Fix the host check in a ping handler

Your internal ops tool pings a host the user types, to check it is reachable. The first version ran sh -c "ping -c 1 " + host, so a host like 8.8.8.8; cat /etc/passwd ran a second command. The new version runs ping directly with a list of arguments (argv) and no shell.

An assistant wrote build_ping_argv(host), which decides what that argument list is. It is in the editor. It has a bug. Find it and fix it.

Input and output:

• host: the text the user typed.
• Return ["ping", "-c", "1", host] when the host is valid, with host unchanged. Return [] when it is not, and nothing runs.

Rules the tests check. A host is valid only when all of these hold:

• It is 1 to 253 characters long, dots included.
• Split on ., every part (a label) is 1 to 63 characters long. So no empty labels: a..b and a leading or trailing . are invalid.
• Every character of a label is an ASCII letter, a digit from 0 to 9 or -.
• No label starts or ends with -. This also stops -f, which ping would read as an option, not a host.
• If every label is made only of digits, the host must be an IPv4 address. That means exactly 4 labels, each from 0 to 255, with no leading zero. 0 is fine, 010 is not, because some resolvers read 010 as octal 8.

Examples:

• build_ping_argv("8.8.8.8") returns ["ping", "-c", "1", "8.8.8.8"].
• build_ping_argv("8.8.8.8; cat /etc/passwd") returns [].
• build_ping_argv("internal-db.corp.example.com") returns ["ping", "-c", "1", "internal-db.corp.example.com"].

Then, in a sentence or two: how would you catch this kind of bug in review?