buildroot: Makefile: Allow target board makefile fill in default target_skeleton path.

If the target skeleton path is empty in config, don't change variable TARGET_SKELETON in buildroot Makefile. Allow the target board makefile fill in a default path.

Le Wed, 8 Aug 2012 14:57:19 +0800,
Sonic Zhang <[email protected]> a ?crit :

Quoted 3 lines

If the target skeleton path is empty in config, don't change variable TARGET_SKELETON in buildroot Makefile. Allow the target board makefile fill in a default path.

The concept of "target board makefile" does not exist. It used to exist in earlier versions of Buildroot, but it no longer exists in newer versions. Thus, I have the feeling that rather than fixing this, it is the way your boards are supported that should be modified according to the newer Buildroot good practices.

The practices in terms of board support are:

* One defconfig file per board in configs/, generally named <vendor>_<boardname>_defconfig. Those files must be generated with 'make savedefconfig'

* Utility files (kernel patches, kernel configuration files and other stuff), should be stored in board/<vendor>/<boardname>/. Kernel configuration files must be generated with 'make savedefconfig' as well.

* Using a different skeleton from the default one provided by Buildroot is really *not* recommended. Instead, if anything needs to be tweaked in the filesystem, use the post-build script for that effect. This allows to avoid duplicating the skeleton for each and every board, causing a maintenance nightmare when we want to change something in this skeleton.

If "target board makefile" doesn't exit, where should we add arch specific makefile targets and compile/link flags? Where should arch specific Config options be put? Take blackfin specific makefile as an example bellow.

Makefile.in
----------------
TARGET_CFLAGS += -D__NOMMU__ -D__uClinux__ -D_GNU_SOURCE

ifeq ($(BR2_ABI_FLAT),y)
TARGET_LDFLAGS += -Wl,-elf2flt
endif
ifeq ($(BR2_BFIN_SHARED_FLAT), y)
TARGET_LDFLAGS += -mid-shared-library -mshared-library-id=0 TARGET_CFLAGS += -mid-shared-library -mshared-library-id=0 TARGET_CXXFLAGS += -mid-shared-library -mshared-library-id=0 endif

CROSS_COMPILE_SHARED_ELF ?= bfin-linux-uclibc-

romfs.shared.libs.elf:

        set -e; \
        t=`$(CROSS_COMPILE_SHARED_ELF)gcc $(CPUFLAGS)
-print-file-name=libc.a`; \
        t=`dirname $$t`/../..; \
        for i in $$t/lib/*so*; do \
                i=`readlink -f "$$i"`; \
                soname=`$(CROSS_COMPILE_SHARED_ELF)readelf -d "$$i" |
sed -n '/(SONAME)/s:.*[[]\(.*\)[]].*:\1:p'`; \
                $(INSTALL) -D $$i $(TARGET_DIR)/lib/$$soname; \
        done

CROSS_COMPILE_SHARED_FLAT ?= bfin-uclinux-
romfs.shared.libs.flat:

        set -e; \
        t=`$(CROSS_COMPILE_SHARED_FLAT)gcc $(CPUFLAGS)
-mid-shared-library -print-file-name=libc`; \
        if [ -f $$t -a ! -h $$t ] ; then \
                $(INSTALL) -D $$t $(TARGET_DIR)/lib/lib1.so; \
        fi

Config.in
----------------------------------------------------
config BR2_TARGET_ANALOGDEVICES_INSTALL_ELF_SHARED
bool "Install ELF shared libraries" if !BR2_BFIN_FDPIC default y

config BR2_TARGET_ANALOGDEVICES_INSTALL_FLAT_SHARED
bool "Install FLAT shared libraries" if !BR2_BFIN_SHARED_FLAT default y

Blackfin Linux distribution has a different file system layout from current default one in buildroot. And we have no plan to deviate from current one used for years. Anyway without this patch, we can still set a custom skeleton for blackfin.

Hello,

Le Thu, 9 Aug 2012 18:26:54 +0800,
Sonic Zhang <[email protected]> a ?crit :

Quoted 4 lines

If "target board makefile" doesn't exit, where should we add arch specific makefile targets and compile/link flags? Where should arch specific Config options be put? Take blackfin specific makefile as an example bellow.

There is no general rule for this, it depends on each thing you want to add. For things that must be added globally, package/Makefile.in is usually the current place (though this file would probably need some cleanup, but this is a different topic).

Quoted 3 lines

Makefile.in
----------------
TARGET_CFLAGS += -D__NOMMU__ -D__uClinux__ -D_GNU_SOURCE

I don't think the -D_GNU_SOURCE should be passed globally to all packages. Each package should define it if it is needed to build this package.

Also, before globally defining __NOMMU__ and __uClinux__, I'd like to understand why they are needed. I think __NOMMU__ is generally used to test whether fork() or vfork() should be used, but this should probably be tested using autoconf tests when possible.

That said, if such flags are indeed necessary, package/Makefile.in is a good place for that.

Quoted 9 lines

ifeq ($(BR2_ABI_FLAT),y)
TARGET_LDFLAGS += -Wl,-elf2flt
endif
ifeq ($(BR2_BFIN_SHARED_FLAT), y)
TARGET_LDFLAGS += -mid-shared-library -mshared-library-id=0 TARGET_CFLAGS += -mid-shared-library -mshared-library-id=0 TARGET_CXXFLAGS += -mid-shared-library -mshared-library-id=0 endif

Same thing for these flags, package/Makefile.in.

Quoted 1 line

CROSS_COMPILE_SHARED_ELF ?= bfin-linux-uclibc-

No, this should use the existing variables to get the cross-compiler location.

Quoted 12 lines

romfs.shared.libs.elf:

        set -e; \
        t=`$(CROSS_COMPILE_SHARED_ELF)gcc $(CPUFLAGS)
-print-file-name=libc.a`; \
        t=`dirname $$t`/../..; \
        for i in $$t/lib/*so*; do \
                i=`readlink -f "$$i"`; \
                soname=`$(CROSS_COMPILE_SHARED_ELF)readelf -d "$$i" |
sed -n '/(SONAME)/s:.*[[]\(.*\)[]].*:\1:p'`; \
                $(INSTALL) -D $$i $(TARGET_DIR)/lib/$$soname; \
        done

Why is this needed? How is this romfs.shared.libs.elf target being used?

Quoted 9 lines

CROSS_COMPILE_SHARED_FLAT ?= bfin-uclinux-
romfs.shared.libs.flat:

        set -e; \
        t=`$(CROSS_COMPILE_SHARED_FLAT)gcc $(CPUFLAGS)
-mid-shared-library -print-file-name=libc`; \
        if [ -f $$t -a ! -h $$t ] ; then \
                $(INSTALL) -D $$t $(TARGET_DIR)/lib/lib1.so; \
        fi

Ditto.

Quoted 7 lines

Config.in
----------------------------------------------------
config BR2_TARGET_ANALOGDEVICES_INSTALL_ELF_SHARED
bool "Install ELF shared libraries" if !BR2_BFIN_FDPIC default y

Quoted 3 lines

config BR2_TARGET_ANALOGDEVICES_INSTALL_FLAT_SHARED
bool "Install FLAT shared libraries" if !BR2_BFIN_SHARED_FLAT default y

Any reason to have these in addition to the existing selection of ABI? What are the use cases for these?

Quoted 4 lines

Blackfin Linux distribution has a different file system layout from current default one in buildroot. And we have no plan to deviate from current one used for years. Anyway without this patch, we can still set a custom skeleton for blackfin.

How does the filesystem layout differs? Depending on how it differs, having a separate skeleton may or may not be the right solution.