docs(web/guides): correct 20 audited testing-guide claims (batch 2)

[bpamiri]

Batch 2 of the guide behavioral-audit docs wave — all 20 docs-affected claims for work item p1-15-testing (testing/*.mdx). Every correction below was live-verified by the audit (raw evidence in the p1 batch-2 verifier output); broken-but-unfixed behavior is documented as-is with the tracking issue cited.

index.mdx

• IDX-10 — full engine x DB matrix runs on a weekly cron + manual dispatch, not "on every PR"; engine list corrected to Lucee 6/7, Adobe 2023/2025, BoxLang (compat-matrix.yml:1-23). Same fix applied to the CI-integration paragraph lower on the page for consistency with ci-integration.mdx.
• IDX-11 — wheels test --ci is accepted for forward-compat but currently changes nothing (ciMode parsed, never used; verified byte-identical output and exit codes). Browser gating is the server env (WHEELS_CI/WHEELS_BROWSER_CI_ENABLE), not this flag.
• IDX-07 — runner-URL example used the wrong directory shape for the core endpoint; rewritten around /wheels/app/tests?directory=tests.specs.models with the core form (directory=wheels.tests.specs.model) noted. Unrecognized directory values silently run the FULL suite — cited #3083.
• IDX-08 — populate is opt-out (?populate=false), defaulting true when absent (runner.cfm:113); page previously said opt-in via ?populate=true, contradicting its sibling fixtures-and-test-data.mdx.
• Setup-files table: tableName() → table() (the setter; tableName() is a getter no-op, docs+model: guides and CLAUDE.md use non-existent tableName("x") setter — silent no-op, models fall back to the convention table (real setter is table()) #3079 finding).

model-tests.mdx

• MOD-03 — .persisted() does not exist (throws); the API is isPersisted() / isNew() (miscellaneous.cfc:244/228).

view-and-form-tests.mdx

• VFT-01 — assertSee/assertDontSee/assertSeeInOrder are case-insensitive (FindNoCase, verified live); page said case-sensitive.
• VFT-06 — the encoded apostrophe is ' (hex) on both Lucee 7 and Adobe 2023, so the assertSee("can't be empty") example failed everywhere probed. Example rewritten to the split-assertion pattern (primary recommendation) with the corrected ' literal documented; trailing bullet fixed too.

functional-tests.mdx

• FUN-03 — both engines emit text/html;charset=UTF-8 (no space) and assertHeader is exact-match; the "text/html; charset=UTF-8" example failed as written. Fixed the literal + added an exact-match note.
• Same ' → ' correction in the split-assertion explanation.

fixtures-and-test-data.mdx

• FIX-03 — application.wo.execute(...) does not exist ("No matching function [EXECUTE] found", verified live); execute() lives only on wheels.migrator.Migration. The minimal populate.cfm example would crash at boot. Rewritten to native queryExecute(..., {datasource: application.wheels.dataSourceName}) via a small runSql() helper; scope-gotcha section corrected (application.wo.model() still works and is kept).
• tableName("test_posts") → table("test_posts") in the TestPost example (getter no-op; all framework test models use table()). This also covers the p1-13 cross-file sketch entry for this file.
• Added an isolation caution: the runner mutates the live application.wheels (settings/path swaps leak into normal requests) — cites #3025.

browser-tests.mdx

• BRW-04 — browserDescribe()'s aroundEach auto-skips when browserTestSkipped is set (BrowserTest.cfc:140-152); the hand-written if (this.browserTestSkipped) return; guard is only needed inside plain describe(). Prose updated and the guard line stripped from all example specs.
• BRW-08 — dialog error type is Wheels.BrowserDialogNotSupported, not Wheels.DialogSupportMissing.
• BRW-15 — form helpers emit a dashed id="post-title", not camelCase id="postTitle" (matches sibling page).
• BRW-16 — no beforeAll fixture-route re-include exists; replaced with a warning that route-clearing specs must restore the route table.
• BRW-14 — stale cfwheels/cfwheels CLAUDE.md link → wheels-dev/wheels (branding rule).
• loginAs: noted app-side mounting requires set(loadBrowserTestFixtures=true) (default false) in testing/development.

running-tests-locally.mdx

• RUN-03 — --core auto-detect is removed (Module.cfc:738 comment; live wheels test at the framework repo root runs the app suite). Aside rewritten: pass --core explicitly; --core flag description aligned.
• RUN-06 — single-spec-file filtering via directory= matches 0 bundles; the working form is testBundles=wheels.tests.specs.model.callbacksSpec (verified live; also replaced the nonexistent postSpec). Filter table updated; added the directory allowlist silent-fallback note citing #3083.
• RUN-10 — compose.yml ships mysql, postgres, sqlserver, cockroachdb, oracle services; there is no mssql or h2 service (H2/SQLite are embedded). The docker-compose.db-*.yml files are demo-app dev overlays, now noted as such. "You'll learn" DB list aligned.
• RUN-11 — the soft-fail CI database is Oracle (SOFT_FAIL_DBS="oracle"), not CockroachDB.
• RUN-12 — JAVA_HOME preflight qualified: the brew wrapper exports JAVA_HOME itself, so the refusal can't trigger on brew installs (prereqs + failure-modes sections).
• --ci example comment aligned with IDX-11 (non-zero exit applies to every run).

ci-integration.mdx

• CI-01 — reporter table was stale: on released 4.0.3, --reporter=tap emits TAP v13 and --reporter=json emits the raw JSON document (verified live; Module.cfc reporter dispatch). Table rewritten; --ci row kept with the universal non-zero-exit note.
• Pinned LuCLI bumped v0.3.3 → v0.3.7 to match pr.yml's LUCLI_VERSION.
• Browser-gating paragraph aligned with BRW-04 (auto-skip via browserDescribe, no per-it guard).

Verification

pnpm verify:docs over all 8 changed files: 38 tagged blocks, 38 passed, 0 failed (exit 0).

🤖 Generated with Claude Code

[wheels-bot]

Wheels Bot — Reviewer

TL;DR: This PR corrects 20 audited factual claims across the eight v4-0-0 testing-guide pages (batch 2 of the behavioral audit). I verified every load-bearing claim in the diff against the framework source and CI config, and they all check out — several at the exact line numbers cited in the PR body. No correctness, cross-engine, or security findings; the commit message conforms to commitlint. One minor docs nit below (an H2 omission in the index page's matrix database list). Verdict: comment — nothing blocking.

Verification audit (claims checked against source)

Claim	Evidence
Full matrix is weekly cron + manual dispatch; engines Lucee 6/7, Adobe 2023/2025, BoxLang	.github/workflows/compat-matrix.yml:5-23 (cron: '0 2 * * 0', workflow_dispatch, matrix list)
Oracle (not CockroachDB) is the soft-fail DB	compat-matrix.yml:293 — SOFT_FAIL_DBS="oracle"
--ci is parsed but currently changes nothing	cli/lucli/Module.cfc:731,755,4735 — ciMode is captured and passed into runTests() but never read in its body
--reporter=tap emits TAP v13, --reporter=json emits the raw document	Module.cfc:4816-4824 reporter dispatch; TAP v13 emitters at Module.cfc:4869+
--core auto-detect removed; app suite is the default	Module.cfc:738-745 — comment documents exactly the removal and the silent-wrong-suite rationale the doc now states
Populate is opt-out, default true	vendor/wheels/tests/runner.cfm:112 — StructKeyExists(url, "populate") ? url.populate : true
Unrecognized directory silently runs the full suite (#3083)	runner.cfm:126-135 — allowlist regex falls through to the default wheels.tests.specs scope
testBundles= filters to a single spec file	vendored TestBox reads url.testBundles (vendor/wheels/wheelstest/system/TestBox.cfc:411-412); the runner itself only passes directory=
.persisted() doesn't exist; isPersisted()/isNew() do	vendor/wheels/model/miscellaneous.cfc:244 / :228 — exact lines cited in the PR body
table() is the setter, tableName() the getter	miscellaneous.cfc:35 (table) vs :177 (tableName)
assertSee/assertDontSee/assertSeeInOrder are case-insensitive	vendor/wheels/wheelstest/TestClient.cfc:251,264,280 — all FindNoCase
assertHeader is exact-match	TestClient.cfc:374 — hdrs[arguments.name] != arguments.value
browserDescribe() auto-skips via aroundEach when browserTestSkipped is set	vendor/wheels/wheelstest/BrowserTest.cfc:138-152 — the skip path even has a comment saying hand-written guards are unnecessary
Dialog error type is Wheels.BrowserDialogNotSupported	vendor/wheels/wheelstest/BrowserClient.cfc:950,970; no DialogSupportMissing anywhere in the tree
No automatic fixture-route re-include in beforeAll (BRW-16)	BrowserTest.cfc:61+ — no route include exists; the old claim was indeed wrong
Form helpers emit dashed ids (post-title)	vendor/wheels/view/miscellaneous.cfc:596-617 ($tagId joins with -); the doc comment at :626 even calls ##post-title the historical selector
loadBrowserTestFixtures defaults false, gated to testing/development	vendor/wheels/events/init/views.cfm:30; vendor/wheels/Global.cfc:1526-1531
execute() exists only on migration CFCs	only definition under vendor/wheels/ is vendor/wheels/migrator/Migration.cfc; both real populate files use native <cfquery datasource="#application.wheels.dataSourceName#"> (tests/populate.cfm:63-95), matching the rewritten example
compose ships mysql, postgres, sqlserver, cockroachdb, oracle (no mssql/h2 service); the docker-compose.db-*.yml files are dev overlays	compose.yml service list; docker-compose.db-h2.yml / db-mysql.yml / db-postgres.yml exist at the repo root
Pinned LuCLI 0.3.7 matches CI	.github/workflows/pr.yml:33 — LUCLI_VERSION: "0.3.7"

The commit (05f57a869) is a valid conventional commit (docs(web/guides): …, 66-char header), DCO-signed matching the author identity. A changelog fragment is not required for a docs-type change. The PR body reports pnpm verify:docs green over all 8 files (38/38 tagged blocks); the script exists at web/sites/guides/package.json:13.

Docs

• web/sites/guides/src/content/docs/v4-0-0/testing/index.mdx:70 — the corrected matrix sentence now reads "…and every supported database (MySQL, PostgreSQL, SQL Server, SQLite, CockroachDB, Oracle)", dropping H2 from the list (the pre-PR text included it). But the weekly matrix does run H2 on the Lucee legs — compat-matrix.yml:41-44 appends h2 to DATABASES for Lucee engines — and this same PR's running-tests-locally.mdx "You'll learn" list correctly names H2 as a supported target. Suggest keeping the enumeration complete with its caveat, e.g. "(MySQL, PostgreSQL, SQL Server, SQLite, CockroachDB, Oracle, plus H2 on Lucee)". Minor — does not block.

Everything else in the diff is accurate, internally consistent across the sibling pages, and properly cites the open tracking issues (#3025, #3083) for broken-but-unfixed behavior instead of papering over it.