diff options
| author | Bruce Ashfield <bruce.ashfield@gmail.com> | 2026-06-15 14:13:10 +0000 |
|---|---|---|
| committer | Bruce Ashfield <bruce.ashfield@gmail.com> | 2026-06-15 14:13:10 +0000 |
| commit | 7acf8f44150841091fa7bcb2db46fe8e937916d7 (patch) | |
| tree | 4fcac4020872ea304f01a744bc8cfb47fbd4f36b /dynamic-layers/raspberrypi | |
| parent | cfa93b77c8ee3ec2e96861ce82e2f8a458689067 (diff) | |
| download | meta-virtualization-7acf8f44150841091fa7bcb2db46fe8e937916d7.tar.gz | |
tests: use CONTAINER_PROFILE and surface runqemu errors in lxc fixture
Two correctness improvements to the lxc runtime test fixture.
1. Switch the build from CONTAINER_IMAGE_HOST_EXTRA_INSTALL = "lxc"
to CONTAINER_PROFILE = "lxc".
The original fixture set CONTAINER_IMAGE_HOST_EXTRA_INSTALL to
force lxc into the rootfs without depending on a profile fragment,
because at the time no lxc profile existed. Now that
container-host-lxc.conf is in tree, the test should exercise the
same path real users take — which is the entire point of a
runtime test. Using the profile also pulls in lxc-networking and
lxc-templates automatically (via packagegroup-lxc), where the
original "just install lxc" approach silently left a half-broken
LXC the test couldn't usefully verify.
2. Capture the runqemu output to a file and surface it in the
pexpect.EOF failure message.
pexpect's default 100-character "before" buffer truncates
runqemu's actual error to the trailing fragment, which is rarely
enough to diagnose the failure. Boot failures during fixture
setup now print the last 60 lines of the runqemu log into the
pytest failure message and leave the full log at
/tmp/test_lxc_runtime-runqemu.log for follow-up. When the lxc
image still wasn't booting cleanly, this is what made the
"lxc-net.service could not be found" / bridge-missing root cause
visible without re-instrumenting the fixture.
Signed-off-by: Bruce Ashfield <bruce.ashfield@gmail.com>
Diffstat (limited to 'dynamic-layers/raspberrypi')
0 files changed, 0 insertions, 0 deletions
