> On Feb 10, 2016, at 10:54 AM, Phil Blundell <pb@pbcl.net> wrote:
> 
> On Wed, 2016-02-10 at 17:43 +0000, Khem Raj wrote:
>> +        cwd=$PWD;cd ${S}
>> +        STAGING_INCDIR=${STAGING_INCDIR} \
>> +        STAGING_LIBDIR=${STAGING_LIBDIR} \
>> +        BUILD_SYS=${BUILD_SYS} HOST_SYS=${HOST_SYS} \
>> +        ${STAGING_BINDIR_NATIVE}/${PYTHON_PN}-native/${PYTHON_PN} ${S}/setup.py build --build-base=${B} ${DISTUTILS_BUILD_ARGS} || \
>> +        bbfatal "${PYTHON_PN} setup.py build execution failed."
>> +        cd $cwd
> 
> Why do you need to save and restore the cwd?  If this is a shell task
> then the "cd" would be local to that shell invocation anyway, right?
> 
> Also, given that you're invoking it by full pathname as
> "${S}/setup.py ..." anyway, do you even need ${S} to be the cwd in the
> first place?  That seems a bit redundant.

yes however there are packages like
http://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/recipes-devtools/python/rpio_0.10.0.bb
and
http://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/recipes-devtools/python/python-rtimu_git.bb

whose build systems were assuming that they were building inside S and were referring to headers include paths
relative to source dirs.

There might be more of these kinds

> 
> p.
> 
>