ra8_io_swap_demo Directory Reference

Directory dependency graph for ra8_io_swap_demo:

Directories
	src

Detailed Description

The ra8_io capstone (epic #155, #264): a single binary showing every abstraction at once – two block-device backends swapped behind one interface, a VFS round-trip, and stdio retargeted to two different sinks. Where each sibling ra8_io_*_demo binds one backend, this ties the fabric together.

1. Drop-in block-device swap. One backend-agnostic engine runs the identical round-trip over a table of backends. Row 0 is an in-SRAM RAM block device: volatile, erases to zero, no erase-before-write. Row 1 is the on-board Octo-SPI NOR flash: an erase-before-write medium that reads back 0xFF after erase and does a whole-sector read-modify-write on every 512-byte block write. The engine names no peripheral, so two capability-different media are proven interchangeable behind the one vtable; adding a third is one more table row and the engine body never changes.
2. VFS open / read / write / close. Each backend is bridged to ra8_fs, formatted and mounted as FAT12, and registered in the VFS under a short name. The round-trip opens for write, writes, closes, re-opens for read, reads, closes and byte-compares – the streaming file API, not a whole-file shortcut.
3. Targetable stdio, retargeted to two sinks. During the swap phase the engine's progress stream is an in-RAM capture sink, so the same puts / put_u32 calls that would otherwise reach the serial console land in a byte buffer; afterwards the captured bytes are replayed out of the UART sink. One writer, two destinations.

The engine logic, the deterministic payload pattern and the read-back verdict are covered on the host in tests/mocks/src/test_app_ra8_io_swap_demo.c, which runs the same swap over a RAM backend and the register-level xSPI NOR model plus the in-RAM stdio sink, with MC/DC vectors for the compound read-back verdict and the pattern-equality guard.

Blocked on

The xSPI leg. It programs the non-volatile on-board OSPI NOR and has not been captured on the bench; it is emulator-proven and register-identical to ra8_io_xspi_demo, but only a bench run against the real IS25LX512M closes it. The RAM, VFS and stdio legs are fully exercised off-target and on the host.