hi,

i'm surprised and annoyed to find that my backup linux server fails to 
compute a md5sum properly about every 3rd time i check a CD image.  so i 
guess that machine can't add, and i guess i don't really want it as my 
backup server afterall!

but waitaminute, how do i really know it's hardware?  i can imagine a 
software bug that is latent enough to only surface on occational 
machines.  how would you test such a bugger to see what's going on?

and then good grief, my XP machine..  i burned from it, the CD is fine, 
but cygwin sha1sum always fails the iso.  somehow this one doesn't seem 
so likely to be hardware, but how would i know?  what tests would you run?

in an earlier life, before many of you were born i'm sure, working on 
data general clones my (then) company (kurzweil computer products) was 
building, i discovered (guess how) that the hardware could be shown to 
fail only when stressed with certain patterns passing via DMA while 
simultaneously testing memory.  i had to write the diagnostic to prove 
it.  i wasn't immediately popular with our hardware designers, but the 
test became standard in QA.  in the intervening 3 decades it seems i 
have been blessed with working on hardware that just works.  but now.. 
my current life won't afford me the luxury of writing my own diagnostics.

are there some good open source diagnostics that stress the machine 
fully, like in particular stressing disc DMA while simultaneously 
testing memory?

tia,
greg

Greg Whitley Mott
IT Coordinator
NonviolentPeaceforce.org