[OE-core] [PATCH] kernel-arch.bbclass: Add xtensa and arc into valid_archs table

[Bystricky, Juro]

Thanks. I will add the test for TARGET_OS in a separate patch.

> -----Original Message-----
> From: Khem Raj [mailto:raj.khem at gmail.com]
> Sent: Tuesday, October 18, 2016 10:53 AM
> To: Bystricky, Juro <juro.bystricky at intel.com>
> Cc: openembedded-core at lists.openembedded.org; jurobystricky at hotmail.com
> Subject: Re: [OE-core] [PATCH] kernel-arch.bbclass: Add xtensa and arc into
> valid_archs table
> 
> 
> > On Oct 18, 2016, at 9:13 AM, Bystricky, Juro <juro.bystricky at intel.com>
> wrote:
> >
> > The single purpose of "map_kernel_arch" is to set
> >
> > export ARCH = "arch_xxx"
> >
> > Presently, we must set ARCH even when building baremetal toolchains only,
> > without any need to build kernel. However, in this case we may
> > encounter a bogus error such as "cannot map arch_xxx to a linux kernel
> architecture".
> > The simplest way to avoid this is to place arch_xxx into the table
> > "valid_archs". (That was the main reason for the patch).
> >
> > A better but more complex way would be to modify the code so
> > map_kernel_arch is not called at all when NOT building kernel or kernel
> > modules. This would also allow building baremetal cross-toolchains for
> arches
> > that don't have Linux kernels at all.
> >
> > Another way to consider is perhaps just getting rid of the table
> > "valid_archs" entirely.
> >
> > "map_kernel_arch" maps arch_aaa -> arch_xxx for some cases.
> > If there is no explicit mapping, then it could simply map
> >
> > arch_aaa->arch_aaa
> >
> > resulting in export ARCH="arch_aaa"
> > There would be no error generated. If ARCH is not a valid
> > arch for a Linux kernel, we would encounter a build error at some point.
> > But that, I believe is the same case now if we try to build a kernel
> > with a valid kernel ARCH but not supported actively in OE or extended OE
> ecosystem.
> 
> I think you should add a check for testing TARGET_OS and if its not linux
> then
> ignore the valid kernel arches check.
> 
> >
> >
> >
> >
> >> -----Original Message-----
> >> From: Khem Raj [mailto:raj.khem at gmail.com]
> >> Sent: Monday, October 17, 2016 4:40 PM
> >> To: Bystricky, Juro <juro.bystricky at intel.com>
> >> Cc: openembedded-core at lists.openembedded.org; jurobystricky at hotmail.com
> >> Subject: Re: [OE-core] [PATCH] kernel-arch.bbclass: Add xtensa and arc
> into
> >> valid_archs table
> >>
> >>
> >>> On Oct 11, 2016, at 9:31 AM, Juro Bystricky <juro.bystricky at intel.com>
> >> wrote:
> >>>
> >>> Both "arc" and "xtensa" are valid Linux architectures, add
> >>> them into valid_archs table.
> >>>
> >>> Signed-off-by: Juro Bystricky <juro.bystricky at intel.com>
> >>> ---
> >>> meta/classes/kernel-arch.bbclass | 2 +-
> >>> 1 file changed, 1 insertion(+), 1 deletion(-)
> >>>
> >>> diff --git a/meta/classes/kernel-arch.bbclass b/meta/classes/kernel-
> >> arch.bbclass
> >>> index 8a4bef1..ea976c6 100644
> >>> --- a/meta/classes/kernel-arch.bbclass
> >>> +++ b/meta/classes/kernel-arch.bbclass
> >>> @@ -14,7 +14,7 @@ valid_archs = "alpha cris ia64 \
> >>>               parisc s390  v850 \
> >>>               avr32 blackfin \
> >>>               microblaze \
> >>> -               nios2"
> >>> +               nios2 arc xtensa”
> >>
> >> there parisc etc as well so in essence patch is ok but I wonder if we
> >> should
> >> remove the arches we dont actively support in OE or extended OE
> ecosystem.
> >>
> >>>
> >>> def map_kernel_arch(a, d):
> >>>    import re
> >>> --
> >>> 2.7.4
> >>>
> >>> --
> >>> _______________________________________________
> >>> Openembedded-core mailing list
> >>> Openembedded-core at lists.openembedded.org
> >>> http://lists.openembedded.org/mailman/listinfo/openembedded-core
> >