On 06/05/2012 03:20 PM, Allen Martin wrote:
In preparation for splitting out the armv4t code from tegra20, move the tegra20 SoC code to arch/arm/cpu/tegra20-common. This code will be compiled armv4t for the arm7tdmi and armv7 for the cortex A9.
diff --git a/arch/arm/cpu/armv7/tegra20/Makefile b/arch/arm/cpu/tegra20-common/Makefile
# The AVP is ARMv4T architecture so we must use special compiler # flags for any startup files it might use. -CFLAGS_arch/arm/cpu/armv7/tegra2/ap20.o += -march=armv4t -CFLAGS_arch/arm/cpu/armv7/tegra2/clock.o += -march=armv4t -CFLAGS_arch/arm/cpu/armv7/tegra2/warmboot_avp.o += -march=armv4t +CFLAGS_arch/arm/cpu/tegra2-common/ap20.o += -march=armv4t +CFLAGS_arch/arm/cpu/tegra2-common/clock.o += -march=armv4t
Don't you still need the CFLAGS override for warmboot_avp.o here too? I assume that a later patch will just remove it, but it'd be nice if each patch along the way built and booted OK to, so that "git bisect" works through this series.
-COBJS := $(COBJS-y) -SRCS := $(SOBJS:.o=.S) $(COBJS:.o=.c) -OBJS := $(addprefix $(obj),$(COBJS) $(SOBJS)) +SRCS := $(SOBJS:.o=.S) $(COBJS-y:.o=.c) +OBJS := $(addprefix $(obj),$(SOBJS) $(COBJS-y))
At least in this patch, I don't think SOBJS is set. Does using it here cause build issues?
diff --git a/spl/Makefile b/spl/Makefile
+ifneq ($(CONFIG_TEGRA2),) +LIBS-y += arch/$(ARCH)/cpu/$(SOC)-common/lib$(SOC)-common.o +endif
In another file in this patch, that first line was the non-inverted test:
+ifeq ($(SOC),tegra20)
That seems simpler. Can we use the same here?