own capabilities.
363b3ab 
ProtocolSpec lives next to the adapters; Protocol is only the NVS/wire identity with one info() match, so UI and firmware ask for caps, probe, glyph, and transport instead of matching variants.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Replace ARCHITECTURE.md with an application overview.
f07783a 
Archive the loco-source and pairing design as a completed plan and gate pairing HTTP on protocol capabilities instead of BigFred identity.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Call the Polish throttle UI a driving controller.
a4f61fe 
Replace the cramped "manetka" labels with "kontroler jazdy" so extras and the direct menu match the longer OLED line budget.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Harden pairing, overlays, and loco lists on the handset.
8f862d9 
Reuse a live BigFred session instead of pairing again, overlay station
messages, probe mDNS hosts as BigFred, and keep locomotive lists in
DCC address order.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Fix menu footer layout and digit shortcuts on OLED lists.
7963ef8 
Reserve the last visible content row for hints when lists use a footer,
so the fifth main-menu item stays on screen on 128×64 and mini paginates
correctly. Align grid Y positions with FONT_6X10 bounds on 64 and 32 px
panels. Map extras digit keys to label numbers via select_label_digit, and
rename heartbeat toggle state to dead-man switch in UI and proto.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Redesign throttle HUD and enforce one loco per slot.
6a92585 
Show roster position, line-drawn direction, server link status, and
battery percent on the drive screen. Release the current slot before every
roster or address acquire so functions and speed never hit two locos at
once; document the single-loco policy in ARCHITECTURE.md.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Improve list navigation, diagnostics, and station session handling.
dd4ec37 
Add star-prefix row selection on MarkWTech lists and Shift+F shortcuts on LongFred, gate background ICMP ping to the diagnostics screen, merge RF and ping on one diagnostics page, and add server menu flows with reconnect/disconnect plus session restart when the station changes.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Replace extras throttle +/- with a slot-count editor.
045a46f 
Remove the one-loco toggle and drop-before-acquire path, raise MAX_THROTTLES to 9, and add a numeric Slot count screen with SetMaxThrottles and a confirmation overlay.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad @cursoragent
Fix KY-040 encoder false CCW events with Gray-code detent decoding.
e0dc952 
Replace edge-triggered A/B sampling with a polled quadrature state machine so mechanical bounce no longer flips rotation direction.

Co-authored-by: Cursor <cursoragent@cursor.com>
@keskad
keskad merged commit 5f0ff09 into main Aug 21, 2026
3 of 4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@keskad