fix: pre-encode XML to UTF-8 to avoid surrogate pair corruption in JSZip

[Yuof]

Problem

When JSZip processes large XML content (> 16KB), it chunks strings using \substring()\ which operates on UTF-16 code units. This can split surrogate pairs for characters above U+FFFF (like emoji or Material Design Icons).

Each surrogate then gets encoded as a separate 3-byte UTF-8 sequence, producing invalid CESU-8 instead of proper UTF-8. This corrupts the docx file and causes XML parsing errors.

Root Cause

JSZip's \DataWorker\ uses:
\\javascript
DEFAULT_BLOCK_SIZE = 16 * 1024
data.substring(index, nextIndex)
\\

The \Utf8EncodeWorker\ then processes each chunk independently, without handling split surrogates. I've opened a fix PR to JSZip: Stuk/jszip#963

Solution

This PR works around the issue by pre-encoding strings to UTF-8 using \TextEncoder\ before passing to \zip.file(). Since we pass \Uint8Array\ instead of strings, JSZip skips the problematic \Utf8EncodeWorker\ entirely.

Changes

• Added \�ncodeUtf8()\ helper function in
  ext-compiler.ts\ and \rom-docx.ts\
• Modified \zip.file()\ calls to use pre-encoded UTF-8 bytes for XML content

Testing

• Build succeeds
• TypeScript compiles without errors
• This is a minimal, low-risk change that ensures UTF-8 encoding happens correctly before JSZip processing

[Copilot]

Pull request overview

This PR addresses a bug in JSZip where string chunking can split UTF-16 surrogate pairs when processing large XML content (> 16KB), resulting in invalid CESU-8 encoding instead of proper UTF-8. The fix pre-encodes XML strings to UTF-8 using TextEncoder before passing them to JSZip, bypassing the problematic chunking behavior.

Changes:

• Added encodeUtf8() helper function to pre-encode strings to UTF-8 bytes
• Modified all zip.file() calls for XML content to use the pre-encoded UTF-8 bytes
• Binary content (images, fonts) continues to be passed directly without encoding

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
src/patcher/from-docx.ts	Added encodeUtf8() helper and applied it to XML content in the patching workflow
src/export/packer/next-compiler.ts	Added encodeUtf8() helper and applied it to all XML file generation in the compiler

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (ec4fdc2) to head (0373b68).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##            master     #3329   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files          291       291           
  Lines         8862      8864    +2     
  Branches      1435      1436    +1     
=========================================
+ Hits          8862      8864    +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[dolanmiu]

Looks good. I fixed tests and removed that TextEncoder fallback. All major browsers have this already, so it's ok.

Removed Buffer dependency, there was a PR which removed Buffer, so best to adhere to that