On Mon, Dec 06, 2021 at 11:51:22PM +0000, Joel Stanley wrote: > On Mon, 6 Dec 2021 at 12:16, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote: > > > > > + depends on OF && LITEX > > > > > > I don't like having litex drivers depend on the LITEX kconfig. The > > > symbol is not user visible, and to enable it we need to build in the > > > litex controller driver, which platforms may or may not have. > > > > > > The microwatt platform is an example of a SoC that embeds some LITEX > > > IP, but may or may not be a litex SoC. > > > > I do like the LITEX dependency, as it allows us to gate off a bunch of > > related drivers, and avoid annoying users with questions about them, > > using a single symbol. > > I appreciate your concern. > > We could do this: > > depends on PPC_MICROWATT || LITEX || COMPILE_TEST What about the current OF dependency? Is that covered by COMPILE_TEST, or do we need an additional `depends on` line for it? Thanks, --G > It's unfortunate that kconfig doesn't let us describe the difference > between "this driver requires this symbol" as it won't build and "this > driver is only useful when this symbol is enabled". Traditionally I > write kconfig to represent only the former, whereas you prefer both. > > > Originally, people told me the system controller is always present, > > hence the current logic to have LITEX_SOC_CONTROLLER visible, and > > an invisible LITEX (which is shorter to type) for individual drivers > > to depend on. > > That's another option. I think LITEX either needs to become visible, > become selected by microwatt, or we adopt the proposal I made above > for the litex drivers that the microwatt soc uses.
