Roo-Code/pr_description.md

Description

Fixes #5633

This PR resolves the critical HTML entity escaping/unescaping issue that was causing diff search/replace operations to fail or produce unintended content modifications.

Problem

The apply_diff tools were unescaping HTML entities (e.g., & → &) in search content for non-Claude models, but the actual file content still contained the escaped entities. This caused:

1. Search failures: When LLMs provided search content with HTML entities, the unescaping made it not match the actual file content
2. Unintended modifications: Partial matches could lead to incorrect replacements in files

Solution

• Removed HTML entity unescaping logic from applyDiffTool.ts and multiApplyDiffTool.ts for non-Claude models
• Preserved HTML entities in search content to ensure exact matching with file content
• Removed unused imports of unescapeHtmlEntities from both files
• Added comprehensive test coverage to prevent regression

Changes Made

• src/core/tools/applyDiffTool.ts:

  • Removed HTML entity unescaping for non-Claude models (lines 26-28)
  • Removed unused import of unescapeHtmlEntities (line 13)

• src/core/tools/multiApplyDiffTool.ts:

  • Replaced HTML entity unescaping logic with simple assignment (lines 414-418)
  • Removed unused import of unescapeHtmlEntities (line 13)

• src/core/tools/tests/applyDiffHtmlEntity.spec.ts:

  • Added comprehensive test suite with 5 test cases
  • Tests verify HTML entities are preserved in search/replace operations
  • Tests cover various entity types: &, <, >, ', "
  • Tests verify both Claude and non-Claude model behavior

Testing

• All existing tests pass
• Added comprehensive tests for HTML entity handling scenarios:
  • HTML entities preserved in search content for non-Claude models
  • HTML entities still unescaped for Claude models (backward compatibility)
  • Various entity types tested (&, <, >, ', ")
  • Both single and multiple entity scenarios covered
• Manual testing completed:
  • Verified search operations work with HTML entities in content
  • Confirmed no unintended content modifications occur
• All linting and type checking passes

Verification of Acceptance Criteria

• Search failures resolved: HTML entities in search content now match file content exactly
• Unintended modifications prevented: No more partial matches causing incorrect replacements
• Backward compatibility maintained: Claude model behavior unchanged
• No regressions: All existing functionality preserved
• Comprehensive test coverage: Edge cases and scenarios covered

Related Issues

This issue was previously attempted to be fixed in:

• PR #5608 (closed without merging)
• Issue #4077 (related HTML entity handling)

This PR provides a complete and tested solution that addresses the root cause.

Checklist

• Code follows project style guidelines
• Self-review completed
• Comments added for complex logic
• Documentation updated (test documentation)
• No breaking changes
• All tests passing
• Linting and type checking passes