Skip to content

testing.md

Latest commit

 

History

History
66 lines (49 loc) · 2.92 KB

testing.md

File metadata and controls

66 lines (49 loc) · 2.92 KB
Raw

Testing

The library's logic runs on the host, against fakes, so it is tested without a board:

pio test -e native     # unit tests only
scripts/check.sh       # unit tests, plus every example built for each board

CI runs scripts/check.sh on every push and pull request.

The fake Arduino core

test/fake/Arduino.h is just enough of the Arduino core for the headers to compile and run on the host. Only the native environment puts test/fake/ on the include path; board builds use their real core. Nothing touches hardware, so a test controls everything a sketch would get from outside:

Area Behaviour
Time Stands still until the test moves it (fake::advanceUs), or moves a fixed amount on every micros() call (fake::setAutoAdvance), so busy-wait loops such as homing make progress.
Writes Every digitalWrite is recorded with its time, so tests can count step pulses and check their spacing.
Reads digitalRead returns what the test scripted, per pin or through a handler that can follow the axis (for example, a switch that closes at some step).
Serial Collects everything printed, log output included.

fake::reset() returns to power-on: time 0, no writes, every input reading HIGH (a pulled-up pin), Serial empty. Call it from setUp().

F() strings keep their own type, as on AVR, so overloads resolve exactly as they do on a board.

Known gap: unsigned long is 64 bits on the host and 32 bits on the boards, so the micros() rollover cannot be reproduced. Tests keep time well clear of it.

FakeEngine

test/fake/FakeEngine.h is a StepEngine with no pins and no timing, for testing the logic above the engine (Stepper, homing, the fleet) on its own. It records what it was told and moves in one of two modes:

  • POLLED: one step per service(), which returns true, like PolledEngine.
  • OWN: stepsPerService steps per service(), which returns false, like an engine that steps on its own. The steps come in a burst, the worst case for how late the endstop guard sees a switch.

Each step moves fake time on by the step interval, so endstop debounce runs as it would on a real axis. shaft() is where the motor really is: unlike position(), setPosition() does not reset it, so homing tests can check where step 0 ended up.

Fake FastAccelStepper

test/fake/FastAccelStepper.h stands in for FastAccelStepper 1.4.0, with only the calls FasEngine makes. It copies the library's behaviour where that shapes FasEngine:

  • moveTo() while running is ignored when it would reverse the move, or stop it where it is.
  • stopMove() ramps down over stepsToStop() steps, never past the target.
  • forceStopAndNewPosition() stops at once.
  • Intervals faster than the hardware can step are refused.

Nothing moves on its own: a test calls run(n) to let up to n steps go out.