On 12/02/14 09:08, Sachin Kamat wrote: > On 11 February 2014 19:57, Tomi Valkeinen <tomi.valkeinen@xxxxxx> wrote: >> On 11/02/14 14:01, Sachin Kamat wrote: >>> On 10 February 2014 17:48, Tomi Valkeinen <tomi.valkeinen@xxxxxx> wrote: >>>> On 17/01/14 06:32, Sachin Kamat wrote: >>>>> Exynos is now a DT only platform. Hence there is no need >>>>> for an explicit OF dependency. Remove it. >>>> >>>> But the driver still depends on OF, doesn't it? I don't think it's very >>>> good for the driver Kconfig to make presumptions about what ARCH_* >>>> depend on. >>> >>> Depending upon nested dependencies is redundant IMHO. >> >> Well, a driver should be independent of the underlying arch. In >> practice, we have ARCH dependencies, as many of the devices only exist >> on that arch. But I think the drivers should still be designed to be >> arch-independent, as far as possible (omapdss compiles fine on x86). >> >> If the driver depends on OF, it should depend on OF in the Kconfig, no >> matter if the arch also depends on OF. >> >> I don't really care if the EXYNOS_LCD_S6E8AX0 has OF dependency or not, >> but to me this just looks unneeded cleanup, cluttering git logs, and in >> my opinion it's even going to the wrong direction. > > Your argument makes sense. Upon further experimentation I found that even the > Exynos video drivers are ARCH independent (i.e., they build on x86 too) and do > not need to depend on OF for compilation. So I believe, we can remove both these > dependencies. What is your opinion? Indeed, the driver doesn't even seem to call any of_* funcs. Looking at the commit f9b1e013f1c6723798b8f7f5b83297e2837aaef7 (video: exynos_dp: remove non-DT support for Exynos Display Port), it kind of sounds to me that the OF dependency was put there just to prevent non-DT use. I'm fine with removing OF dependency, if the commit description is updated to say that it can be removed as the driver doesn't actually depend on OF at all. As for the ARCH dependency, I think we should keep it. I once removed ARCH_OMAP dependency from omapdss, but Linus wasn't impressed when his kernel compilation started to ask him if he wants to enable OMAPDSS this, OMAPDSS that =). So I think it's fine to keep ARCH dependencies in cases where the driver is clearly used only on some architecture. However, you can use COMPILE_TEST kconfig option if you want to compile test on other archs. I.e.: depends on ARCH_EXYNOS || COMPILE_TEST Tomi
Attachment:
signature.asc
Description: OpenPGP digital signature
