On Fri, Jan 19, 2018 at 05:39:32PM +0100, Matthijs van Duin wrote:
On 19 January 2018 at 16:41, Tom Rini trini@konsulko.com wrote:
OK. For v2, if you can convert CONFIG_SYS_NS16550_MEM32 to Kconfig as well I'd appreciate it.
I'm not hugely comfortable doing that, since that would affect even more targets. To be honest, I don't understand why it even exists as a separate var instead of just testing whether CONFIG_SYS_NS16550_REG_SIZE is 4 or -4. Are there actually some crazy targets that allocate 4 bytes per register yet fail if you use 32-bit access?
It's most likely a case that just wasn't cleaned up. If you want to correct things by dropping CONFIG_SYS_NS16550_MEM32 and checking CONFIG_SYS_NS16550_REG_SIZE instead, that's fine.
Would you object hugely to me just implicitly setting CONFIG_SYS_NS16550_MEM32 in ns16550.h for the c6x uart variant? I think it makes a lot of sense to do so, since in that case the driver itself explicitly depends on being able to write a value that doesn't fit in a byte.
Well, it's just SOC_DA8XX that isn't doing it today, and we need to migrate that CONFIG symbol or do away with it.