swm: package -- ready_for_testing should pass when in later pockets
In a partial respin the kernel is still ready_for_testing of some of it
is in -proposed and other packages are in -updates or -security. Handle
split occupancy when considering our ready_for_testing status.
Signed-off-by: Andy Whitcroft <email address hidden>
swm: boot-testing -- overrides should work even in proposed
When trackers are in the middle of a partial respin we can end up with
the kernel in proposed with boot-testing overriden via tags, but with the
boot-testing task itself pulled back to New or Confirmed. Ensure our
overrides are correctly applied.
Signed-off-by: Andy Whitcroft <email address hidden>
swm: automated-testing -- report SEED-AWOL and similar as manual-intervention required
When the adt-matrix has no seed data for a kernel, it cannot sensibly
determine regression status. In this case it reports the testing as
ongoing, which we calmly report as testing in-progress. This is easy to
miss and assume testing will complete. Detect this state and convert it
into TaskError so that it is front and center.
Signed-off-by: Andy Whitcroft <email address hidden>
swm: wft/verification_testing -- we cannot spam till our parent is ready to spam
We do not wish to spam a derivative until we have a set of bugs for our
parent. Without this parental list we would not be able to elide the
duplicates there and spam the appropriate list.
Signed-off-by: Andy Whitcroft <email address hidden>
Ensure we only short-circuit testing checks when we are actually fully
in -proposed. This allows testing to move to "Failed because the thing
I am testing has dissappeared" even once promotion was marked complete.
Signed-off-by: Andy Whitcroft <email address hidden>
swm: promoter -- Won't Fix is now a valid testing outcome
Now that Won't Fix is a valid testing outcome representing that we will
not perform testing nor monitor for it, make sure we count that as a
completed state.
Signed-off-by: Andy Whitcroft <email address hidden>
When using bug-update to shove things won't-fix I hit some unsupported
cases in the previous heuristic. Switch to a forward mapping from the
actual states to realistic aliases (lower-case single quotes removed
etc).
Signed-off-by: Andy Whitcroft <email address hidden>