[OE-core] [PATCH 3/6] distro: Add defaultsetup.conf, a set of default configuration providing sane overrridable default for commonly used options

[Khem Raj]

On (11/05/11 10:37), Richard Purdie wrote:
> On Tue, 2011-05-10 at 16:31 +0200, Koen Kooi wrote:
> > Op 10 mei 2011, om 16:00 heeft Richard Purdie het volgende geschreven:
> > 
> > > From: Richard Purdie <richard.purdie at linuxfoundation.org>
> > > 
> > > The intent is to allow distros to share common core config but still allow
> > > customisations. The core should work with no distro set but users
> > > can still customise in any ways needed.
> > > 
> > > Signed-off-by: Richard Purdie <richard.purdie at linuxfoundation.org>
> > > ---
> > > meta/conf/bitbake.conf                             |    4 +-
> > > meta/conf/distro/defaultsetup.conf                 |   23 ++++++++++
> > > .../include/{poky-eglibc.inc => tclibc-eglibc.inc} |    6 ++-
> > > .../include/{poky-glibc.inc => tclibc-glibc.inc}   |    6 ++-
> > > .../include/{poky-uclibc.inc => tclibc-uclibc.inc} |    4 ++
> > > .../{poky-default.inc => tcmode-default.inc}       |   13 +++---
> > > meta/conf/distro/poky.conf                         |   46 +++++--------------
> > > 7 files changed, 54 insertions(+), 48 deletions(-)
> > > create mode 100644 meta/conf/distro/defaultsetup.conf
> > > rename meta/conf/distro/include/{poky-eglibc.inc => tclibc-eglibc.inc} (92%)
> > > rename meta/conf/distro/include/{poky-glibc.inc => tclibc-glibc.inc} (91%)
> > > rename meta/conf/distro/include/{poky-uclibc.inc => tclibc-uclibc.inc} (86%)
> > > rename meta/conf/distro/include/{poky-default.inc => tcmode-default.inc} (85%)
> > > 
> > > diff --git a/meta/conf/bitbake.conf b/meta/conf/bitbake.conf
> > > index d843e70..ce74a9b 100644
> > > --- a/meta/conf/bitbake.conf
> > > +++ b/meta/conf/bitbake.conf
> > > @@ -627,9 +627,7 @@ include conf/build/${BUILD_SYS}.conf
> > > include conf/target/${TARGET_SYS}.conf
> > > include conf/machine/${MACHINE}.conf
> > > include conf/machine-sdk/${SDKMACHINE}.conf
> > > -include conf/distro/include/default-providers.inc
> > > -include conf/distro/include/default-versions.inc
> > > -include conf/distro/include/world-broken.inc
> > > +include conf/distro/defaultsetup.conf
> > > include conf/distro/${DISTRO}.conf
> > 
> > Please include it after $DISTRO, otherwise ?= in $DISTRO won't work as intended.
> 
> Fixed, thanks.
> 
> 
> > > include conf/documentation.conf
> > > require conf/sanity.conf
> > > diff --git a/meta/conf/distro/defaultsetup.conf b/meta/conf/distro/defaultsetup.conf
> > > new file mode 100644
> > > index 0000000..af5ef7b
> > > --- /dev/null
> > > +++ b/meta/conf/distro/defaultsetup.conf
> > > @@ -0,0 +1,23 @@
> > > +include conf/distro/include/default-providers.inc
> > > +include conf/distro/include/default-versions.inc
> > > +include conf/distro/include/world-broken.inc
> > > +
> > > +TARGET_VENDOR ?= "-oecore"
> > > +
> > > +TARGET_FPU_arm ?= "soft"
> > > +TARGET_FPU_armeb ?= "soft"
> > 
> > Something more elaborate would be better, this is what angstrom has:
> > 
> > conf/distro/include/angstrom.inc:TARGET_FPU_arm = "soft"
> > conf/distro/include/angstrom.inc:TARGET_FPU_armeb = "soft"
> > conf/distro/include/angstrom.inc:TARGET_FPU_ixp4xx = "soft"
> > conf/distro/include/angstrom.inc:TARGET_FPU_ppc405 = "soft"
> > conf/distro/include/angstrom.inc:TARGET_FPU_armv6 = "hard"
> > conf/distro/include/angstrom.inc:TARGET_FPU_armv6-novfp = "soft"
> > conf/distro/include/angstrom.inc:TARGET_FPU_armv7a = "hard"
> > conf/distro/include/angstrom.inc:TARGET_FPU_ppc603e = "hard"
> > 
> > you really want to use hard (don't confuse it with hardfp) on armv7a.
> 
> Agreed but we have a problem here as the default OVERRIDES in OECore
> don't include BASE_PACKAGE_ARCH. I'm kind of reluctant to do so too as
> the number of cases we need this are small and the strings aren't very
> unique. I'b feel happier if the override was something like
> parch-${BASE_PACKAGE_ARCH} as you could at least easily spot where it
> was being used.
> 
> I appreciate we need to do something here, its just a question of what.
> 
> > > --- a/meta/conf/distro/include/poky-eglibc.inc
> > > +++ b/meta/conf/distro/include/tclibc-eglibc.inc
> > > @@ -2,6 +2,10 @@
> > > # eglibc specific configuration
> > > #
> > > 
> > > +TARGET_OS = "linux"
> > > +TARGET_OS_arm = "linux-gnueabi"
> > > +TARGET_OS_armeb = "linux-gnueabi"
> > 
> > Maybe something like angstroms version:
> > 
> > TARGET_OS = "linux"
> > TARGET_OS .= "${@['','-gnueabi'][bb.data.getVar('TARGET_ARCH',d,1) in ['arm', 'armeb']]}"
> > TARGET_OS .= "${@['','-gnuspe'][bb.data.getVar('BASE_PACKAGE_ARCH',d,1) in ['ppce500', 'ppce500v2']]}"
> 
> I decided the _arm was more readable. Once we address the OVERRIDE
> issue, we can add the spe bits.
> 
> > or the sane-toolchain version:
> > 
> > def compute_os_portion_of_target_triplet (d):
> >     import bb
> >     arm_eabi_unsupported_arches = "armv1 armv2 armv3"
> >     ppc_spe_supporting_arches = "ppce500v2 ppce500"
> >     gnu_suffix = ""
> >     if bb.data.getVar('LIBC', d, 1) == "uclibc":
> >         libc_suffix = "uclibc"
> >     else:
> >         libc_suffix = ""
> > 
> >     if bb.data.getVar('TARGET_ARCH',d,1) in ['bfin']:
> >         if libc_suffix is not "uclibc":
> >             bb.fatal("bfin is not supported on glibc/eglibc. Please choose uclibc")
> >         else:
> >             os_suffix = "uclinux"
> >     else:
> >         os_suffix = "linux"
> >     bparch = bb.data.getVar('BASE_PACKAGE_ARCH', d,1)
> > 
> >     if bb.data.getVar('DISTRO_FEATURES',d,1) is not None and \
> >     bparch is not None:
> >         if 'eabi' in bb.data.getVar('DISTRO_FEATURES',d,1).split() and \
> >         bb.data.getVar('TARGET_ARCH', d, 1) in [ 'arm', 'armeb' ]:
> >             if bparch in arm_eabi_unsupported_arches.split():
> >                 bb.fatal("DISTRO requested EABI but selected machine does not support EABI")
> >                 abi_suffix = ""
> >             else:
> >                 if libc_suffix is not "uclibc":
> >                     gnu_suffix = "gnu"
> >                 abi_suffix = "eabi"
> >         elif bparch in ppc_spe_supporting_arches.split():
> >             if libc_suffix is not "uclibc":
> >                 gnu_suffix = "gnu"
> >             abi_suffix = "spe"
> >         else:
> >             abi_suffix = ""
> >     else:
> >         bb.note("DISTRO_FEATURES is not set abi suffix not set")
> >         abi_suffix = ""
> > 
> >     if bb.data.getVar('TOOLCHAIN_BRAND', d, 1) is not None and \
> >     "csl" in bb.data.getVar('TOOLCHAIN_BRAND', d, 1):
> >         gnu_suffix = "gnu"
> > 
> >     if libc_suffix + gnu_suffix + abi_suffix is not "":
> >         return os_suffix + "-" + libc_suffix + gnu_suffix + abi_suffix
> >     else:
> >         return os_suffix
> 
> My eyes! We must be able to something more readable than that. Do we
> even support armv1?


oe had lot of combinations to consider e.g.
oabi/eabi/spe/uclinux/uclibc/external-toolchain when it came to
compute the TARGET_OS and while not so readable may be it did address
all of the combinations and permutations well. oe-core may not *yet* worry for many of them
and some like oabi can be dropped but eventually it might need something
handy to compute TARGET_OS may be a map would be more readable.

-Khem