redirected output out of control

[jc at trillian.mit.edu (John Chambers)]

| Hi BLU,
|
| Maybe this is a newbie error and I've been lucky not to encounter it before,
| but your feedback is appreciated.
|
| I ran a command and unintentionally created an enormous text file that ate up
| all the gigs of open space on the server:
| find -print0 | xargs -0 grep -i blah > blahInstances.txt
|
| I'm puzzled because when I ran the output to the screen, it was only something
| like 1000 lines and it quit like I expected. What's also strange to me is
| that I ran the command on my local suse 9.3 box and it didn't get caught up
| and generate a huge file. It ran as expected. Not sure exactly what OS
| version is on the machine that had the issue.
|
| My main confusion is why running output to the screen would be fine, but
| running it to the file would get caught in a loop or something similar.

What happened was that on that one system, a file  containing  "blah"
was  found  before  the  find  found  blahInstances.txt.   Then  when
blahInstances.txt was finally found, it contained one or  more  lines
that  matched  "blah",  so a new line was added to blahInstances.txt,
and you're now in a loop trying to find the end of a file that  grows
as fast as you read it, because you're copying the data to the end as
you go.

On the suse box, blahInstances.txt  was  scanned  before  "blah"  was
found  anywhere,  so  it  was  empty.  And when you write the data to
stdout, no file is created, so the loop isn't possible.

This is an inherent problem with find.  It's best to make  sure  that
the  output  isn't written into any directory in the search tree.  If
you must, try using  the  -prune  arg  to  skip  the  directory  that
contains the output file. /tmp is always a good place for such files,
and a good directory to prune, since you  never  know  what  sort  of
intermediate file may be there as you pass through.

I've seen people do the same sort of thing by  putting  tar's  output
into the directory being tarred.  I did it myself once.  Oops.


--
   _,
   O   John Chambers
 <:#/> <jc at trillian.mit.edu>
   +   <jc1742 at gmail.com>
  /#\  in Waltham, Massachusetts, USA, Earth
  | |