test(view): regression spec for inline-SVG icon convention in migrator view

[bpamiri]

Summary

Adds a regression test (vendor/wheels/tests/specs/migrator/MigratorViewIconsSpec.cfc) that locks in the inline-SVG icon convention for the migrator dashboard.

The original bug (broken icon glyph on the Database Error placeholder, #2427 / #2425) has already been fixed on develop via two earlier PRs:

• #2562 replaced the offending Semantic UI icon-font tag on line 299 of vendor/wheels/public/views/migrator.cfm with an inline SVG that matches the sibling "No migration files found" placeholder.
• #2563 inlined the Semantic UI icon font globally (base64 data URI in _header.cfm / _header_simple.cfm), so font-based glyphs now render across every framework dev page (/wheels/guides, /wheels/info, /wheels/migrator, etc.).

This PR was originally opened against the unfixed state and conflicted with develop after #2562 / #2563 landed. Rebased to drop the duplicate view change + duplicate CHANGELOG entry and keep just the new regression spec — convention enforcement for migrator.cfm going forward.

Related Issue

Closes #2427 (duplicate of #2425; substantive fix is on develop).

Type of Change

• Bug fix
• New feature
• Enhancement to existing feature
• Documentation update
• Refactoring
• Test-only (regression coverage)

What this test asserts

1. The Database Error placeholder in vendor/wheels/public/views/migrator.cfm uses an inline <svg> (not an icon-font tag).
2. The whole file contains zero <i ...icon...> opener tags.

Both checks build their regex literals via Chr(60) & ... to avoid the Lucee tag-scanner trap documented in .ai/wheels/testing/.

Feature Completeness Checklist

• Tests — The new spec at vendor/wheels/tests/specs/migrator/MigratorViewIconsSpec.cfc is the entire change.
• Framework Docs — n/a (test-only, no public API change)
• AI Reference Docs — n/a
• CLAUDE.md — n/a
• CHANGELOG.md — n/a (the user-facing fix already has an entry from fix(view): inline Semantic UI icon font so dev page icons render #2563; this PR only adds a test)
• Test runner passes — Verified with an equivalent Python regex run against the current vendor/wheels/public/views/migrator.cfm on develop: both checks pass (anchor found at byte ~16700, <svg present in the 1500-char window before it, zero icon-font hits in the 50KB file). CI will run the CFML spec end-to-end.

Test Plan

# Local
bash tools/test-local.sh migrator

# CI
# Runs as part of every PR via pr.yml and snapshot.yml across the matrix.

The test reads migrator.cfm directly via FileRead(ExpandPath("/wheels/public/views/migrator.cfm")) — no database, no HTTP, no fixtures required. It will pass on every engine × DB combo so long as migrator.cfm keeps its inline-SVG convention.

🤖 Generated with Claude Code

[wheels-bot]

Wheels Bot — Reviewer A

TL;DR: This PR adds a single regression spec (vendor/wheels/tests/specs/migrator/MigratorViewIconsSpec.cfc) that enforces the inline-SVG icon convention in the migrator dashboard view. The change is test-only, correct, cross-engine safe, and follows BDD conventions. Approving with two minor nits.

---

Correctness

The test logic holds up under scrutiny:

• Find("Database Error", source) correctly returns the byte position of the anchor text (line 300 of migrator.cfm). The expect(anchor).toBeGT(0, ...) guard catches a missing anchor with a helpful message.
• The window arithmetic (windowStart = Max(anchor - 1500, 1), windowLen = anchor - windowStart) captures exactly 1500 characters immediately before the anchor — sufficient to contain the preceding <svg> block, which is confirmed to be ~380 chars in the current file. The comment explaining the 700-char estimate refers to path data length; 1500 is a safe upper bound.
• reFindNoCase returns 0 when no match is found. expect(fontIconHit).toBe(0) is the correct idiom for that return value.
• reMatchNoCase returns an array of all matches; expect(ArrayLen(hits)).toBe(0) is correct for "no matches".
• toBeGT, toInclude, and toBe all accept a message second argument as defined in vendor/wheels/wheelstest/system/Expectation.cfc lines 624, 560, and 168 respectively — usage is valid.

Conventions

The Chr(60) & "i[\s>][^>]*\bicon\b" pattern correctly avoids embedding a literal < in a CFML string expression, following the guidance in .ai/wheels/testing/ and consistent with the # scanner trap documented in CLAUDE.md. The spec extends wheels.WheelsTest (not the legacy wheels.Test) and uses BDD describe/it/expect syntax throughout — correct.

Nit — duplicate FileRead call (lines 23 and 59):

// it block 1 (line 23)
var source = FileRead(ExpandPath("/wheels/public/views/migrator.cfm"));

// it block 2 (line 59)
var source = FileRead(ExpandPath("/wheels/public/views/migrator.cfm"));

Both it blocks read the same file independently. A beforeAll (or a describe-level local.source read once at the top of run()) would eliminate the duplicate disk hit. This is cosmetic — the file is ~50 KB and reads in milliseconds — but worth a one-line fix.

Cross-engine

All functions used (FileRead, ExpandPath, Find, Mid, Max, reFindNoCase, reMatchNoCase, Chr, ArrayLen, ArrayToList) are standard CFML and work on Lucee 6/7, Adobe CF 2023/2025, and BoxLang. Arrow-function closures (() => {}) are supported by every engine in the CI matrix. No struct.map() on CFC objects, no application scope members, no client scope usage. The Left(str, 0) Lucee 7 gotcha (CLAUDE.md "Known cross-engine gotchas") does not apply here — Mid() is used, and the degenerate windowLen = 0 case (which would require "Database Error" to appear at byte position 1 in the file) cannot occur in practice.

Tests

This PR is entirely test code and is self-verifying. Both happy-path assertions (<svg present, zero icon-font tags) pass against the current migrator.cfm on develop as confirmed by the PR description's Python pre-check. ExpandPath("/wheels/public/views/migrator.cfm") follows the same /wheels/ virtual-path convention used in migratorSpec.cfc line 386 — path resolution is correct.

Docs

Nit — PR labels: The PR carries docs and enhancement labels, but the type-of-change checklist correctly marks [x] Test-only (regression coverage). The labels do not reflect the test-only nature and may mislead triage — worth updating.

Commits

Single commit: test(view): regression spec for inline-SVG icon convention in migrator view

• Type test: valid
• Scope view: valid (the spec tests migrator.cfm, a view template)
• Subject 60 chars, not ALL-CAPS: valid
• Describes the why ("regression spec for ... convention"), not just the what: valid

No commitlint violations.

Security

No user input, no SQL, no XSS surface. FileRead reads a known framework-owned path resolved via ExpandPath — no path traversal risk.

---