The tests to ext2, ext3, ext4, and ntfs all fail on mips64 (presumably also sparc64). A quick glance at the code shows no attempt to adjust the endianness of constants before comparison. For example this bit from ntfs.c: #define NTFS_FILEMAGIC ((uint32_t)(0x454C4946)) ... if (fr->fr_hdrmagic != NTFS_FILEMAGIC) goto fail;
It's certainly unintentional, but I'm split on whether it's a bug, or a feature :-) Basically, the intent of fstyp is to return filesystem type, when it finds a mountable filesystem. Do we support endianess-independent ext[234] and NTFS? If not, then the filesystems are not mountable.
As implemented it's a bug. We could choose to not compile in support for these file systems on big endian systems, but we shouldn't match bogus byte-swapped filesystems that shouldn't exist and we definitely shouldn't run the tests since they make no sense.
FWIW, ext#fs and NTFS are both pure little-endian filesystems. The fstyp implementation would find totally bogus byte swapped filesystems which should never be supported, but fail to find valid ones.