Magic can't get a proper mime type from a MP3 file

[kskadart]

Hey folks,

I try to get a mime type from an audio asset:

import magic
magic.from_file('test_audio.mp3', mime=True)

output:

'application/octet-stream'

As I can see from the ffprobe output the asset contains corrupted metadata:

ffprobe test_audio.mp3
...
[mp3 @ 0x7fcbb6904300] Skipping 48 bytes of junk at 70.

But I can get mime by the built-in python library mimetypes:

from mimetypes import MimeTypes
mime = MimeTypes()
mime.guess_type('test_audio.mp3')

output:

('audio/mpeg', None)

I can imagine that the magic and mimetypes have different logic... But probably you can suggest some robust approach how I can extract a mime type by the magic lib? I attach the audio asset and change ext from MP3 to MP4 because the github doesn't like MP3 format: https://github.com/ahupp/python-magic/assets/120260513/b614b76c-5651-47fb-a8a3-0001a60e5bfa

Thanks in advance

[ahupp]

I'm guessing mimetypes just looks at the extension but haven't looked closely at it. Interestingly, libmagic reports the correct type when not asking for MIME:

❯ file ~/home/Downloads/276412087-b614b76c-5651-47fb-a8a3-0001a60e5bfa.mp4
/home/adam/home/Downloads/276412087-b614b76c-5651-47fb-a8a3-0001a60e5bfa.mp4: Audio file with ID3 version 2.3.0

framework in ~
❯ file --mime ~/home/Downloads/276412087-b614b76c-5651-47fb-a8a3-0001a60e5bfa.mp4
/home/adam/home/Downloads/276412087-b614b76c-5651-47fb-a8a3-0001a60e5bfa.mp4: application/octet-stream; charset=binary

Unfortunately this is a libmagic bug outside of python-magic so I'm not sure how to resolve; maybe you can dig through the definition files and see how the mime and non-mime paths differ? (surprised they can differ though)

[kskadart]

Hey Adam, thanks for the fast answer! I'll try to open a ticket directly for the magiclib