Patrick Welche prlw1 at
Wed Dec 4 19:39:01 GMT 2002

On Wed, Dec 04, 2002 at 07:12:20PM +0000, jra at wrote:
> Ah you see this is the *exact* problem. This is why I ditched the
> code that uses stat.st_blksize. It was causing 64-bit copies on
> IRIX to fail.
> The fundamental misunderstanding (and it was mine also when I wrote
> this code) is that stat.st_blksize is the units of stat.st_blocks.
> It is *NOT*. stat.st_blksize is a "guide" to the "best" I/O size
> for the filesystem. It has nothing to do with stat.st_blocks.
> The unfortunate problem is that neither stat.st_blocks, or
> stat.st_blksize are defined by POSIX so vendors set them as
> they wish.

So is your STAT_ST_BLOCKSIZE the units of stat.st_blocks? In that case

     st_blocks      The actual number of blocks allocated for the file in
                    512-byte units.  As short symbolic links are stored in the
                    inode, this number may be zero.

the ubiquitous 512 should do it for netbsd too...



More information about the samba-technical mailing list