[U-Boot] [PATCH v3 0/6] Second step towards Kbuild: Descend down like Kbuild

[Masahiro Yamada]

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 at chromium.org>
Cc: Tom Rini <trini at 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

-- 
1.8.1.2