timothy
2a2dcacd58
PR Gates / CI image pin matches docker/ci (pull_request) Successful in 11s
PR Gates / Docs update reminder (pull_request) Successful in 12s
PR Gates / decisions lifecycle (pull_request) Successful in 25s
Review verdict / Set review-verdict status (pull_request) Successful in 24s
PR Gates / Script tests (pytest) (pull_request) Successful in 50s
Build ErsatzTV Image / API docs in sync (OpenAPI + endpoint index) (pull_request) Successful in 1m35s
Build ErsatzTV Image / Formatting (changed .cs conform to .editorconfig) (pull_request) Successful in 1m24s
review-verdict/h10 Review-verdict: MERGEABLE @ 2a2dcac (base: main)
Build ErsatzTV Image / Functional E2E (curl + UI contracts) (pull_request) Successful in 6m14s
Build ErsatzTV Image / EF migration integrity (SQLite + MySql) (pull_request) Successful in 19m26s
Build ErsatzTV Image / Build & test (.NET) (pull_request) Successful in 23m38s
Build ErsatzTV Image / Build & push image (amd64) (pull_request) Has been skipped
Additive tests for properties #666 shipped correctly but left unguarded. No behaviour change. The stub's status read was hardcoded to "no verdict yet", so two whole branches of the classify step were unreachable from any test. Four mutations survived the full suite because of it — including re-introducing the literal ersatztv#647 fail-open, and overwriting an existing human verdict. The stub now models a transport error, a garbage body, and an existing verdict. `test_an_EMPTY_enumeration_is_not_exempt_even_for_a_BOT` needs the bot author to test anything: with a non-bot, the blank line an empty list produces already fails DOCS_ONLY, so the `count -eq 0` guard never decides the outcome. On the bot path it is the ONLY thing between an enumeration that read nothing and an unattended success. Verified by mutation — `grep -c .` -> `grep -c ''` grants a bot PR success while every other test stays green. Same short-circuit shape as the PROTECTED/DOCS_ONLY disjointness this file already documents. Two anchors were also unguarded: `grep -qxF` (author `ova` is a substring of `renovate`) and DOCS_ONLY's `$` (`evil.mdx` reads as docs-only). Five of the six mutations are caught behaviourally. The sixth — dropping the shell emptiness check — cannot be caught locally: `jq -e` over empty input exits 4 on jq 1.8 so the guard still fires on a dev Mac, and 0 on the runner's 1.6 where it is the actual bug. A structural assertion closes that gap, with comments stripped first, since a raw substring search is satisfiable by moving the guard into a comment while deleting the real one — verified. refs #649, #672