I feel like this is probably a pebcak issue, but I'm puzzled why the change we're making is dealing with the first token following the keyword not being a # while our test input has # as the first token.

Am I misunderstanding and/or missing something obvious?

I would've thought the test input we'd use would include some non-# token in the first position, e.g. the input from the referenced issue:

cfg_if::cfg_if! {
    if {
    } else if #(&cpus) {
    } else [libc::CTL_HW, libc::HW_NCPU, 0, 0]
}

@calebcartwright Thank you for catching that. We should be using the snippet from the referenced issue.

While I was investigating what was going on I modified the snippet by adding the # just to see if that would prevent the ICE and accidentally committed that test case. Going to update it now.