// Unit tests for the RadioLib driver calls Meshtastic's radio interfaces make, one set per chip
// family: SX126x (sx126x_tests.h), SX127x (sx127x_tests.h), SX128x (sx128x_tests.h), LR11x0
// (lr11x0_tests.h) and LR2021 (lr2021_tests.h), against the RadioLib pinned in platformio.ini.
//
// No chip is needed. RecordingHal (RecordingHal.h) logs every SPI transaction and answers with a
// status byte every status-byte family reads as success, plus the few scripted replies the setters
// check before acting (the packet type); for SX127x it keeps a register file instead. That reaches
// every setter and mode command the interfaces use, and the tests pin what reaches the chip: the
// opcode, the frame length, and where the value is fixed, the payload.
//
// The regressions guarded are driver changes that send the wrong bytes, or read or write past a
// buffer, on paths no hardware-free test reached before. Under [env:coverage] (-fsanitize=address)
// an out-of-bounds access aborts the program at the call. The first such bug is jgromes/RadioLib#1864:
// the LR2021 DC-DC workaround passes sizeof(uint32_t) as a word count, overrunning the stack on
// every setRxPath() and LoRa modulation change. The pin here, 7.7.1, predates that workaround, so
// the LR2021 DC-DC tests register only on a RadioLib that has it (see lr2021_tests.h). On one that
// also carries the overrun (7.8.0) they abort, so the LR2021 set and the LR20x0Interface set, whose
// setup runs the same workaround, run last and the rest report first.
//
// CAD (channel activity detection, the listen-before-talk check) is pinned per family: the command
// bytes, and that a command the chip refuses comes back as an error, never CHANNEL_FREE. The LR2021
// refuses SetLoraCadParams exit mode 0x02, the value of RadioLib's RADIOLIB_LR2021_CAD_EXIT_MODE_RX
// (a SetCadParams value; jgromes/RadioLib#1882 adds the LoRa ones), and LR20x0Interface reads the
// error as a free channel, so a node sending that define would transmit over live frames. develop
// sends 0x01 from a local constant and does not, which is what is pinned here. lr2021_tests.h
// models that refusal; lr20x0_interface_tests.h runs LR20x0Interface::isChannelActive() against it,
// so a firmware change that sends a refused exit byte fails here.
//
// Anything that decodes a chip reply (begin(), readData(), getRSSI(), updateFirmware()) needs replies
// scripted per opcode; each family's header lists those under "Grows here".
#include "TestUtil.h"
#include <RadioLib.h>
#include <unity.h>

#include "lr11x0_tests.h"
#include "lr2021_tests.h"
#include "lr20x0_interface_tests.h"
#include "sx126x_tests.h"
#include "sx127x_tests.h"
#include "sx128x_tests.h"

void setUp(void) {}
void tearDown(void) {}

void setup()
{
    initializeTestEnvironment();
    UNITY_BEGIN();
    runSx126xTests();
    runSx127xTests();
    runSx128xTests();
    runLr11x0Tests();
    runLr2021Tests();          // after the rest: its DC-DC tests abort under ASan on a RadioLib with the #1864 overrun
    runLr20x0InterfaceTests(); // its setSpreadingFactor() runs that same DC-DC workaround
    exit(UNITY_END());
}

void loop() {}