Various office files wrong mimetype

[cars10]

According to this the mimetype of .oft (outlook email emplate) should be application/vnd.ms-outlook, but marcel currently reports application/x-tika-msoffice.

Some other issues include:

• .accdb, .mdb and .mde are reported as font/ttf
• .mht is reported as multipart/related

All of these files worked on marcel 0.3.3

[gmcgibbon]

oft and accsdb I can't get to work by name with 0.3.3, maybe they only work by content matching. I'll work on patching the others into the current types db in the meantime. If you could edit this test case to reproduce the type matching you describe in 0.3.3, that would be helpful. Thanks!

# frozen_string_literal: true

require "bundler/inline"

gemfile(true) do
  source "https://rubygems.org"

  git_source(:github) { |repo| "https://github.com/#{repo}.git" }

  # gem "marcel", github: "rails/marcel", branch: "main"
  gem "marcel", "0.3.3"
  gem "minitest"
end

require "minitest"
require "minitest/autorun"
require "marcel"

class BugTest < Minitest::Test
  { oft: "application/octet-stream", accesdb: "application/octet-stream", mdb: "application/vnd.ms-access", mht: "application/x-mimearchive" }.each do |ext, type|
    define_method("test_#{ext}") do
      assert_equal type, Marcel::MimeType.for(name: "file.#{ext}")
    end
  end
end

[cars10]

office_files.zip
Hi, some errors on my side:

• its accdb not accsdb, and this was reported as application/octet-stream in marcel 0.3.3. Not great, but better than font/ttf
• .oft was reported before as application/x-ole-storage, which is what we currently use for our whitelist. But we can change that ofc if you want to report it as application/vnd.ms-outlook like the link that i found in the opener suggests.

The following code works with the attached files:

# frozen_string_literal: true

require 'bundler/inline'

gemfile(true) do
  source 'https://rubygems.org'

  git_source(:github) { |repo| "https://github.com/#{repo}.git" }

  # gem "marcel", github: "rails/marcel", branch: "main"
  gem 'marcel', '0.3.3'
  gem 'minitest'
end

require 'minitest'
require 'minitest/autorun'
require 'marcel'

class BugTest < Minitest::Test
  {
    oft: 'application/x-ole-storage',
    accdb: 'application/octet-stream',
    mdb: 'application/vnd.ms-access',
    mde: 'application/vnd.ms-access',
    mht: 'message/rfc822'
  }.each do |ext, type|
    define_method("test_#{ext}") do
      assert_equal type, Marcel::MimeType.for(Pathname.new("files/#{ext}.#{ext}"))
    end
  end
end

office_files.zip

[gmcgibbon]

It looks like the TTF identification bug is being caused by a manual mime extension:

marcel/lib/marcel/mime_type/definitions.rb

Line 40 in a525d5b

Marcel::MimeType.extend "font/ttf", magic: [[0, "\x00\x01\x00\x00"]], extensions: %w( ttf ttc )

We either need to define magic for accdb/mdb files or adjust the magic matcher for TTFs. Preferably, both access DB types should be able to identify as application/vnd.ms-access.

[gmcgibbon]

I can't find a good magic matcher for oft files. Microsoft's specs for file naming only mention Word, PowerPoint, and Excel.

If it isn't feasible to match to a more specific type for outlook files, we may want to look at falling back to application/x-ole-storage in cases where we would return the new tika type.