You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This gives us some additional confidence that all tests that run either:
* pass
* skip, but were expected to given the test hardware/guest configuration
Importantly, not all PHD tests pass on all hardware/guest OS
combinations, and not all PHD tests even *run* on all hardware/guest
combinations. As a bit of a fluke, *Intel* happens to run one additional
test that is skipped on AMD processors, even.
The strategy is to run "skipped" tests and verify they skip because if
guest artifacts are changed unexpectedly we may need to better
understand how the artifact differs from PHD expectations. Or if
Propolis/PHD implementation changes such that the skip is no longer
taken, failing the test instead of silently passing will force us to
notice if that change is wanted, if the test *should* pass, etc.
As an example, the test Alpine image is currently expected to be an
`alpine-virt` on an ISO 9660 (read-only) filesystem. Since this means
the EFI system partition is also read-only,
`guest_can_adjust_boot_order` is not a meaningful test (for now). We
run this test anyway and *expect* a skip, because if it passed then
either:
* the guest Alpine is on a writable root FS; are we missing test
coverage involving read-only guest images?
* we've moved EFI NvVars to out-of-guest storage, and efivarfs writes
succeed.
In either case, the corresponding changes should have some consideration
for how they relate Propolis and guest operations.
0 commit comments