- Generated by
1.9.8
|
phyphox BLE
Plot your Arduino's data in the phyphox app and control the board from the phone
|
| tier | what | runs where |
|---|---|---|
| T0 | host unit tests of the core (tests/host), conformance of the generated documents against the phyphox file-format spec (tests/conformance), compile matrix of every example on every supported board with a size report | every push, GitHub Actions |
| T1 | the bench: every supported board flashed in turn and driven by a host BLE client (tools/bench) that plays the phone's part — transfer, data, inputs, buttons, events, MTU, disconnects | the lab machine, before a release |
| T2 | the phones: one Android and one iOS phone against every board, for what only the real app can show | the lab, release gate |
.github/workflows/t0.yml runs the tests; .github/workflows/compile.yml compiles every example for every supported board with a size report per pull request (arduino/compile-sketches).
The conformance step needs phyphox-docs checked out beside this repository and lxml; it uses that repository's own validators (RELAX NG, Schematron, validate_experiments.py), with the Schematron's version-gate warnings counted as failures. tests/host/golden/ holds the documents the example configurations must produce byte for byte; regenerate them with tests/host/build/gen_examples tests/host/golden and review the diff before committing.
Flashes tools/bench/benchSketch (a rich experiment that echoes every callback on serial), then plays the phone from this machine's Bluetooth adapter with bleak: discovery, the expected characteristics, the transfer on subscription and on the control write, byte-for-byte equality with the document the board prints on serial, CRC, data rate, one callback per input write, a button press per write, a sensor write fanning out to its channels, an event write, repeated transfers over fresh connections, and the board's own statistics (no aborted transfers). The served document is validated against the phyphox spec. Give each board on the desk its own --name. Results go to tools/bench/results/<label>.json. The script takes the same .bench-lock-<hostname> in the working root as the phyphox-docs lab.
With --custom-xml it flashes tools/bench/benchCustomXml instead, a sketch that serves a hand-written document, and checks that it is served verbatim with the config characteristic cddf1003.
Drives a real phyphox app on an Android phone and on an iPhone against a board running the bench sketch, from this Linux host. Android needs the app's debug and androidTest APKs installed (./gradlew installRegularDebug installRegularDebugAndroidTest in phyphox-android); the connect step is the app's BleCompatConnectTest seam. iOS needs a development build of the app on the phone, developer mode, and sudo pymobiledevice3 remote tunneld running; the app is launched with its -phyphoxBleConnect seam through the tunnel. Everything after the connect goes through the remote-access API: start and stop, buffer writes, button triggers, readings — with the board's serial echo as the other half of each assertion. It checks change-only delivery of every input element, buttons, submit-with-button, the phone's accelerometer streaming to the board, start/pause/clear events, and whether an input set while paused reaches the board (it does not, on either app).
--transfer-loop 20 loads the experiment from the board twenty times per phone over fresh connections and reports the board's transfer statistics (refused and retried packets, abandoned transfers); run it with the phone far from the board or with other boards advertising to see the flow control at work. PHYPHOX_BENCH_SERIAL_LOG=<file> keeps everything the board prints; on the ESP32, building the bench sketch with -DPHYPHOX_BLE_ESP32_TRACE adds the Bluetooth stack's refused confirmations ([phyphoxBLE] conf ... lines).
Flashes every shipped example as it is and lets both real apps load and start it, checking what the example is about: the board's data arriving in the app, or the board's own serial output for the examples that only receive. Run it per board before a release. Every other board on the desk must be off the air — an example flashed earlier keeps advertising its name, and the phone connects to whichever phyphox-Arduino it sees first — so the sweep flashes a do-nothing sketch to the board under test and to every board named in --silence PORT=FQBN, then scans and refuses to start while an example's name is still in the air. --capture DIR keeps each document exactly as the Android phone received it.
sensebox:samd:sb, board package URL https://raw.githubusercontent.com/sensebox/senseBoxMCU-core/master/package_sensebox_index.json): when the Arduino SAMD core is installed as well, arduino-cli upload fails with bossac: extra arguments found because the senseBox recipe passes -U true and arduino-cli picks bossac 1.9.1. The bench scripts upload by hand: open the port at 1200 baud to enter the bootloader, wait for the port to come back, then run bossac 1.7.0 (packages/arduino/tools/bossac/1.7.0-arduino3/bossac -i -d --port=ttyACM0 -U true -i -e -w -v <sketch>.bin -R in the Arduino data directory). Native USB: the bench sketch waits up to 3 s for the host to open the port so the READY line is not lost.PIN_XB1_PWR) with the pins released at begin(); a test sketch must not power-cycle the module with Serial3 open. A module at u-blox factory settings has CTS/RTS flow control on and does not transmit in this socket; the library turns it off with AT+UMRS=115200,2,8,1,1, AT&W, AT+CPWROFF (not yet verified on a factory-fresh module).DSR/CS on the Bee, unconnected on the senseBox) takes u-connectXpress's factory-reset gesture: one second of silence, five high-to-low transitions within one second, one second of silence. Wired to a senseBox GPIO it reboots the module about 2.8 s later, but has not been seen to restore the factory UART settings — treat it as a reset. 1.9.8