On 30/08/2022 13:35, Mark Brown wrote:
On Sat, Aug 27, 2022 at 12:02:39PM +0300, Krzysztof Kozlowski wrote:
True, it's not the same build coverage need as subsystem/interface dependencies, but still:
- Drivers can be built independently, so why not?
- There is no drawback in allowing to building them independently
(compile test). 3. The parent MFD device could also depend on something (RK817 and MFD_LOCHNAGAR depends on I2C and OF; CPCAP on SPI), so it's not always the case of just enabling parent. IOW, you can compile test codecs without I2C and SPI which is valuable itself.
I find the last reason actually quite valid and applicable here. If you wish I can add it to the commit msg.
You've already found one case where there's a dependency being pulled in by the parent, I suspect there's more. There's a tradeoff between getting coverage and having to deal with the noise from randconfigs.
Yeah, that was my mistake because did not spot it has "select REGMAP_I2C". For such cases the benefit is indeed gone.
Best regards, Krzysztof