Re: [Buildroot] [PATCH v6 02/10] package/gcc-bare-metal: new package

[Thomas Petazzoni via buildroot <buildroot@buildroot.org>]

Hello,

On Fri, 20 Oct 2023 12:42:28 +0100
Neal Frager via buildroot <buildroot@buildroot.org> wrote:

> diff --git a/package/gcc-bare-metal/Config.in.host b/package/gcc-bare-metal/Config.in.host
> new file mode 100644
> index 0000000000..ca7735340c
> --- /dev/null
> +++ b/package/gcc-bare-metal/Config.in.host
> @@ -0,0 +1,12 @@
> +config BR2_PACKAGE_HOST_GCC_BARE_METAL
> +	bool "host gcc-bare-metal"
> +	help
> +	  Build GCC for a bare-metal toolchain
> +
> +if BR2_PACKAGE_HOST_GCC_BARE_METAL
> +
> +config BR2_PACKAGE_HOST_GCC_BARE_METAL_VERSION
> +	string
> +	default "13.2.0"

Both of these options are useless, they should be dropped.


> diff --git a/package/gcc-bare-metal/gcc-bare-metal.hash b/package/gcc-bare-metal/gcc-bare-metal.hash
> new file mode 100644
> index 0000000000..5c5c28fa7e
> --- /dev/null
> +++ b/package/gcc-bare-metal/gcc-bare-metal.hash

Symlink to package/gcc/gcc.hash ?


> diff --git a/package/gcc-bare-metal/gcc-bare-metal.mk b/package/gcc-bare-metal/gcc-bare-metal.mk
> new file mode 100644
> index 0000000000..fe6f931008
> --- /dev/null
> +++ b/package/gcc-bare-metal/gcc-bare-metal.mk
> @@ -0,0 +1,68 @@
> +################################################################################
> +#
> +# gcc-bare-metal
> +#
> +################################################################################
> +
> +HOST_GCC_BARE_METAL_VERSION = \
> +	$(call qstrip,$(BR2_PACKAGE_HOST_GCC_BARE_METAL_VERSION))
> +ifeq ($(HOST_GCC_BARE_METAL_VERSION),)
> +HOST_GCC_BARE_METAL_VERSION = 13.2.0
> +endif

Just harcode the version.

> +
> +HOST_GCC_BARE_METAL_SITE = \
> +	https://ftp.gnu.org/gnu/gcc/gcc-$(HOST_GCC_BARE_METAL_VERSION)
> +HOST_GCC_BARE_METAL_SOURCE = gcc-$(HOST_GCC_BARE_METAL_VERSION).tar.xz
> +
> +HOST_GCC_BARE_METAL_LICENSE = GPL-2.0, GPL-3.0, LGPL-2.1, LGPL-3.0
> +HOST_GCC_BARE_METAL_LICENSE_FILES = COPYING COPYING3 COPYING.LIB COPYING3.LIB
> +
> +HOST_GCC_BARE_METAL_DEPENDENCIES = \
> +	host-binutils-bare-metal \
> +	host-gmp \
> +	host-mpc \
> +	host-mpfr \
> +	host-isl
> +
> +# gcc doesn't support in-tree build, so we create a 'build'
> +# subdirectory in the gcc sources, and build from there.
> +define GCC_BARE_METAL_CONFIGURE_SYMLINK
> +	mkdir -p $(@D)/build
> +	ln -sf ../configure $(@D)/build/configure
> +endef
> +
> +HOST_GCC_BARE_METAL_PRE_CONFIGURE_HOOKS += GCC_BARE_METAL_CONFIGURE_SYMLINK
> +HOST_GCC_BARE_METAL_SUBDIR = build
> +
> +HOST_GCC_BARE_METAL_MAKE_OPTS = \
> +	$(HOST_GCC_COMMON_MAKE_OPTS) \
> +	all-gcc \
> +	all-target-libgcc
> +
> +HOST_GCC_BARE_METAL_INSTALL_OPTS = install-gcc install-target-libgcc
> +
> +GCC_BARE_METAL_ARCH = \
> +	$(call qstrip,$(BR2_PACKAGE_HOST_TOOLCHAIN_BARE_METAL_BUILDROOT_ARCH))

From a patch sequencing perspective, this is a bit weird because
BR2_PACKAGE_HOST_TOOLCHAIN_BARE_METAL_BUILDROOT_ARCH does not even
exist yet. But I don't have a very good suggestion on how to handle
this, so it's fine as is.

Should the variable be named HOST_GCC_BARE_METAL_ARCH, for consistency
with the other variables?

> +HOST_GCC_BARE_METAL_CONF_OPTS = \
> +	--target=$(GCC_BARE_METAL_ARCH)-elf \

I'm wondering if BR2_PACKAGE_HOST_TOOLCHAIN_BARE_METAL_BUILDROOT_ARCH
shouldn't be the full tuple, instead of assuming it's also going to end
in -elf. Whenever we've assumed things like this, it sometimes turned
out to be wrong. So I would suggest that
BR2_PACKAGE_HOST_TOOLCHAIN_BARE_METAL_BUILDROOT_ARCH carry the full
tuple instead.

> +	--disable-initfini_array \
> +	--disable-__cxa_atexit \
> +	--disable-libstdcxx-pch \
> +	--with-newlib \
> +	--disable-threads \
> +	--enable-plugins \
> +	--with-gnu-as \
> +	--disable-libitm \
> +	--without-long-double-128 \
> +	--without-headers \
> +	--enable-languages=c \
> +	--disable-multilib \
> +	--with-gmp=$(HOST_DIR) \
> +	--with-mpc=$(HOST_DIR) \
> +	--with-mpfr=$(HOST_DIR) \
> +	--with-isl=$(HOST_DIR) \
> +	AR_FOR_TARGET=$(O)/host/bin/$(GCC_BARE_METAL_ARCH)-elf-ar \
> +	RANLIB_FOR_TARGET=$(O)/host/bin/$(GCC_BARE_METAL_ARCH)-elf-ranlib

$(HOST_DIR) instead of $(O)/host/bin

Thanks!

Thomas
-- 
Thomas Petazzoni, co-owner and CEO, Bootlin
Embedded Linux and Kernel engineering and training
https://bootlin.com
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot