Auto-detect compression appears to be broken for some files about createrepo_c HOT 6 CLOSED

I found that file at https://download.copr.fedorainfracloud.org/results/audron/dexed/fedora-38-aarch64/repodata/af12533499469117e3824bc938bfc3208539deb44bbf56ec2612fc383b28a1f4-primary.xml.gz. Supplying the whole file reveals the failure:

$ gzip -tv af12533499469117e3824bc938bfc3208539deb44bbf56ec2612fc383b28a1f4-primary.xml.gz 
af12533499469117e3824bc938bfc3208539deb44bbf56ec2612fc383b28a1f4-primary.xml.gz:         OK
$ file af12533499469117e3824bc938bfc3208539deb44bbf56ec2612fc383b28a1f4-primary.xml.gz 
af12533499469117e3824bc938bfc3208539deb44bbf56ec2612fc383b28a1f4-primary.xml.gz: DOS/MBR boot sector; partition 1 : ID=0x4b, active 0xd0, start-CHS (0x63,239,44), end-CHS (0x33d,134,0), startsector 316955931, 1708404748 sectors; partition 3 : ID=0x53, active 0xaa, start-CHS (0x33a,156,2), end-CHS (0x22c,176,51), startsector 1315648692, 2595702924 sectors; partition 4 : ID=0x3a, active 0xca, start-CHS (0x1d5,97,37), end-CHS (0x28b,241,55), startsector 2173204403, 2875341233 sectors
$ rpm -q file-libs
file-libs-5.44-3.fc39.x86_64

That's a clear failure of file tool (libmagic database).

Please understand that the detection is always only approximate. For createrepo_c, please always keep a suffix in the file name. It's a way how YUM repository metadata delivers the compression algorithm. And it's the only way which avoids the detection.

I recommend closing this createrepo_c issues as we cannot fix it on createrepo_c side.

from createrepo_c.

createrepo_c uses file (magic) library underneath. Hence it's correct that both crearerepo_c and file tool agree. I suspect that's an regression in an update of file-libs RPM package. Could you try an older file-libs version? Or share with us the misdetected archive?

from createrepo_c.

createrepo_c uses file (magic) library underneath. Hence it's correct that both crearerepo_c and file tool agree. I suspect that's an regression in an update of file-libs RPM package. Could you try an older file-libs version? Or share with us the misdetected archive?

from createrepo_c.

Sorry, for the double post.
I reversed your af12533499469117e3824bc938bfc3208539deb44bbf56ec2612fc383b28a1f4 sample into a binary file and I cannot reproduce the misdetection. I tried file-libs-5.41-4.fc36.x86_64 and file-libs-5.44-3.fc39.x86_64. What's your file-libs version?

from createrepo_c.

I'm on Fedora 36 but I believe this can also be reproduced on EL8

from createrepo_c.

I looped in @vmihalko on Slack as the maintainer of the file library

from createrepo_c.