Misidentifies jpeg as "audio/vnd.dts.hd"

[rwstauner]

I have 10 photos.
file and identify (imagemagick) think all are jpegs, however this gem classifies one of them incorrectly.

I used the "magic" found here:
https://github.com/minad/mimemagic/blob/v0.3.0/lib/mimemagic/tables.rb#L1487
to verify:

$ for i in *.jpg; { echo $i; file $i; identify $i; head -c 18725 $i | grep -Eo 'dX %'; }                                                                                                                        
1470445_01.jpg
1470445_01.jpg: JPEG image data, JFIF standard 1.02
1470445_01.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 85.6KB 0.000u 0:00.000
1470445_02.jpg
1470445_02.jpg: JPEG image data, JFIF standard 1.02
1470445_02.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 74.6KB 0.000u 0:00.000
1470445_03.jpg
1470445_03.jpg: JPEG image data, JFIF standard 1.02
1470445_03.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 73.4KB 0.000u 0:00.000
1470445_04.jpg
1470445_04.jpg: JPEG image data, JFIF standard 1.02
1470445_04.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 69.6KB 0.000u 0:00.000
1470445_05.jpg
1470445_05.jpg: JPEG image data, JFIF standard 1.02
1470445_05.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 82.5KB 0.000u 0:00.000
Binary file (standard input) matches
1470445_06.jpg
1470445_06.jpg: JPEG image data, JFIF standard 1.02
1470445_06.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 68.4KB 0.000u 0:00.000
1470445_07.jpg
1470445_07.jpg: JPEG image data, JFIF standard 1.02
1470445_07.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 49.6KB 0.000u 0:00.000
1470445_08.jpg
1470445_08.jpg: JPEG image data, JFIF standard 1.02
1470445_08.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 30KB 0.000u 0:00.000
1470445_09.jpg
1470445_09.jpg: JPEG image data, JFIF standard 1.02
1470445_09.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 35.9KB 0.000u 0:00.000
1470445_10.jpg
1470445_10.jpg: JPEG image data, JFIF standard 1.02
1470445_10.jpg JPEG 800x600 800x600+0+0 8-bit DirectClass 55.2KB 0.000u 0:00.000

[rwstauner]

FTR if I cut the file short (first 13,000 bytes) it works correctly:

💥  docker run --rm -it -v ~work/tmp/1470445_05.jpg:/file.jpg ruby:2.3 bash -c 'gem install mimemagic; dd if=/file.jpg of=/file2.jpg bs=1 count=13000; ruby -rmimemagic -e "%w[/file.jpg /file2.jpg].each { |f| puts %Q|#{f}:#{ MimeMagic.by_magic(File.open(f)) }| }"'
Fetching: mimemagic-0.3.1.gem (100%)
Successfully installed mimemagic-0.3.1
1 gem installed
13000+0 records in
13000+0 records out
13000 bytes (13 kB) copied, 0.663014 s, 19.6 kB/s
/file.jpg:audio/vnd.dts.hd
/file2.jpg:image/jpeg

So the gem is selecting based on the a worse match (a match somewhere between bytes 13000 and 18725) instead of a better one (first 2 or 3 bytes).

[natematykiewicz]

data = File.binread('1470445_05.jpg')
MimeMagic::MAGIC.select {|type, matches| MimeMagic.send(:magic_match_str, data, matches) }
=> [["audio/vnd.dts.hd", [[0..18725, "dX %"]]], ["image/jpeg", [[0, "\xFF\xD8\xFF"], [0, "\xFF\xD8"]]]]

So, this file does match the image/jpeg format, but find is returning the first one.

[minad]

Probably this is a flaw in the algorithm or in the priority order in the upstream freedesktop file. Maybe someone could take a look on how this could be done better or how other tools based on the freedesktop database operate?

[janko]

This issue should be solved by #40, because the image/jpeg is now on top of the table, so it will be returned first.

[minad]

I assume this is fixed?