Skip to content

Changelog

What changed in each release of the driver. Entries under Breaking need source changes in code that used the affected API.

[Unreleased]

Changed

  • The repository moved to kokeb-labs-drivers; URLs follow.

[0.1.0] - 2026-11-10

Added

  • Driver. bmp581::BMP581<Port> for the Bosch BMP581 (datasheet BST-BMP581-DS004-13, revision 1.13), on I2C (I2cInterface, I2cPort) or four-wire SPI (SpiInterface, SpiPort), the bus chosen by the port type (BusPort).
  • Start-up. EnsureInitialized(): power-up wait, the SPI selection read, a soft reset, CHIP_ID, NVM ready and INT_STATUS = 0x10 then 0x00 (section 4.3.9), then standby with deep standby off (the reset leaves the chip in deep standby, where writes need not land, section 4.3), the reset marker in NVM_DATA, and the configuration one register per transfer, ODR_CONFIG last, read back in full. Deinitialize(): a soft reset into deep standby, proved by INT_STATUS. Every soft reset is preceded by standby with deep standby off.
  • Measurements. MeasureOnce() (forced mode, polled until it shows as started, at most t_startup, then polled to standby within the datasheet maximum), SetPowerMode() (standby, normal, continuous, through standby, read back; in normal mode OSR_EFF.odr_is_valid polled too, ConfigRejected only at the deadline; with the FIFO on, t_reconf for the FIFO flush before Ok), WaitDataReady(), ReadMeasurement() (one 12-byte read of TEMP_DATA to STATUS); PressurePa(), TemperatureC() and SignExtend24(), exact.
  • FIFO and interrupts. ReadFifo() with a validity per frame, ReadFifoCount(), ReadInterruptStatus(), the latch (GetLatchedInterrupts()), the INT pin, the four sources and the out-of-range window in Config.
  • Values. No frame of a stuck or floating line (reserved registers, reserved bits, all 0x00, all 0xFF), no reset value 0x7F7F7F and nothing outside 30 kPa to 125 kPa or -40 °C to 85 °C is returned as a value. Verify::ReadBack on ReadMeasurement() and MeasureOnce() reads the data again, so a bit flipped inside a data byte is outvoted or refused.
  • Resets. OSR_CONFIG.press_en (always written 1) is the reset sentinel, read with ODR_CONFIG and OSR_EFF and their reserved bits, so no constant byte passes; a marker in NVM_DATA, verified or written in standby before every configuration sequence and counted only until the next measurement from deep standby (whose NVM download may overwrite it, 0x00 0x00 included), shows a reset that a sequence's own writes covered; INT_STATUS.por, the reset value of the data registers, an empty FIFO and an empty or out-of-range FIFO frame run the reset check; on SPI every status, FIFO and raw read runs it before returning (the frame that only selects SPI, section 5.1); evidence is confirmed by a second read; setters read INT_STATUS and the sentinel before they write; a por the sentinel contradicts is decided by the whole configuration and the marker; a check that cannot be read stays pending (DeviceReset, CheckForReset()).
  • Doubt. A failed standby or mode write leaves the configuration in doubt (IsConfigKnown(), StateUnknown); a configuration sequence that fails after its first write leaves the driver uninitialised; standby and Deinitialize() are never refused, also after a failed first start that reached the chip.
  • Configuration. Config with the datasheet reset values, Validate() with table 7 and the INT pad drive rule of section 7.5 (8 and up with open drain only), CheckedConfig, the presets ForcedPreset(), LowPowerPreset(), HighResolutionPreset(), FifoPreset(), and KconfigDefaults().
  • Tests and docs. Host tests over a register model that measures and resets on its own and offers each open reading of the datasheet as a knob (deep standby losing writes, odr_is_valid and mode lags, the SPI selection fill), a fuzz target with the safety rules as invariants and no skips, chip diagrams (pinout, blocks, the I2C and SPI connections, a MISO pull-up on SPI), the start-up and measurement figures, and the host and ESP32 examples.