Note: Update for v3 is the only patch subjects. The patch body is the same as v2.
I have been just wondering why the U-Boot top Makefile is so dirty.
It is sprinkled with SoC-specific code as follows:
ifneq ($(CONFIG_OMAP_COMMON),) LIBS-y += $(CPUDIR)/omap-common/libomap-common.o endif
ifneq (,$(filter $(SOC), mx25 mx27 mx5 mx6 mx31 mx35 mxs vf610)) LIBS-y += arch/$(ARCH)/imx-common/libimx-common.o endif
ifeq ($(SOC),s5pc1xx) LIBS-y += $(CPUDIR)/s5p-common/libs5p-common.o endif ifeq ($(SOC),exynos) LIBS-y += $(CPUDIR)/s5p-common/libs5p-common.o endif ifneq ($(CONFIG_TEGRA),) LIBS-y += arch/$(ARCH)/cpu/$(SOC)-common/lib$(SOC)-common.o LIBS-y += arch/$(ARCH)/cpu/tegra-common/libcputegra-common.o LIBS-y += $(CPUDIR)/tegra-common/libtegra-common.o endif
And it describes drivers not right under the top directory, which should be essentially cared by drivers/Makefile.
LIBS-y += drivers/bios_emulator/libatibiosemu.o LIBS-y += drivers/block/libblock.o LIBS-$(CONFIG_BOOTCOUNT_LIMIT) += drivers/bootcount/libbootcount.o LIBS-y += drivers/crypto/libcrypto.o LIBS-y += drivers/dma/libdma.o ...
This series adds the directory descending feature like Kbuild.
1/6 tweaks scripts/Makefile.build to support 'obj-y += foo/' syntax. If the build system finds 'foo/' (trailed by a slash), it visits foo directory.
2/6-6/6 demonstarate how we can refactor makefiles with this syntax.
2/6: Move some drivers LIBS from top Makefile to drivers/Makefile 3/6: Move some fs LIBS from top Makefile to fs/Makefile 4/6: Move Tegra specific lines from top Makefile under arch/arm/ 5/6: Move OMAP specific lines from top Makefile under arch/arm/ 6/6: Move Samsung SoC specific lines from top Makefile under arch/arm/
Note: This series uses [PATCH v2]First step towards Kbuild: Use Kbuild-style makefiles (19 patch files). as a prerequisite. Please apply those patches first.
I believe no boards are broken by this refactoring. In order to confirm my claim, I compiled all boards except nds32 and nios2 architrecutures. (I could not get an appropriate prebuilt cross compilers for those two arch.)
For arm, avr32, sandbox, x86, I could perfectly compile all boards without any warnings.
For the other architectures, I could compile some boards and I could not the others. But the SUMMARY result is still the same after applying this series. In addition I compared all the error and warning messages for all boards and they are the same. This means these patches do not introduce any additional errors, warnings.
Cc: Simon Glass sjg@chromium.org Cc: Tom Rini trini@ti.com
Changes for v3: 1/6 - 3/6: No change 4/6 - 6/6: Fix mesleading patch subject (No change in body)
Changes for v2: 1/6: No change 2/6: additionally refactor drivers/pcmcia drivers/rtc too 3/6, 4/6: No change 5/6: rebase on the master 6/6: No change
Masahiro Yamada (6): Makefile: support descending down to subdirectories drivers: move some drivers to drivers/Makefile fs: move some file system to fs/Makefile ARM: tegra: delete Tegra specific code from toplevel Makefile ARM: omap: delete OMAP specific code from toplevel Makefile ARM: s5pc, exynos: delete Samsung ARM SoC specific code from toplevel Makefile
Makefile | 44 +++---------------------------------------- arch/arm/cpu/Makefile | 2 ++ arch/arm/cpu/arm720t/Makefile | 2 ++ arch/arm/cpu/armv7/Makefile | 9 +++++++++ board/LEOX/elpt860/u-boot.lds | 1 - board/tqc/tqm8xx/u-boot.lds | 4 ++-- drivers/Makefile | 15 +++++++++++++++ fs/Makefile | 11 +++++++++++ scripts/Makefile.build | 15 +++++++++++++++ spl/Makefile | 14 +------------- 10 files changed, 60 insertions(+), 57 deletions(-) create mode 100644 arch/arm/cpu/Makefile create mode 100644 drivers/Makefile