tests: fix lxc lifecycle suite hang and tighten download-skip predicateHEADmaster

Two diagnostics-quality fixes to test_lxc_runtime.py uncovered by
running the full suite after lxc-networking landed in the lxc profile:

1. test_start used `exit $rc` after the lxc-start invocation to
   propagate the exit code through run_cmd's marker harness. With
   lxc-networking now installed, lxc-start succeeds (rc=0) and the
   `exit 0` terminates the guest's interactive login shell, which
   triggers getty respawn. Every subsequent lifecycle test then
   times out waiting for its marker because the new getty is sitting
   at the login prompt with no auto-login.

   Wrap the cleanup in `( exit $LXC_RC )` so only a subshell exits;
   the outer login shell stays alive for the next test. While here,
   pre-flight checks for lxc-net.service and the lxcbr0 link surface
   in the failure message — the "ABORTING" path that bit us during
   bring-up gave no hint which precondition was missing.

2. test_create_alpine_via_download treated any non-zero lxc-create
   rc as "network unreachable" and skipped. That swallowed a stale
   "Container already exists" left over from an interrupted previous
   run (or from manual poking), silently turning a real test-harness
   bug into a green skip.

   Pre-clean with `lxc-destroy --force` (best-effort, ignores rc),
   then match a list of known network-error markers explicitly. Any
   other failure is a real bug and fails the test with the full
   lxc-create output instead of hiding behind pytest.skip.

Signed-off-by: Bruce Ashfield <bruce.ashfield@gmail.com>

0 files changed, 0 insertions, 0 deletions