LibraryFolderDedupeMigrationTests claims CI sets ETV_REQUIRE_MYSQL_TESTS=1 — nothing does #689

Open
opened 2026-07-27 22:08:19 +02:00 by timothy · 0 comments
Owner

Found while working #668 (PR #687), where I copied this sentence into a new fixture and a review blocked on it. The copy was fixed; the original is still on main.

The false claim

ErsatzTV.Tests/Integration/LibraryFolderDedupeMigrationTests.cs (~lines 37-39) says:

MySQL needs a live server, supplied via ETV_TEST_MYSQL_CONNECTION. Without it the MySQL fixture ignores — a visible skip, never a silent pass. CI sets ETV_REQUIRE_MYSQL_TESTS=1, which turns that skip into a hard failure, so the gate cannot quietly degrade into "connected to nothing and passed".

Grep the repo: nothing sets ETV_REQUIRE_MYSQL_TESTS, and nothing sets ETV_TEST_MYSQL_CONNECTION. The only other mention is the note at .gitea/workflows/docker-build.yml (~411-419) explaining that running this fixture against the live mysql service was implemented and then removed as non-deterministic, with re-arming tracked by #627.

So in CI the MySQL half silently Assert.Ignores — precisely the degradation the sentence claims is impossible.

Why it matters

The sentence does not merely describe a mechanism; it asserts a guarantee that a reader will rely on. Anyone extending MySQL coverage reads it as "CI enforces this" and reasonably concludes their new fixture is protected. That is exactly what happened in #687: the sentence was copied verbatim into a new fixture, and it took an independent review to catch it.

The mechanism it describes is real and correct — ETV_REQUIRE_MYSQL_TESTS=1 does turn the skip into a failure. What is false is that anything sets it.

Suggested fix

Reword to separate the mechanism from its (absent) arming, as #687 did for the new fixture:

Setting ETV_REQUIRE_MYSQL_TESTS=1 turns that skip into a hard failure, so an armed lane cannot degrade into "connected to nothing and passed". CI does not currently arm it (see the note in .gitea/workflows/docker-build.yml; re-arming is tracked by #627), so this half skips there.

Check whether the decision record for #491/#627 carries the same claim, and sweep it too.

Done-when

  • The sentence in LibraryFolderDedupeMigrationTests no longer asserts that CI arms the lane
  • Any decision record repeating the claim is corrected (Decisions-Edit: yes trailer if it is rationale prose)
  • A grep for ETV_REQUIRE_MYSQL_TESTS shows no remaining claim that something sets it
  • Adversarial review passed

Context: #687, #668, #627, #491.

Found while working #668 (PR #687), where I copied this sentence into a new fixture and a review blocked on it. The copy was fixed; **the original is still on `main`.** ## The false claim `ErsatzTV.Tests/Integration/LibraryFolderDedupeMigrationTests.cs` (~lines 37-39) says: > MySQL needs a live server, supplied via `ETV_TEST_MYSQL_CONNECTION`. Without it the MySQL fixture **ignores** — a visible skip, never a silent pass. CI sets `ETV_REQUIRE_MYSQL_TESTS=1`, which turns that skip into a hard failure, so the gate cannot quietly degrade into "connected to nothing and passed". Grep the repo: **nothing sets `ETV_REQUIRE_MYSQL_TESTS`, and nothing sets `ETV_TEST_MYSQL_CONNECTION`.** The only other mention is the note at `.gitea/workflows/docker-build.yml` (~411-419) explaining that running this fixture against the live `mysql` service was implemented and then **removed** as non-deterministic, with re-arming tracked by #627. So in CI the MySQL half silently `Assert.Ignore`s — precisely the degradation the sentence claims is impossible. ## Why it matters The sentence does not merely describe a mechanism; it asserts a **guarantee that a reader will rely on**. Anyone extending MySQL coverage reads it as "CI enforces this" and reasonably concludes their new fixture is protected. That is exactly what happened in #687: the sentence was copied verbatim into a new fixture, and it took an independent review to catch it. The mechanism it describes is real and correct — `ETV_REQUIRE_MYSQL_TESTS=1` *does* turn the skip into a failure. What is false is that anything sets it. ## Suggested fix Reword to separate the mechanism from its (absent) arming, as #687 did for the new fixture: > Setting `ETV_REQUIRE_MYSQL_TESTS=1` turns that skip into a hard failure, so an **armed** lane cannot degrade into "connected to nothing and passed". **CI does not currently arm it** (see the note in `.gitea/workflows/docker-build.yml`; re-arming is tracked by #627), so this half skips there. Check whether the decision record for #491/#627 carries the same claim, and sweep it too. ## Done-when - [ ] The sentence in `LibraryFolderDedupeMigrationTests` no longer asserts that CI arms the lane - [ ] Any decision record repeating the claim is corrected (`Decisions-Edit: yes` trailer if it is rationale prose) - [ ] A grep for `ETV_REQUIRE_MYSQL_TESTS` shows no remaining claim that something sets it - [ ] Adversarial review passed Context: #687, #668, #627, #491.
timothy added the ci-cdpriority: low labels 2026-07-27 22:08:19 +02:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: timothy/ersatztv#689