This is a test for a filesystem model aiming for binary compatibility
with FAT32 (specifically, the Linux implementation.) The model relies
on ACL2's single threaded objects (stobjs) in order to slurp FAT32
disk images into memory quickly by placing their contents in stobj
arrays, which provide constant time random access. This is in contrast
to a list representation, which is expensive to construct because
there are as many cons pairs as there are elements in the
list. Another performance improvement comes from the use of
read-file-into-string, which allows us to use a string representation
of the contents of a file, which again allows us to have constant time
random access to the characters within the string. This is in contrast
to the functions described in read-bytes$-n, which use list
representations for the data read from files and suffer from the same
drawbacks described above. In this test, we attempt to test the
ability of our stobj model (and the "cat" implementation which rests
atop it) to reckon with regular files which span across multiple FAT32
clusters. This requires a traversal of the file allocation table,
which contains a linked list of cluster indices for each file.