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
fix: HMP echo detection and CNI dirs in the VM record (#43)
* fix: treat only the echoed command as HMP echo
A password beginning with a double quote makes QEMU answer 'set_password: string expected' and 'Try "help set_password"'; both lines contain set_password, so the substring test took them for echo and the launch reported a password that was never set. The echo is now the first line carrying the command prefix, and only that line.
* fix: remember the CNI dirs in the VM record so rm can release the NIC
vm rm has no --cni-conf-dir/--cni-bin-dir, so a CNI VM created against a non-default CNI installation kept its NIC forever ('nic release incomplete'). run and clone now persist both dirs and every provisioning verb reads them from the record; records without them keep the defaults.
* fix: match the HMP echo once by substring; rm takes the CNI dirs for older records
QEMU's readline echoes the typed command with redraw sequences, so a prefix test would have taken every successful echo for a rejection. The echo is the first line containing 'set_password ' and only that line. Records written before the CNI dirs were persisted resolve flag, then record, then default, so rm --cni-conf-dir/--cni-bin-dir can release their NICs.
* review: flagOr already carries the fallback; docs name the rm recovery flags
// newProvider builds the cocoon network provider: tap/bridge both use the bridge backend (QEMU opens the TAP in the host netns, so it must be a host-side bridge port); cni's TAP lives in a netns.
0 commit comments