The idea is to start with a clean workspace in the self hosted runner
The previous job could left the workspace in a bad state as it is
happening when it adds a submodule.
This reverts commit 5296059c30.
Now that the v1 tag has been updated to point to 1.0.9, we can go back to
tracking v1 instead of tracking specifically v1.0.9, which will prevent us from
getting updates in the future if there is i.e. a 1.0.10 or something.
Signed-off-by: Ian Johnson <ian.johnson@canonical.com>
This is necessary to avoid the failure to refresh LXD - the images already have
LXD and so can build the snap, they just can't refresh LXD due to store
outages sometimes, so this new version helps with that when it happens and
keeps out CI moving along to the extent that it can.
Signed-off-by: Ian Johnson <ian.johnson@canonical.com>
Uploading coverage to codecov fails on occasion. It is not really useful to
block the whole test run because of this.
Signed-off-by: Maciej Borzecki <maciej.zenon.borzecki@canonical.com>
We want the spread tests to also run on pushes to release/** branches, so the
operator here should just be || otherwise these jobs will not run on pushes.
Thanks to Alberto for pointing this out to me.
Furthermore, we don't want to run the snap-builds job on anything other than
pull requests, since we already have Launchpad snap build recipes setup for the
release and master branches and this would just be duplicating that effort and
use up artifact storage space.
Signed-off-by: Ian Johnson <ian.johnson@canonical.com>
