Hi, Richard:

No, it's not about multilib, but just some special requirement from my
company, we have 2 IMX5 machines and we want to separate the sdk install
path for them, so it's decided we append the ${MACHINE} to
REAL_MULTIMACH_TARGET_SYS, but in current poky, to achieve it, I need
extend two recipes, meta-environment.bbappend and populate_sdk_base
inheritor, since I can not override it in a conf file, otherwise some other
"REAL_MULTIMACH_TARGET_SYS = "none"" setting will be overriden as well.

BTW, do you mean this patch: "multilib_header: Update wrapper to handle arm
32/64 bit"?

//Ming Liu


2017-03-30 17:34 GMT+02:00 Richard Purdie <
richard.purdie@linuxfoundation.org>:

> On Thu, 2017-03-30 at 09:57 +0200, liu.ming50@gmail.com wrote:
> > From: Ming Liu <peter.x.liu@external.atlascopco.com>
> >
> > REAL_MULTIMACH_TARGET_SYS is being defined and referred in several
> > recipes, which is redundant and not easy to be overriden, and
> > SDKTARGETSYSROOT is also defined in two recipes.
> >
> > So move their definitions to bitbake.conf.
>
> I have been wondering why you're needing to customise
> REAL_MULTIMACH_TARGET_SYS? Are you having trouble with arm multilibs?
> If so, the patches I just sent out might be useful...
>
> Cheers,
>
> Richard
>