🚀 go-pugleaf

G'Day Everyone,

I've been asked to provide some statistics to management on disc usage
profiles to help determine if we need to buy more disc or can get by
simply by skimming unused files.

Of course the real answer is buy some more discs, but Pyramid's prices
are unfortunately prohibitive.

For a while I thought I couldn't produce relevant stats because every
file's access time is updated every evening when cpio writes the file
to our will-it-crash-the-system-this-time Gigatape DAT drive.

Then I read about the -a switch to cpio that apparently uses the
O_NOATIME switch on open to leave files' access times alone.

But now I'm back to thinking that I can't produce the relevant stats
because the find command I use to pipe filenames to cpio actually
changes the access time before cpio gets a chance to not change it.

Is there something I'm missing here? An undocumented switch to find
perhaps? Something thats fixed in a later version of OSx than I'm
running (OSx64MNQ coombs 5.1a-901106 21Jan91 Pyramid-9825)?

What should I do? Binary edit find to always do a NO_ATIME open? Wait
a sec! Why would find do an open anyway? Surely it'd just stat(2) each
file and according to the stat(2) man page the st_atime is only
updated on a mknod, utimes, read or write call.

Should I get a PD find and modify that? BSD 4.[34] find is too
difficult to compile on my machine at the moment (but I'm slowly
working on a 'bsd' universe for the 4.4 code :-). Perhaps GNU find?

I think Pyramid's -a switch to cpio is going in the right direction:
backups should not modify a file's access time. Or is my thinking
wrong in this?

Any help appreciated.

Sean
--
------------- Sean Sebastian Batt - sean@coombs.anu.edu.au --------
-------- Coombs Computing Section - Telephone: +61 6 249 3296 -----
-- Australian National University - GPO Box 4 Canberra City 2601 --
-------------------------------------------------------------------