Hi Johannes,
On Tue, Nov 8, 2022 at 6:37 AM SCHNEIDER Johannes johannes.schneider@leica-geosystems.com wrote:
gave it another round, and discussed it with Grygorii - he pointed out that it would be better to do the "fifo empty" or even better "byte sent" checks in "our" code instead of adding more complexity to the serial_mxc for handling a buffer/TXFIFO to be properly flushed prior to each of the multiple calls to set_baudrate during startup between spl and uboot; i tried a couple of approaches which would add checks to "pending" to either a new "puts"-method, or "set_bgr" with partial success - the problem remains that long printfs during board_init - which happens before the serial device is actually fully initialized - get cut of near the end... anyway: added complexity -> !KISS
so i'd say go for the revert; which reduces use of the fifo back to just the first byte
Thanks for investigating it. I have sent the revert.
Regards,
Fabio Estevam