Docx files are corrupted when zipped using RubyZip

[tinabel]

Hi there -- I'm having an issue with docx file headers being corrupted whenever I zip them up using RubyZip. I've tried to use the write_buffer solution, but I'm also zipping files recursively and it's not quite working right. Will there be any solution for docx that doesn't involve using write_buffer?

[tinabel]

I was able to mitigate this issue by using Nokogiri to parse the XML while zipping the files. Here is the gist for what I was able to do: https://gist.github.com/tinabel/ddd5cc9b0dd762986918520a132800d2

[hainesr]

I'd love to have a crack at investigating this issue, but I know nothing about Word documents and I'm struggling to reproduce it.

What I've tried:

1. In Windows: Create new test.docx file in Word with a few random words in it.
2. In Linux (in IRB):

zipfile = Zip::File.open("/opt/windows/word.zip", true)
zipfile.add("zipped.docx", "/opt/windows/test.docx")
zipfile.close

3. In Linux:

$ zipinfo /opt/windows/word.zip 
Archive:  /opt/windows/word.zip
Zip file size: 9683 bytes, number of entries: 1
-rw-r--r--  5.2 unx    12278 t- defN 20-May-29 18:38 zipped.docx
1 file, 12278 bytes uncompressed, 9563 bytes compressed:  22.1%

4. In Windows: Unzipped archive using 7Zip
5. In Linux (no differences):

$ diff /opt/windows/test.docx /opt/windows/zipped.docx

6. In Windows: Loaded zipped.docx into Word.

This all works. I'm using ruby 2.4.6p354 (2019-04-01 revision 67394) [x86_64-linux] and RubyZip HEAD.

I'm sure I'm missing something important!

[hainesr]

I have now tried downloading the gist linked above and running it on a directory of assorted .docx and .xlsx files - both with and without the Nokogiri step - and both times all the resulting files can be unzipped and then loaded by Word and Excel with no issues at all. The .docx files were complex and had background images, tables, foreground images, etc.

I'm using ruby 2.7.2 and have repeated this with RubyZip 2.3.x and 3.0 (HEAD).

If someone can send me a file that is corrupted when zipped with RubyZip I'd love to investigate this further, or is there another way I should be testing this? Otherwise I wonder if files are getting corrupted in another step before RubyZip gets hold of them?

I cannot reproduce this issue with the current information available, sorry.

[sandstrom]

@hainesr I'd open the discussions tab on the repo, and move this issue there.