This GPL-3.0-or-later fixture compiles the repository's actual C stack into an Arduino sketch. The live LabWired catalog used for this demonstration provides the STM32G071, so the hosted run is a G071 protocol demonstration. It is not evidence for the STM32G0B1RE, a particular application BSP, or IAR EWARM.
The fixture injects bytes through a virtual PHY, starts its vectors explicitly in OPERATE, and checks replies with an independent parity-matrix checksum implementation. Its clock advances synthetically between vectors. It does not measure physical UART baud rate, C/Q tA, inter-byte gaps, or sustained throughput.
From the repository root:
bash validation/labwired-protocol/run_host.shExpected final counters are checks=3078 failures=0 replies=400 selectors=400,
followed by PROTOCOL_REGRESSION_PASS. The same code is a normal CTest target,
test_labwired_protocol, and runs in the protocol-rig CI job.
The vectors cover 100 READ and 100 WRITE cycles per COM3 and COM2 configuration, current-cycle PDout-dependent selection of 32-byte prepared inputs, OD1/code4 advertisement, corrupt-request suppression, and a 1024-byte transfer reconstructed from five ISDU records. ISDU checks include the 232-byte record limit, a rejected 233-byte response, 16-bit indices, pending-dispatch Busy, COUNT wrap, and repeated response segments including the final segment. These counts describe executed assertions and vectors, not a certification suite.
python3 validation/labwired-protocol/package.py build/labwired-protocol/source_request.jsonThe generated JSON contains entryPath, source, and a files array suitable
for the hosted firmware source request. Its default --hosted mode fits the
262144-byte source limit by omitting four unused PHY driver headers and
dll_internal.h, and removing Doxygen documentation from included headers.
Header line counts and license notices are preserved. The ten runtime C files
and the fixture remain verbatim. This produces 29 companion files, including
17 stack headers. Use --full to retain all headers and documentation for
review; that larger request may exceed the hosted limit.
Keep the request and resulting firmware tied to the exact Git commit being demonstrated. The script performs local packaging only; compiling, running, and creating a lab use the LabWired tools and their selected board/toolchain configuration.
main.ino reports results over the console. PA6 marks request processing, PA7
marks a virtual reply callback, and PA5 goes high after all assertions pass.
Those pins are execution markers, not C/Q timing probes. A transceiver shown
in a lab netlist is not evidence of analog C/Q or physical-master behavior.
The fixture keeps a 256-by-32-byte prepared buffer table so changing selectors cannot pass by repeating one payload. Products can choose a smaller table and must define their own cross-record snapshot consistency for segmented ISDU.
Physical timing, reset-to-OPERATE negotiation, DMA behavior, analog transceiver behavior, and external-master interoperability require their own evidence.