finfo returns wrong mime type for (.tar).xz files

[nilllzz]

Description

Running a mime type detection on .xz archive files returns an incorrect mime type.

This code:

$path = "[Path to .xz file]";
$finfo = new \finfo(FILEINFO_MIME_TYPE);
$mimeData = $finfo->file($path);
echo $mimeData;

Resulted in this output:

"application/octet-stream"

But I expected this output instead:

"application/x-xz; charset=binary"

This is apparently an issue with libmagic version 5.40 and was fixed in 5.41 (ref https://bugs.astron.com/view.php?id=257).
The specific fix to the magic file can be found here: file/file@9b0459a

I suggest a similar fix to GH-8805 that was done in 34fa65a, since the issue seems very similar.

PHP Version

PHP 8.2.3

Operating System

No response

[iluuu1994]

@nilllzz Thank you for the detailed report!

@weltling Hi! Is this process documented somewhere? I see ext/fileinfo/data_file.c is generated by ext/fileinfo/create_data_file.php, but I'm not sure how the patch files are related, and where/when to apply them. Also, my local magic.mgc file produces massive changes to data_file.c, some I'm guessing there's some canonical one to use?

[weltling]

@iluuu1994 no much docs on that and in many cases things have to be done manually. We carry magicdata.patch which has to be applied to libmagic before creating the magic db binary. That one is then used for embedding in the PHP sources.

Thanks