Gather disk space used in addition to file size

Conversely, a multi-GB file consisting only of zeroes might be stored in only a couple of blocks due to sparse files (http://www.lrdev.com/lr/unix/sparsefile.html) or other types of compression (such as the LZ77 compression provided by NTFS.)

A naive approach to finding the actual disk usage of a file is to use stats(2)'s st_blksize*st_blocks. This is what du uses to determine disk usage. However, this does not account for the capability of some modern file systems (e.g. Reiser4) to handle small files efficiently by putting the data of multiple small files into a single block, or rather by doing away with block alignment requirements altogether.

Quoting http://www.namesys.com:

"Reiser4 uses dancing trees (...). This makes Reiser4 more space efficient than other filesystems because we squish small files together rather than wasting space due to block alignment like they do."

It turns out that st_blksize*st_blocks is not the correct formula to calculate disk usage of a file. According to the stat(2) man page on both OS X 10.4 and a somewhat old Linux distribution (glibc 2.3.2), st_blksize is not to be used as the scale of st_blocks; rather, st_blocks is given in 512-byte blocks regardless of the actual block size of the file system. (st_blksize doesn't necessarily give the actual block size of the file system either; rather, it's the "recommended block size for optimal I/O.")

In summary, the disk usage of a file should be calculated as st_blocks * 512.

Can I close this ticket now?

Earth meanwhile stores disk space usage so closing this ticket.