Shell Programming and Scripting

I am trying to create a script that will report (email and ftp results) on any file, system wide, that has the word password= in it. (and any variation of password=). Basicaly looking for any plain text passwds. I figured the easiest way would be like this......

Code:

for i in `cat find-pw.dat`
 do
 grep -i $i= `find / -type f ` >> out.log
done

But, on Linux I get "grep: /bin/grep: cannot execute [Argument list too long]" and CPU usage hits 99%
on AIX I get "0403-029 There is not enough memory available now."

Is there a more effective cpu/mem friendly way to do this?

Thanks

Try:

Code:

while read line
do
   find / -type f | xargs grep -i "${line}=" >> out.log
done < find-pw.dat

(Deliberately not posting solution).

I got this to work, however it takes forever to run. This isn't what it sounds like methyl

Code:

for i in `cat find-pw.dat`
 do
 find / -type f -exec grep $i= {} \; >> out.log
done

Thanks anurag your versions gives me better output to sort through.

What process produced the file find-pw.dat ?


Btw. I have no idea where the dubious construct
"for i in open_ended_selection"
comes from. The only place I have ever seen it is this board and always from posters in India. Is this construct taught in some commonly available training course notes in India?
The construct has no use whatsoever in the commercial world.

find-pw.dat is a manually created file containing every possible spelling of the word.
Not sure what they teach in india, I was taught that many years ago when I started in Unix by an old white guy. I disagree it has no use whatsoever. I have worked for several fortune 500 companies and I've seen this at all of them. It works great, so I'm not sure why you seem to despise it.

Next I need to create some filters on the output, you would not beleive how many hits on the word passw* this returns.

I think this will be easier in perl.

Mainly because it does not correctly deal with input lines containing space characters and it is can easily break the maximum command line length in common Operating Systems such as IBM's AIX.
As a troubleshooter since early unix I found that script authors using "for" with open-ended lists proved to be the most common scripting mistake. I've even seen it in manufacturer-supplied cleanup crons in Unix SV R3 .

I have no problem for with finite lists with the values in double quotes. The issue is with open-ended lists.


Btw. Your script would run a lot faster if you test the result of the unix "file" command first to eliminate non-ascii files which are unsuitable for a character string "grep". I have used this technique to search over 100,000 files where only about 2,000 were suitable for searching.
Afterthought: The matching process would be a lot quicker with "egrep" on a file-by-file basis.