> It's weekend, and not many people are active now. Let's wait for a > few more days if anybody shows up with arguments to support your > patches.
In my opinion, the serial device should be the first device initialized and utilized before all other devices where that is possible. For example, it would be very useful to have serial output while debugging RAM initialization. An early serial port prior to U-Boot relocating itself would be very useful for porting U-Boot to new boards. An early debug serial port would be a good candidate for mainline U-Boot: CONFIG_DEBUG_EARLY_SERIAL has been implemented in a few forks of U-Boot. I also favor an early serial debug console as advocated by this patch, but perhaps its code should be self contained. I'm not sure it is a good idea to modify some commands so they can be run before RAM relocation. If such a thing can be done simply, it should be done for all commands. Is the patch submitted clean enough? No, the implementation could be improved with some effort; a few #if - #else - #endif clauses could be eliminated for example. Would an easier to maintain implementation even be feasible? I agree that a JTAG debugger is extremely useful for board bring-up, but they are very expensive. I'm strongly in favor of having some early serial functionality whether in the form of simply being able to send output to the serial port or having a semi-functional to fully functional shell prior to RAM relocation. Sincerely, Ken Fuchs _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot