Jonathan Nieder [off-list ref] writes:
Thoughts? Maybe it would be enough to check that the intended get
commands are present in the completion list and other git commands are
not, ignoring binaries that might live elsewhere on the $PATH?
Yeah, it would be a robust alternative approach to verify that (1)
output is a superset of what we expect, and (2) they all share the
string we fed to the machinery as the common prefix, I would think.