Stop the tag scanner ending a tag inside an arrow operator

[ghedwards]

Closes #44.

My diagnosis in the issue was wrong

#44 says the expression rule likely commits before the arrow is seen. It doesn't — the grammar never gets the chance. CFMLStartTag.getEnd() returns at the first unquoted >, and the second character of => is one:

<cfset d = t => t.b()>       scanned as the tag  <cfset d = t =>
<cfset e = (x) -> x + 1>     scanned as the tag  <cfset e = (x) ->

The truncated text e = (x) = then parses to e = (x) — a well-formed expression the visitor accepts, with the body gone. That's why it was silent.

Fix

getEnd already special-cases quotes and the ---> of a comment. An arrow joins them.

The - case needs care: <cfset a = i--> is a decrement followed by the tag end, not an arrow. So - only counts when it doesn't follow another -. The fixture pins that alongside the arrows, alongside gt (where the > never appears) and a quoted subject="x=>y" (already handled by the quote tracking).

Fixture, not a JUnit case

This is the tag path, so TestTagFiles. Nothing under src/test/resources/tag exercised an arrow function, which is exactly why the suite stayed green. The recorded output shows each <cfset> capturing its whole expression and building a lambdaDeclaration:

nested = getURLs().map((target) -> target.toString())
countdown = i--
comparison = total gt 1

Reverting cfml.parsing/src/main fails it.

Verification

• 313 cfparser tests (including the 26 TestGetStartTagEnd cases, which is what I most expected to break), ./gradlew build
• Differential harness 8 → 7, nothing newly broken
• CFLint 675 against a clean build — this changes how every CFML tag is scanned, so the downstream run mattered more than usual here

---

Generated by Claude Code