Re: [flashrom] [PATCH] Autodetect target processor architecture

[Carl-Daniel Hailfinger]

Am 01.07.2011 04:45 schrieb David Hendricks:
> On Tue, Jun 28, 2011 at 7:26 PM, Carl-Daniel Hailfinger <
> [e-mail]> wrote:
> > Autodetect target processor architecture.
> > Enable architecture dependent compilation of individual sub-drivers for
> > the internal programmer.
> >
> > With this patch, you no longer have to edit the Makefile to compile the
> > internal driver on MIPS/ARM/...
> >
> > TODO: arch.h is not suitable for inclusion in a .c/.h file because of
> > its last line. Any ideas how to change that (move arch.h as "here
> > document" into the Makefile, use other trickery like more #ifdefs)?
> >
> > Signed-off-by: Carl-Daniel Hailfinger
> > Index: flashrom-target_arch/Makefile
> >
> > --- flashrom-target_arch/Makefile       (Revision 1357)
> > +++ flashrom-target_arch/Makefile       (Arbeitskopie)
> > @@ -37,7 +37,10 @@
> >  CFLAGS += -Werror
> >  endif
> >
> > -# FIXME We have to differentiate between host and target arch.
> > +# Determine the destination processor architecture
> > +ARCH = $(strip $(shell LC_ALL=C gcc -E arch.h|grep -v '^#'))
>
> Use $(CC) instead of "gcc" so stuff like "armv7a-cros-linux-gnueabi-gcc"

Ouch, yes! This was a cut-n-paste error. Thanks for spotting it.
> gets handled properly. I still think assigning ARCH conditionally is wise
> since package managers should set it when cross-compiling, though
> using $(CC) should yield the same result.
>
> Aside from that, I was able to cross-compile successfully for ARM (using the
> same CONFIG_* settings as before) after adding this minor patch on top of
> yours:
>
> diff -Nru a/Makefile b/Makefile
> --- a/Makefile  2011-06-30 19:35:53.410049173 -0700
> +++ b/Makefile  2011-06-30 19:35:39.140049283 -0700
> @@ -38,7 +38,7 @@
>  endif
>
> # Determine the destination processor architecture
> -ARCH = $(strip $(shell LC_ALL=C gcc -E arch.h|grep -v '^#'))
> +ARCH = $(strip $(shell LC_ALL=C $(CC) -E arch.h|grep -v '^#'))

Right.
> # FIXME We have to differentiate between host and target OS architecture.
>  OS_ARCH        ?= $(shell uname)
> @@ -228,6 +228,10 @@
>  ifeq ($(ARCH),"x86")
>  PROGRAMMER_OBJS += it87spi.o it85spi.o ichspi.o sb600spi.o wbsio_spi.o
> mcp6x_spi.o
>  else
> +ifeq ($(ARCH),"arm")
> +PROGRAMMER_OBJS += tegra2_spi.o
> +else
> +endif
>  endif
>  NEED_PCI := yes
>  endif
> diff -Nru a/arch.h b/arch.h
> --- a/arch.h    2011-06-30 19:35:50.190944388 -0700
> +++ b/arch.h    2011-06-30 19:35:35.280103884 -0700
> @@ -27,5 +27,7 @@
>  #define __FLASHROM_ARCH__ "mips"
>  #elif defined(__powerpc__) || defined(__powerpc64__) || defined(__ppc__) ||
> defined(__ppc64__)
>  #define __FLASHROM_ARCH__ "ppc"
> +#elif defined(__arm__)
> +#define __FLASHROM_ARCH__ "arm"
>  #endif
>  __FLASHROM_ARCH__

Should I extend my patch to cover ARM as well, or should we perform all
AR related changes in your big ARM suppport patch?

Regards,
Carl-Daniel

[PATCH] Autodetect target processor architecture

[David Hendricks]

On Thu, Jun 30, 2011 at 10:42 PM, Carl-Daniel Hailfinger <
[e-mail]> wrote:
> Should I extend my patch to cover ARM as well, or should we perform all
> AR related changes in your big ARM suppport patch?

I'll add ARM support in the other patches. I only used it here since it's
the only non-x86 platform I have set up for testing at the moment.