[OE-core] [PATCH 1/1] toolchain-scripts.bbclass: use MLPREFIX to instead of MULTILIB_VARIANTS

[Richard Purdie]

On Tue, 2015-07-07 at 00:17 -0700, Robert Yang wrote:
> Fixed when enable multilib:
> require conf/multilib.conf
> MULTILIBS = "multilib:lib32"
> DEFAULTTUNE_virtclass-multilib-lib32 = "x86"
> 
> $ bitbake core-image-minimal
> 
> No lib32 package should be built since we don't build
> lib32-core-image-minimal, but a lot them are built:
> lib32-opkg-utils
> lib32-cryptodev-linux
> lib32-sqlite3
> lib32-libtool-cross
> lib32-ncurses
> lib32-db
> lib32-diffutils
> lib32-zlib
> lib32-gcc-cross-i686
> lib32-gcc-cross-initial-i686
> lib32-flex
> lib32-libgcc-initial
> lib32-libffi
> lib32-linux-libc-headers
> lib32-expat
> lib32-openssl
> lib32-glibc
> lib32-binutils-cross-i686
> lib32-gcc-runtime
> 
> This is because MULTILIB_VARIANTS is always avaliable when multlib is
> enabled, use MLPREFIX to fix the problem so that lib32/lib64 packages will
> only be built when build lib32/lib64-core-image-minimal.
> 
> [YOCTO #7961]
> 
> Signed-off-by: Robert Yang <liezhi.yang at windriver.com>
> ---
>  meta/classes/toolchain-scripts.bbclass |    5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/meta/classes/toolchain-scripts.bbclass b/meta/classes/toolchain-scripts.bbclass
> index 9378918..d05a5a3 100644
> --- a/meta/classes/toolchain-scripts.bbclass
> +++ b/meta/classes/toolchain-scripts.bbclass
> @@ -144,8 +144,9 @@ python __anonymous () {
>      deps = ""
>      for dep in (d.getVar('TOOLCHAIN_NEED_CONFIGSITE_CACHE', True) or "").split():
>          deps += " %s:do_populate_sysroot" % dep
> -        for variant in (d.getVar('MULTILIB_VARIANTS', True) or "").split():
> -            clsextend = oe.classextend.ClassExtender(variant, d)
> +        mlprefix = d.getVar('MLPREFIX', True)
> +        if mlprefix:
> +            clsextend = oe.classextend.ClassExtender(mlprefix[:-1], d)
>              newdep = clsextend.extend_name(dep)
>              deps += " %s:do_populate_sysroot" % newdep
>      d.appendVarFlag('do_configure', 'depends', deps)

I'm starting to get mildly annoyed with this. Back in 2014 it was
insisted the correct behaviour here was that the cache artefacts for
*all* multilibs shoul be pulled into the SDK. This was done with:

http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=bb213d8e60746d61d80505487d5f14a4eb45231b

where the code quite clearly iterates MULTILIB_VARIANTS for this reason.
We've therefore gone to some lengths to preserve that.

So which is this code supposed to do? I don't honestly know any more. I
do agree its extremely annoying that the images trigger the multilib
builds though and I think we need to fix that. This class is used from:

populate_sdk_base
meta-ide-support
meta-environment

so perhaps the best way to do this is have a function like:

def toolchain_get_depends(d):
    import oe.classextend
    deps = ""
    for dep in (d.getVar('TOOLCHAIN_NEED_CONFIGSITE_CACHE', True) or "").split():
        deps += " %s:do_populate_sysroot" % dep
        for variant in (d.getVar('MULTILIB_VARIANTS', True) or "").split():
            clsextend = oe.classextend.ClassExtender(variant, d)
            newdep = clsextend.extend_name(dep)
            deps += " %s:do_populate_sysroot" % newdep
    return deps

then meta-ide-support can do:

do_populate_ide_support[depends] += "${@toolchain_get_depends(d)}"

meta-environment can do:

do_generate_content[depends] += "${@toolchain_get_depends(d)}"

and I can't actually see why populate_sdk_base needs the
toolchain-scripts class at all. Perhaps someone can spot why it needs
the inherit or the dependency?

Cheers,

Richard