New `--rm` flag deletes the VM once the ssh session or `-- cmd` exits, making `vm run` one-shot. Exit code from command mode still propagates correctly. Semantics: - Create fails → no VM to delete, nothing to do. - SSH-wait timeout → VM intentionally kept alive so `vm logs <name>` shows why; the timeout error already pointed users at that. Even with --rm, this path skips delete — a wedged sshd is exactly when you want post-mortem access. - Session/command ends (any exit code, any reason) → VM is deleted via `vm.delete` RPC. Uses a fresh 10s context so Ctrl-C during the session doesn't abort the cleanup. New vmDeleteFunc seam at the top of banger.go alongside the other RPC seams. Two tests cover the happy path (session ends cleanly → delete fires with correct ref) and the skip-on-timeout path (ssh wait errors → delete does NOT fire). README updated with an ephemeral example and a note about the timeout-skip behaviour. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| banger.go | ||
| bangerd.go | ||
| cli_test.go | ||
| make_bundle_test.go | ||