Parse failures on valid DOCX files: 'Failed to read from zip' masks XML parsing errors

[developer0hye]

Summary

We're using docx-rs (v0.4) in office2pdf to parse DOCX files. When processing ~2,800 real-world DOCX files (from LibreOffice and Apache POI test suites), 19 files fail. 16 of these return "Failed to read from zip" even though the zip archives are valid — the error actually comes from XML parsing failures within the zip entries.

Related to #832.

Failing files (19 files)

"Failed to read from zip" (16 files)

These are valid ZIP files (verified independently). The error message is misleading — the actual failure is in XML parsing:

• cloud.docx, sdt_after_section_break.docx (LibreOffice)
• tdf108350.docx, tdf108408.docx, tdf108714.docx, tdf108806.docx, tdf108849.docx, tdf109306.docx, tdf109524.docx, tdf111550.docx, tdf111964.docx, tdf124670.docx, tdf129659.docx (LibreOffice regression tests)
• tdf171025_pageAfter.docx, tdf171038_pageAfter.docx (page break after section)
• xml_space.docx (xml:space attribute handling)

"Failed to read xml" (1 file)

• math-malformed_xml.docx — intentionally malformed math XML

ParseFloatError panic (1 file)

• tdf79272_strictDxa.docx — unwrap() on parsing table column width as float for Strict OOXML dxa unit values

Deep nesting failure (1 file)

• deep-table-cell.docx (Apache POI) — deeply nested table cells cause parse failure

Issues

1. Misleading error message: "Failed to read from zip" when the actual problem is XML parsing. It would help if the error distinguished between zip I/O errors and XML parsing errors.
2. Panic on Strict OOXML: The dxa parsing calls unwrap() on parse::<f64>(), which panics on Strict OOXML unit formats.
3. Limited OOXML support: Many LibreOffice-specific OOXML extensions (structured document tags, complex section breaks) are not handled.

Expected behavior

• XML parsing errors should be reported as such, not masked as zip errors
• No panics — parsing failures should return Err
• Unsupported OOXML features should be skipped with a warning rather than causing total parse failure

Environment

• docx-rs v0.4
• office2pdf v0.3.0
• Rust 1.85+
• Files from LibreOffice and Apache POI test suites (publicly available)