test(installer): the PowerShell sanitiser's SS3 + floor behaviour had no test

[LukasWodka]

Refs tracebloc/backend#2084 — its carved-out sub-task, not the ticket itself.

The gap

client#736 extended ConvertTo-SanitizedInput with the SS3 family (ESC O <final>) and the unknown-family floor, and shipped both with zero PowerShell coverage. The four cases in that Describe block are all CSI — so on Windows the entire SS3 half and the whole floor were unverified, while the bash and Go copies each had cases.

That asymmetry is the mechanism behind the bug #736 was fixing. The rule is hand-copied into three languages; only one shape was ever tested in one of them; SS3 went missing from all three at once (tracebloc/cli#516) — exactly as the CSI gap had before it (cli#364 / client#362). A rule covered in two of three implementations is a rule that drifts in the third.

What's here

Eight It blocks, mirroring the bats corpus in install-client-helm.bats case-for-case:

• SS3 around real content → stripped
• SS3-only (arrows, Home/End, F1/F2) → empty, so the caller re-prompts
• SS3 + CSI mixed in one value
• a bare O is not an escape (OPTIMUS-01 survives)
• truncated ESC O with no final byte → empty
• an unknown family (ESC N, SS2 — deliberately not in the strip list) alone → refused
• the same family beside content → content kept
• the floor counts non-Latin letters as content

The non-Latin case is asserted as a pair with the ASCII one. Either alone proves nothing: the point is that keep-vs-reject must not depend on the script a name is written in (Bugbot, #736), and only the two together say that.

Mutation-proven

Each anchor confirmed to apply, against the 12 cases in the block:

mutation	reddens
drop SS3 from the strip (the cli#516 bug itself)	2
remove the floor entirely	2
floor uses [A-Za-z] instead of \p{L}	1
match O<final> without requiring ESC (the obvious wrong fix)	4

Full Pester suite: 671 passed, 0 failed, 13 skipped. gen-manifest.sh --check passes (test-only change, no lib touched).

What this deliberately does not do

backend#2084's structural half — one canonical corpus file that all three suites derive from, plus a cross-repo drift check — is untouched. Splitting it out was the point: a fixture wired into two of the three implementations is the same defect wearing a different hat, and the corpus genuinely spans two repos, so it needs a vendoring + fail-closed drift mechanism (scripts/tests/check-drift.sh is the shape) rather than being tacked onto a test PR.

Until that lands, matching case lists across the three suites are the only thing making them comparable by reading — hence mirroring the bats corpus one-for-one rather than inventing a different set here.

🤖 Generated with Claude Code

---

Note

Low Risk
Test-only change with no production code touched; risk is limited to CI/test maintenance.

Overview
Adds eight Pester cases to the ConvertTo-SanitizedInput Describe in install-k8s.Tests.ps1, filling a gap where only CSI-shaped inputs were tested after #736.

The new tests align with the bats corpus in install-client-helm.bats: SS3 stripping and SS3-only → empty (re-prompt contract), mixed SS3/CSI, OPTIMUS-01 surviving (no bare-O false match), truncated ESC O, unknown escape family (SS2) refused vs content preserved, and non-Latin + ASCII pairs for the floor’s \p{L} rule.

No changes to ConvertTo-SanitizedInput or other libs—test-only.

^{Reviewed by Cursor Bugbot for commit d7f8b5c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

/fr-pass

Best-effort functional review passed (triage: merged clean, non-interactive; behavioral evidence limited while e2e journey is red — backend#2206). Advancing to Ready for prod.