Bugfix — Issue #1875: 7 freeTestCases ftc.* keys stayed English in 8 locale bundles

Branch fix/issue-1875-ftc-i18n · Commits 29cc40397 (bundles), 719dcdaa0 (regression suite) · Refs #1262, #1265, #1844

Symptom

On the modernized Free Test Cases screen (gui/templates/results/freeTestCases.html) 7 keys of the ftc.* group rendered their raw English text in 8 of the 10 locale bundles (de/es/fr/it/ja/pt/ru/zh): ftc.allAssigned, ftc.elapsedSeconds, ftc.forProject, ftc.header, ftc.matchCount, ftc.testCases, ftc.testProject. Every other ftc.* key (and ro.json entirely) was already translated.

Root cause

1. The screen binds the keys via data-i18n (freeTestCases.html:84-90) and TLi18n.t()/t() (:154, :180, :228-229, :249).
2. gui/templates/i18n/i18n.js:175 — var str = _strings[key] || key; — a present key always renders its bundle value; there is no "is this English?" detection.
3. The #1262/#1265 translation batches created the 7 keys in the 8 bundles with values byte-identical to en.json — they were never translated, only copied.
4. ai/verify_i18n_coverage.sh (the #1844 gate) compares key sets only: a key that exists with an English value is coverage-green, so the drift shipped silently.

Why this method

The fix is value-only, no code change: translate the 7 keys in the 8 bundles, reusing each locale's own established terminology already present in the same bundles and in the legacy locale/*/strings.txt files, so the new values are consistent with every other screen in that language:

anchor	where it comes from
Test Project	ato.testProject / assign.testProject / cfa.testProject in the same bundle
for test project	ato.forProject in the same bundle
test cases	dsh.testCases / cwt.testCases in the same bundle
Match count	legacy $TLS_match_count / reqsearch.matchCount / cwt.matchCount
Elapsed seconds	reqcov.elapsedSeconds / rbtb.elapsedSeconds / cwt.elapsedSeconds
screen header	legacy $TLS_report_free_testcases_on_testproject (official de/es/ja/pt/zh; fr/it/ru crafted to match)

Alternatives rejected: (a) re-translating from scratch — risks diverging from the vocabulary every translated control already uses on the same screen; (b) a runtime "identical-to-en" fallback — wrong layer, would mask future gaps instead of fixing them; (c) extending the coverage gate to byte-compare values — larger scope, belongs to a separate issue, and would false-positive on intentional identity like the ftc.groupHeader placeholder template ({label}: {value} ({count})).

Files changed

• gui/templates/i18n/de.json, es.json, fr.json, it.json, ja.json, pt.json, ru.json, zh.json — 7 keys each, +56/−56 total.
• tmp/TLU_Test_Cases.md — appended regression suite (8/8 PASS).

ftc.groupHeader is intentionally left identical everywhere (format template with placeholders, no translatable words).

Verification

• python3 -m json.tool → OK on all 8 touched bundles.
• bash ai/verify_i18n_coverage.sh → PASS 9/9 bundles (exit 0).
• Cross-bundle diff → 0 ftc.* keys still identical to en.json outside en.
• Browser ?tproject_id=1&locale=de: header "Testfälle die keinem Testplan zugewiesen sind", "für Testprojekt", "Testprojekt", "Anzahl Treffer: 1", "Vergangene Sekunden: 0" — see screenshot below.
• Browser ?tproject_id=1&locale=zh: "没有关联到任何测试计划的测试用例", "匹配数量: 1", "已用秒数: 0" — see screenshot below.
• Console: no new Errors/Warnings; events table unchanged (only 2 pre-existing AUDIT log_level=16 rows).

Note on the suite gate

verify_test_suites.sh reported 1 FAIL ("line removed … existing-but-denied writes.") that is an EOL false positive: the merge-base 25815085e committed tmp/TLU_Test_Cases.md WITHOUT a trailing newline, so any append shows that unchanged line as removed. Diff between the EOL-normalized base and the candidate is pure additions (0 deletions), and the heading set is identical (41/41) — no content was lost.