A couple dozen upgrade and skew e2e test suites have been broken for exactly the same amount of time (1 month, 23 days). I looked at 2, and they were both broken due to:

/workspace/sh2ju.sh: line 125: bc: command not found

This problem was introduced by #27825 in June, and apparently wasn't entirely fixed by #30556.

We lost 2 months of upgrade test coverage due to this. What can we do to notice such failures earlier?

Visible here, if you scroll all the way to the right:
https://k8s-testgrid.appspot.com/google-upgrade#gke-1.3-1.4-kubectl-skew&width=3

Also, we have repeated problems with the lack of portability of shell scripts. If we're not going to ban them, can we make them more hermetic?

cc @ixdy @kubernetes/sig-testing @freehan @spxtr @lavalamp