On Tue, Dec 07, 2010 at 11:11:23PM +0100, Stefan Reinauer wrote:
- Stefan Reinauer stepan@coreboot.org [101207 23:06]:
- Patrick Georgi patrick@georgi-clan.de [101207 16:35]:
Am 06.12.2010 23:45, schrieb Uwe Hermann:
See patch.
src/cpu/Kconfig defines CACHE_AS_RAM to "!ROMCC", so I think these can be dropped completely. Right?
We basically assume it's CAR unless romcc is used.
This seems to make a lot of sense.
Do we need this stuff in Kconfig at all?
There are only very few places in the Makefile that need the distinction:
./src/arch/i386/Makefile.inc:ifeq ($(CONFIG_ROMCC),y) ./src/arch/i386/Makefile.inc:ifeq ($(CONFIG_ROMCC),y) ./src/pc80/Makefile.inc:romstage-$(CONFIG_CACHE_AS_RAM) += serial.c ./src/console/Makefile.inc:romstage-$(CONFIG_CACHE_AS_RAM) += console.c ./src/lib/Makefile.inc:romstage-$(CONFIG_CACHE_AS_RAM) += ramtest.c
And some in the .c files, but I think it would be safe to replace those #if CONFIG_CACHE_AS_RAM with #ifndef __ROMCC__
That would be pretty confusing IMHO, as sometimes "#ifndef __ROMCC__" is used even in CAR boards (where the bootblock is compiled with romcc AFAIK) in some header files (to "disable" function prototypes which romcc doesn't support). The __ROMCC__ #define would then serve two different purposes, kind of like before we splitted away __PRE_RAM__ from __ROMCC__.
Uwe.