test: select-e2e's full-suite asserts pass on any selection larger than five suites #3652

Summary

Three tests in hack/select-e2e_test.bats assert "the full suite was selected" as [ "$(echo "$output" | wc -w)" -gt 5 ] (lines 67, 76, 117 at the current main). A word count above five is a proxy for "full suite": it passes on any selection of six or more suites, so a regression that drops the selection from the full set to, say, half of it stays green in all three tests. The comment above the first one states the intent plainly ("Full suite means more than 5 suites"), which makes the gap easy to inherit as fact.

Why it matters

These are exactly the tests guarding the escalation paths (shared library change, e2e harness change), where the failure mode worth catching is partial escalation: a selector bug that picks some suites but not all of them. The current assert cannot distinguish "all" from "more than five", so the guard is real against total failure and blind against the partial one.

Fix shape

Assert equality with the full set instead of a threshold: derive the expected list the same way the script's full-suite branch does (or pin the known suite list with the fixture sources the tests already construct) and compare with the actual selection. The fixtures are already synthetic and small, so the exact expected output is knowable in each test.