| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
| |
The full suite logs often contains invalid utf8 sequences.
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
| |
One failure point less
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
| |
Was causing problems on CI
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
|
|
|
|
| |
It did not work when using an empty build dir with sstate-cache,
the proper working way is to bitbake build_sysroots.
Also change the way we fetches some of the akautalizr package files
which was broken too.
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
"Autoprovisioning" or "automatic provisioning" is now known as "shared
credential provisioning". "Implicit provisioning" is now known as
"device credential provisioning". "HSM provisioning" was always a
misnomer, so it is now refered to as "device credential provisioning
with an HSM".
This includes a bump of the aktualizr version as well.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
| |
Signed-off-by: Mike Sul <ext-mykhaylo.sul@here.com>
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
|
| |
If ctest is interrupted (e.g. timeout), its partial output will be in
/tmp/aktualizr-ptest.log.tmp
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
|
| |
For some reason it now often fails if you don't briefly pause first, so
just retry it if it fails a few times.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
|
| |
The secondary discovery mechanism is being overhauled and that tool no
longer exists.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
|
| |
lshw is installed in `/usr/sbin` which does not appear when a command is
launched outside a non-login shell.
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
| |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
| |
Should build `imagename` and not 'core-image-minimal' in all cases
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
| |
To allow for more targeted testing
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
| |
The output was moved from stdout to stderr in a recent aktualizr commit.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
|
|
| |
It was a nice idea when we were less confident about bitbaking and
oe-selftest, but at this point, it is redundant and provides no useful
information while adding extraneous time to the already long test suite.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
| |
CLI has changed
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|
|
|
|
|
|
|
|
| |
OSTree needs UsrMove to be implemented, meaning that everything
should be in /usr. OpenEmbedded has a distro feature usrmerge which
implements exactly that, and makes sure that no package install
anything in /bin, /sbin or /lib. Make use of it.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
|
|
|
| |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The qemucommand.py script uses kvm-ok to determine whether KVM
is available. However, kvm-ok is very Ubuntu specific and not
readily available on Fedora.
Use QEMU_USE_KVM variable which is also used in OpenEmbedded
selftests. The variable must contain True to enable KVM for
machines containing x86 in its name, or contain a list of
machines. For meta-updater this makes sure KVM is used for all
tests:
QEMU_USE_KVM = "intel-corei7-64 qemux86-64"
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
|
|
|
| |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
| |
Bump to the latest version as well.
|
|
|
|
|
|
|
|
|
|
|
| |
Do not write directly to the IMAGE_DEPLOY_DIR but to the new
deploy directory IMGDEPLOYDIR.
Also change the image ending to match the image type name (from
otaimg to ota-ext4). This allows to use the automatic symlinking
and output artefact management too.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Work around a bug in oe-selftest with migration and aktualizr-info
interplay by forcing --allow-migrate for now.
|
| |
|
|
|
|
|
| |
This should not be necessary, but after building for the grub test, the
sysroot does not get repopulated as expected for the implicit/HSM tests.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
STAGING_DIR_NATIVE does not get populated as I had previously believed.
Native recipes do not get packaged, so although we want to use a native
executable, we need to get the config files from the
aktualizr-host-tools package built by the aktualizr (target) recipe.
Thanks to Andre McCurdy and Khem Raj for insight on how to get this
right.
|
|
|
|
| |
Note that the HsmTests still sometimes fail
|
|
|
|
|
|
|
|
| |
Move HSM dependencies to aktualizr recipe, since it's aktualizr that can
depend on HSM support, not the provisioning prepartion.
Remove references to systemd in provisioning recipes, since they have
nothing to do with systemd.
|
| |
|
| |
|
|
|
|
|
| |
Not much else from the example configs that is easy to test with
oe-selftest.
|
| |
|
|
|
|
| |
Also add a test for the 'primary-image'
|
|
|
|
| |
Also, timeout communication to the DUT after 60s.
|
| |
|
| |
|
|
|
|
|
|
| |
This required splitting the aktualizr package into several components, in order
to be able to either install the primary 'aktualzr' or the secondary
'aktualizr-secondary'.
|
|
|
|
|
|
|
|
|
|
| |
There's not much we can do, since booting the device via a test is not
practical. Still, it's useful to make sure that we can build the image.
If trying to build the image in an environment designed for qemu, this
takes some trickery to set things up just right.
Also consistently set up layers before setting the MACHINE, since the
sanity checker sometimes complains otherwise.
|
|
|
|
|
|
|
|
| |
The layers required for the grub test are now also only added if not
already present.
This should make the tests runnable from an environment configured for
raspberry pi.
|