Docx mime type incorrectly guessed

[AlfonsoUceda]

Hi!

I've been experienced an error with Marcel where having a docx file (a real one) gets detected as application/zip instead of application/vnd.openxmlformats-officedocument.wordprocessingml.document, instead using file --mime-type command correctly detects the right mime type.

I've been researching why this happens and it seems the matchers defined here https://github.com/rails/marcel/blob/main/lib/marcel/tables.rb#L2416 expect a right order of those strings in the file.
I've compared two docx files one that is correctly detected by Marcel and the other one not. It seems that identifier aren't in the correct order. E.g. the [Content_Type].xml check is almost at the end of the file.

The following snippet works but I guess the gem reads first bytes for performance reasons so doesn't have to check include in the whole file.

file = Pathname.new('PATH_DOCX').open
content = file.read
content.include?('[Content_Types].xml') # => true
content.include?('word/') # => true
content.include?('_rels/.rels') # => true

Cheers