We recently had a regression in forcing tasks with -f (see bug 5805); since this went undetected for some time and has regressed before, I think we need to have some tests to make sure it continues working. The tests should cover the following: 1) Run bitbake <recipename>, then modify something in ${WORKDIR}/image and bitbake -c package -f <recipename>, check that the files changed in ${WORKDIR}/packages-split, then bitbake <recipename> again and check that do_package_write_(ipk/deb/rpm) re-execute (either by looking at the output, or looking at the log files in ${T} for the recipe) 2) Run bitbake <recipename>, then modify some source file in ${S}, then bitbake -C compile <recipename> and check that do_package and do_populate_sysroot are re-executed (either by looking at the output, or looking at the log files in ${T} for the recipe) We probably ought to have a test for the kernel's -c menuconfig option as well, but that'll require some interaction so I'll enter that as a separate bug.
At the moment we have this test, also automated with oe-selftest, that checks basic functionality of task forcing: https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=163 As explained in bug 5805 this was not enough to catch that issue. We will improve upon what we already have to cover the specified scenario.
Sent patches to oe-core: http://lists.openembedded.org/pipermail/openembedded-core/2015-October/111485.html Updated TC 163 to match TC 2 and added a new one for TC 1 (https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=1354)
Verified