* Decouple ButtonThread from sleep.cpp
Reorganize sleep observables. Don't call ButtonThread methods inside doLightSleep. Instead, handle in class with new lightsleep Observables.
* InkHUD: initial commit (WIP)
Publicly discloses the current work in progress. Not ready for use.
* feat: battery icon
* chore: implement meshtastic/firmware #5454
Clean up some inline functions
* feat: menu & settings for "jump to applet"
* Remove the beforeRender pattern
It hugely complicates things. If we can achieve acceptable performance without it, so much the better.
* Remove previous Map Applet
Needs re-implementation to work without the beforeRender pattern
* refactor: reimplement map applet
Doesn't require own position
Doesn't require the beforeRender pattern to precalculate; now all-at-once in render
Lays groundwork for fixed-size map with custom background image
* feat: autoshow
Allow user to select which applets (if any) should be automatically brought to foreground when they have new data to display
* refactor: tidy-up applet constructors
misc. jobs including:
- consistent naming
- move initializer-list-only constructors to header
- give derived applets unique identifiers for MeshModule and OSThread logging
* hotfix: autoshow always uses FAST update
In future, it *will* often use FAST, but this will be controlled by a WindowManager component which has not yet been written.
Hotfixed, in case anybody is attempting to use this development version on their deployed devices.
* refactor: bringToForeground no longer requests FAST update
In situations where an applet has moved to foreground because of user input, requestUpdate can be manually called, to upgrade to FAST refresh.
More permanent solution for #23e1dfc
* refactor: extract string storage from ThreadedMessageApplet
Separates the code responsible for storing the limited message history, which was previously part of the ThreadedMessageApplet.
We're now also using this code to store the "most recent message". Previously, this was stored in the `InkHUD::settings` struct, which was much less space-efficient.
We're also now storing the latest DM, laying the foundation for an applet to display only DMs, which will complement the threaded message applet.
* fix: text wrapping
Attempts to fix a disparity between `Applet::printWrapped` and `Applet::getWrappedTextHeight`, which would occasionally cause a ThreadedMessageApplet message to render "too short", overlapping other text.
* fix: purge old constructor
This one slipped through the last commit..
* feat: DM Applet
Useful in combination with the ThreadedMessageApplets, which don't show DMs
* fix: applets shouldn't handle events while deactivated
Only one or two applets were actually doing this, but I'm making a habit of having all applets return early from their event handling methods (as good practice), even if those methods are disabled elsewhere (e.g. not observing observable, return false from wantPacket)
* refactor: allow requesting update without requesting autoshow
Some applets may want to redraw, if they are displayed, but not feel the information is worth being brought to foreground for. Example: ActiveNodesApplet, when purging old nodes from list.
* feat: custom "Recently Active" duration
Allows users to tailor how long nodes will appear in the "Recents" applets, to suit the activity level of their mesh.
* refactor: rename some applets
* fix: autoshow
* fix: getWrappedTextHeight
Remove the "simulate" option from printWrapped; too hard to keep inline with genuine printing (because of AdafruitGFX Fonts' xAdvance, mabye?). Instead of simulating, we printWrapped as normal, and discard pixel output by setting crop. Both methods are similarly inefficient, apparently.
* fix: text wrapping in ThreadedMessageApplet
Wrong arguments were passed to Applet::printWrapped
* feat: notifications for text messages
Only shown if current applet does not already display the same info. Autoshow takes priority over notifications, if both would be used to display the same info.
* feat: optimize FAST vs FULL updates
New UpdateMediator class counts the number of each update type, and suggets which one to use, if the code doesn't already have an explicit prefence. Also performs "maintenance refreshes" unprovoked if display is not given an opportunity to before a FULL refresh through organic use.
* chore: update todo list
* fix: rare lock-up of buttons
* refactor: backlight
Replaces the initial proof-of-concept frontlight code for T-Echo
Presses less than 5 seconds momentarily illuminate the display
Presses longer than 5 seconds latch the light, requiring another tap to disable
If user has previously removed the T-Echo's capacitive touch button (some DIY projects), the light is controlled by the on-screen menu. This fallback is used by all T-Echo devices, until a press of the capacitive touch button is detected.
* feat: change tile with aux button
Applied to VM-E290.
Working as is, but a refactor of WindowManager::render is expected shortly, which will also tidy code from this push.
* fix: specify out-of-the-box tile assignments
Prevents placeholder applet showing on initial boot, for devices which use a mult-tile layout by default (VM-E290)
* fix: verify settings version when loading
* fix: wrong settings version
* refactor: remove unimplemented argument from requestUpdate
Specified whether or not to update "async", however the implementation was slightly broken, Applet::requestUpdate is only handled next time WindowManager::runOnce is called. This didn't allow code to actually await an update, which was misleading.
* refactor: renaming
Applet::render becomes Applet::onRender.
Tile::displayedApplet becomes Tile::assignedApplet.
New onRender method name allows us to move some of the pre and post render code from WindowManager into new Applet::render method, which will call onRender for us.
* refactor: rendering
Bit of a tidy-up. No intended change in behavior.
* fix: optimize refresh times
Shorter wait between retrying update if display was previously busy.
Set anticipated update durations closer to observed values. No signifacant performance increase, but does decrease the amount of polling required.
* feat: blocking update for E-Ink
Option to wait for display update to complete before proceeding. Important when shutting down the device.
* refactor: allow system applets to lock rendering
Temporarily prevents other applets from rendering.
* feat: boot and shutdown screens
* feat: BluetoothStatus
Adds a meshtastic::Status object which exposes the state of the Bluetooth connection. Intends to allow decoupling of UI code.
* feat: Bluetooth pairing screen
* fix: InkHUD defaults not honored
* fix: random Bluetooth pin for NicheGraphics UIs
* chore: button interrupts tested
* fix: emoji reactions show as blank messages
* fix: autoshow and notification triggered by outgoing message
* feat: save InkHUD data before reboot
Implemented with a new Observable. Previously, config and a few recent messages were saved on shutdown. These were lost if the device rebooted, for example when firmware settings were changed by a client. Now, the InkHUD config and recent messages saved on reboot, the same as during an intentional shutdown.
* feat: imperial distances
Controlled by the config.display.units setting
* fix: hide features which are not yet implemented
* refactor: faster rendering
Previously, only tiles which requested update were re-rendered. Affected tiles had their region blanked before render, pixel by pixel. Benchmarking revealed that it is significantly faster to memset the framebuffer and redraw all tiles.
* refactor: tile ownership
Tiles and Applets now maintain a reciprocal link, which is enforced by asserts. Less confusing than the old situation, where an applet and a tile may disagree on their relationship. Empty tiles are now identified by a nullptr *Applet, instead of by having the placeholderApplet assigned.
* fix: notifications and battery when menu open
Do render notifications in front of menu; don't render battery icon in front of menu.
* fix: simpler defaults
Don't expose new users to multiplexed applets straight away: make them enable the feature for themselves.
* fix: Inputs::TwoButton interrupts, when only one button in use
* fix: ensure display update is complete when ESP32 enters light sleep
Many panels power down automatically, but some require active intervention from us. If light sleep (ESP32) occurs during a display update, these panels could potentially remain powered on, applying voltage the pixels for an extended period of time, and potentially damaging the display.
* fix: honor per-variant user tile limit
Set as the default value for InkHUD::settings.userTiles.maxCount in nicheGraphics.h
* feat: initial InkHUD support for Wireless Paper v1.1 and VM-E213
* refactor: Heard and Recents Applets
Tidier code, significant speed boost. Possibly no noticable change in responsiveness, but rendering now spends much less time blocking execution, which is important for correction functioning of the other firmware components.
* refactor: use a common pio base config
Easier to make any future PlatformIO config changes
* feat: tips
Show information that we think the user might find helpful. Some info shown first boot only. Other info shown when / if relevant.
* fix: text wrapping for '\n'
Previously, the newline was honored, but the adojining word was not printed.
* Decouple ButtonThread from sleep.cpp
Reorganize sleep observables. Don't call ButtonThread methods inside doLightSleep. Instead, handle in class with new lightsleep Observables.
* feat: BluetoothStatus
Adds a meshtastic::Status object which exposes the state of the Bluetooth connection. Intends to allow decoupling of UI code.
* feat: observable for reboot
* refactor: Heltec VM-E290 installDefaultConfig
* fix: random Bluetooth pin for NicheGraphics UIs
While working on #4378 I noticed a funny problem: the blinking system
LED was on during deep-sleep. Initially I thought it was some weird
sleep hw config thing but it turns out it was easier but more pervasive.
We had two different preprocessor symbols which both meant approximately the same
thing LED_INVERTED and LED_STATE_ON (though their polarity was opposite).
Some variant files were setting one, others were setting the other, and others were
setting both. heh.
In the case of the board I was testing (seeed tracker wio 1100) it was only setting one
and the default behavior for the other (for all boards) was incorrect. So I did a grep
and it seems like LED_STATE_ON was used more often, so I kept that one and removed
LED_INVERTED everywhere.
* Make button timing configurable per variant
* Adjust button timing for T-Echo
Easier multi-clicks for features like "toggle backlight" (4x click)
* Fewer full-refreshes for T-Echo display
Disables ghost pixel tracking: T-Echo ghost pixels are fairly faint.
* Refactor EInkDisplay
A lot of variant specific code is merged, with the macros pushed to the respective variant.h files.
"Dynamic Partial" code has been purged, pending a rewrite.
* fix: declare class only if USE_EINK, init all members
* refactor: move macros to platformio.ini
Responds to https://github.com/meshtastic/firmware/pull/3299#issuecomment-1966425926
* fix: EInkDisplay::connect() references old macros
Usage was in a block of variant-specific code, which had been intentionally left untouched.
* fix: remove duplicate macros from variant.h
---------
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
* Isolate esp32 adc logic gymnastics, try simplifying getBattVoltage
* Set sense resolution for pico platforms
* try silencing cppcheck when variant has no battery pin
* ADC channel for esp-idf calibration
* Missed an rp2040 device
---------
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
* Refactor GPS to not probe if pins not defined
* Use Named Constructor to clean up code
* Move doGPSPowerSave to GPS class
* Make sure to set GPS awake on triple-click
* Cleanup and remove dead code
* Rename GPS_PIN_WAKE to GPS_PIN_STANDBY
* Actually put GPS to sleep between fixes
* add GPS_POWER_TOGGLE for heltec-tracker and t-deck
* Change GPS_THREAD_INTERVAL to 200 ms
* More dead code, compiler warnings, and add returns
* Add Number of sats to log output
* Add pgs enable and triple-click config
* Track average GPS fix time to judge low-power time
* Feed PositionModule on GPS fix
* Don't turn off the 3v3_s line on RAK4631
when the rotary is present.
* Add GPS power standbyOnly option
* Delay setting time currentQuality
to avoid strange log message.
* Typos, comments, and remove unused variable
* Short-circuit the setAwake logic on GPS disable
* heltec-tracker 0.3 GPS power saving
* set en_gpio to defined state
* Fix fixed_position logic with GPS disabled
* Don't process GPS serial when not isAwake
* Add quirk for Heltec Tracker GPS powersave
---------
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
Co-authored-by: mverch67 <manuel.verch@gmx.de>
Co-authored-by: Manuel <71137295+mverch67@users.noreply.github.com>
* lay groundwork for a possible future architecture
* switch from feature opt-out to feature opt-in
* lay groundwork for a possible future architecture
* switch from feature opt-out to feature opt-in
* fix USE_RTC in variant.h for rak4631_epaper and t-echo
* ensure Screen.h is not included without configuration.h
Co-authored-by: Peter Lawrence <12226419+majbthrd@users.noreply.github.com>
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
