Re: [yocto][PATCH 1/1] ptest: user-specified run-ptest script location

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2015-09-29 at 06:59 +0000, Ray, Ian (GE Healthcare) wrote:
> On 28 September 2015, Tudor Florea wrote:
> > > On 28 September 2015, Ross Burton wrote:
> > > > Surely if an upstream is under your control and is ptest-aware, it can also
> > > > install run-ptest into the right place too?
> > > The run-ptest script is currently required to be in WORKDIR.
> > Actually run-ptest is currently required to be present in
> > /usr/lib/<package>/ptest. If a package is "ptest-aware", that  package
> > should contain a  install-ptest make target which will take care of
> > installing the run-ptest file.
> 
> In summary, the options seem to be as follows:
> 
> # option 1 -- package includes run-ptest, recipe points to it
> PTEST_RUN_SCRIPT_PATH = "path/to/run-ptest"
> 
> # option 2 -- recipe copies run-ptest to WORKDIR
> # This leverages meta/classes/ptest.bbclass do_install_ptest_base.
> do_compile_ptest () {
>         oe_runmake buildtest-TESTS
>         cp ${S}/path/to/run-ptest ${WORKDIR}
> }
> 
> # option 3 -- package installs run-ptest
> # The recipe requires custom do_install_ptest because there is no
> # run-ptest in WORKDIR so meta/classes/ptest.bbclass does not call
> # do_install_ptest_base.
> do_install () {
>         oe_runmake DESTDIR=${D} install
> 
>         # Copied from do_install_ptest_base.
>         install -D ${S}/ptest/run-ptest ${D}${PTEST_PATH}/run-ptest
>         oe_runmake DESTDIR=${D}${PTEST_PATH} install-ptest
>         chown -R root:root ${D}${PTEST_PATH}
> 
>         # Hack alert!  had to comment out do_install_ptest_base[cleandirs] in
>         # meta/classes/ptest.bbclass to test this option.
> }
> 
> My inexperience shows here with option #3 which attempted to address the comments
> by Ross and Tudor, but which does not actually work (tested on Fido branch).
> 
> Maybe there is an option #4, where the recipe does *not* inherit ptest, but that would
> place a significant burden on the recipe author (detecting PTEST_ENABLED, adding
> tasks, etc).  No doubt there is an easier way.  Sorry for being such a noob!

I think people were suggesting option 4 where your "make install" puts
the file into the correct place in DESTDIR when the install is run.
There is then nothing the recipe needs to do.

Cheers,

Richard