[BUGFIX] Fix escape handling in TextRoleRule

[CybotTM]

Summary

Fixes escape handling bugs in TextRoleRule for code-type text roles (:rst:, :php:, :code:, etc.).

Resolves: TYPO3-Documentation/render-guides#1188

Bug 1: Escaped backslash stays literal in rawPart

Code-type text roles use $rawContent (via #533). When the author writes \\ to display a single backslash, $rawPart preserved the literal \\ instead of resolving it.

RST input:

:code:`a\\b`

Before (broken):

<code>a\\b</code>

After (fixed):

<code>a\b</code>

Other escapes like \T or \* are intentionally preserved raw in $rawPart — code contexts need literal backslash-letter sequences (e.g., PHP namespaces like \App\Entity).

Bug 2: Escaped backtick swallows closing delimiter

The lexer has a catchable pattern that tokenizes a 3-char sequence (backslash + two backticks) as a single ESCAPED_SIGN token. This swallows the closing backtick of the text role, so TextRoleRule never finds a BACKTICK token to close — it rolls back and the role breaks entirely.

RST input:

:code:`text\``

Before (broken): role is not recognized, rendered as literal text:

:code:`text\``

After (fixed): post-loop recovery detects the swallowed backtick. The 3-char token is split semantically:

\``  =  \`  (escaped backtick → literal `)  +  `  (closing delimiter)

The escape character is consumed, the backtick becomes content:

<code>text`</code>

The recovery is narrowed to only the 3-char token — the only lexer pattern that swallows backticks. Regular 2-char escapes at end-of-input (like \T or \\ without a closing backtick) correctly roll back as genuinely unterminated roles:

RST input (genuinely unterminated — no closing backtick):

:role:`content\T

Result: rolls back, no role node produced (correct — the role was never closed).

Changes

TextRoleRule.php:

• Resolve \\ → \ in $rawPart (other escapes preserved raw)
• Add EOF fallback for the 3-char token: escape consumed, literal backtick in content
• Extract createTextRoleNode() helper to deduplicate role-building logic
• Simplify escape tracking from two variables to one ($lastEscapedToken)

TextRoleRuleTest.php:

• 3 new positive cases: escaped backslash, escaped backtick at end, only escaped backtick
• 2 new negative cases: unterminated roles with \T and \\ at end → assert rollback

Functional tests:

• Updated code-textrole-no-escape.html expected output
• New code-textrole-escape/ test fixture

Test plan

• All 8 unit tests pass (6 positive + 2 negative rollback cases)
• All 3 textrole functional tests pass
• No regressions in existing text role behavior

[CybotTM]

Marking this as ready for review, but I'd appreciate feedback on the approach — I'm not fully sure how this project wants to handle these edge cases.

Specific questions:

1. rawPart resolution for \\: This PR resolves \\ → \ in $rawPart, so that:

   :code:`a\\b`

   renders as a\b. Other escapes like \T are preserved raw for code contexts (PHP namespaces etc.). Is this the right trade-off, or should $rawPart stay fully unresolved?

2. EOF fallback for the 3-char token: The lexer's catchable pattern for backslash + two backticks swallows the closing backtick. The fix detects this post-loop and recovers, but only for that specific 3-char token. Is this recovery acceptable, or would a lexer-level fix be preferred?

3. Consistency of mid-content vs end-of-content escapes: Mid-content escaped backticks are still preserved raw in $rawPart (existing behavior, not changed here). The EOF fallback resolves them. Should these be consistent?

Happy to adjust the approach based on your guidance.

[jaapio]

Thank you for your work on this bugfix. Your contribution helped highlight some interesting edge cases and scenarios that needed verification, which was very valuable.

After further investigation, I discovered that the issue was more complex and required a broader fix. I’ve since implemented and merged a comprehensive solution that addresses the root cause. You might want to verify the changes in #1312.

I’ll go ahead and close this pull request, but I truly appreciate your effort and the insights your work provided. Looking forward to your future contributions!